The x-ray flat panel detector (FPD) is a key component of the next generation x-ray imaging systems which promote digitization of x-ray images. By developing FPD applicable to both fluoroscopy and radiography, it is expected that x-ray diagnostic systems will change dramatically not only in terms of performance, but also in shape and form. The purpose of this research is to develop a selenium-based FPD applicable to both radiography and fluoroscopy. This report presents the results of the work on new technology which can be applied to create clinically-useful view size detector. The prototype detector adopts 1000 micron thick selenium as the photoconductor, 23 cm square of field of view, 150 X 150 micron pixel pitch, 1536 X 1536 pixel number, and is capable of capturing images at up to 30 frames/second. The features of this prototype detector are: simple-structure selenium suitable for real-time readout, and a high voltage protection structure for the TFT array which acts at x-ray overexposure. To realize large field of view, the field uniformity performance of the selenium and TFT array has been improved, and noise from the TFT array has been minimized. In this paper, new physical performance related to MTF, input- and-output characteristics, image lag, blooming, etc., are discussed.

This paper presents results of an image quality evaluation performed on a prototype DirectRay system (Sterling Diagnostic Imaging, Direct Radiography Corp.) operating in a UK radiology department. Physical imaging characteristics measured included characteristic (sensitometric) response, square-wave response function, Wiener spectrum and detective quantum efficiency (DQE). In addition, subjective image quality was assessed using a calibrated threshold contrast detail detectability (TCDD) test object. Results produced by the DirectRay system were compared with those achieved by a modern photo-stimulable computed radiography (CR) system. DirectRay and CR systems exhibit similarly wide exposure dynamic ranges, but in practice exploit this property in differing ways. Measured values of DQE(0) for the DirectRay system were found to be higher compared to those obtained for CR, for the majority of the exposure range examined. These results were confirmed subjectively by TCDD test-object analysis. No evidence of a 'memory artifact' phenomenon was found.

The x-ray imaging performance is reported using polycrystalline lead iodide as a thick semiconductor detector on an active matrix flat panel array. We have developed a test image sensor with 100 micron pixel size in a 512 X 512 format, using amorphous silicon TFTs for matrix addressing. The new 14 bit electronic system allows radiographic and fluoroscopic x-ray imaging. PbI₂ has larger x-ray absorption and higher charge generation efficiency than selenium, and has the potential for higher sensitivity imaging. The films are deposited by vacuum sublimation and have been grown thicker than 100 micrometer. Measurements of the carrier transport and charge collection, together with modeling studies show how the x-ray sensitivity depends on the material properties. Imaging measurements find excellent spatial resolution and confirm models of the x-ray sensitivity. Both radiographic and fluoroscopic imaging are demonstrated. While good overall imaging is obtained, the dark leakage current and image lag need further improvement.

A preliminary investigation on the temperature stability of selenium alloys has been carried out to determine their performance limitations in terms of properties relevant to x- ray imaging applications. The selenium properties tested as a function of temperature were the rate of crystallization, dark current, x-ray sensitivity and the residual signal. Results from crystal growth data show that the lifetime of conventional selenium can be up to 30 years of operation at 40 degrees Celsius but only about 3 years at 50 degrees Celsius. However, selenium alloys produced with a new process exhibited improved thermal stability by a factor of more than 2, with expected lifetime of up to 7 years at 50 degrees Celsius. Long term temperature cycling at 50 degrees Celsius show that the dark current remained below 50 pA/cm² for the test period of about 100 operational days but was projected to remain below 100 pA/cm² for up to 10 operational years. Testing under operational conditions showed a slight but reversible increase in the dark current occurred at 50 and 60 degrees Celsius, accompanied by a corresponding reversible decrease in the x-ray sensitivity in this temperature range. The loss in the x-ray sensitivity however, could be recovered by suitably increasing the operating electric field.

A new technique called Variable-Resolution X-ray (VRX) detection that dramatically increases the spatial resolution in computed tomography (CT) and digital radiography (DR) is presented. The technique is based on a principle called 'projective compression' that allows the resolution element of a CT detector to scale with the subject or field size. For very large (40 - 50 cm) field sizes, resolution exceeding 2 cy/mm is possible and for very small fields, microscopy is attainable with resolution exceeding 100 cy/mm. Several effects that could limit the performance of VRX detectors are considered. Experimental measurements on a 16-channel, CdWO₄ scintillator + photodiode test array yield a limiting MTF of 64 cy/mm (8(mu) ) in the highest-resolution configuration reported. Preliminary CT images have been made of small anatomical specimens and small animals using a storage phosphor screen in the VRX mode. Measured detector resolution of the CT projection data exceeds 20 cy/mm (less than 25 (mu) ); however, the final, reconstructed CT images produced thus far exhibit 10 cy/mm (50 (mu) ) resolution because of non-flatness of the storage phosphor plates, focal spot effects and the use of a rudimentary CT reconstruction algorithm. A 576-channel solid-state detector is being fabricated that is expected to achieve CT image resolution in excess of that of the 26-channel test array.

Large area, flat panel solid state detectors are being investigated for digital radiography and fluoroscopy. These detectors employ an x-ray imaging layer of either photoconductor ('direct' conversion method) or phosphor ('indirect' conversion method) to detect x-rays. In both cases the image formed at the surface of the layer is read out in situ using an active matrix array. Depending upon the resolution of the layer compared to the pixel size, undersampling of the image and hence aliasing may occur. Aliasing is always present regardless of the pixel size in direct detectors based on amorphous selenium because of its high intrinsic resolution. Aliasing gives rise to increased noise which results in reduction of detective quantum efficiency DQE at high spatial frequencies. The aliasing can be reduced or even eliminated by blurring prior to pixel sampling (e.g., by scattering in a phosphor layer). However, blurring, which may be quantified by the spatial frequency f dependent modulation transfer function MTF(f), also has a deleterious effect: the imaging system becomes much more susceptible to noise for example that arising in the charge amplifiers or secondary quantum statistics. Note that in principle, the system MTF can be corrected to any desired values in a digital system thus MTF has no predictive value for the quality of an imaging system, rather it is the DQE(f) which determines the overall signal to noise ratio independently of the MTF enhancement chosen. Nevertheless, determining the ideal level of presampling blurring (i.e., the Presampling Modulation Transfer function) is not straightforward. A problem caused by blurring is that the degree of blurring often depends on the depth of absorption of the x-ray in the imaging layer. In such cases (as pointed out by Lubberts) additional noise is transferred to the image. The predictions of a Lubberts model will be compared with published measurements of DQE for both direct and indirect detectors. A preliminary conclusion, is that blurring by CsI phosphor layers is non-ideal and leads to a significant loss of DQE at high spatial frequencies while no such loss is occurring in (alpha) -Se layers due to the equal MTF (in this case MTF(f) approximately equals 1) at all depths. Thus the only method which appears practical to cause blurring so as to avoid noise aliasing while avoiding the Lubberts depth dependent effect is to have a perfect MTF in the imaging layer and then blur before sampling. Such an approach has been proposed for the direct method based on the use of a partially conducting layer. Theoretical estimates of the final DQE(f) to be expected using this variation of the direct conversion method are produced.

The objective of this paper is to analyze quantitatively and systematically the major electronic noise source and provide design guidelines to improve signal to noise ratio in large area flat panel x-ray imaging systems. A transmission line model combined with a thin-film transistor model and transfer functions of charge-amplifier and correlated-double sampling is employed to simulate the electronic noises arising from the external amplifiers, data lines, gate lines and pixels. Simulation results using simple discrete RC models are presented for comparison. The noise analysis method and noise formula presented will provide guidelines to achieve the goal of optimization in imaging performance and quantum noise limited operation of the detector.

Offset and gain corrections are indispensable to exploit images from large image sensors, because of the pixel to pixel variation in dark current and sensitivity. However, an inappropriate correction may be detrimental to the signal to noise ratio of the raw image. This is especially critical in X-ray imaging, where the quantum noise is filtered by the detector spatial response. The noise power spectrum (NPS) in the corrected image is a combination of the initial noise spectrum in the raw image (quantum noise and electronic noise) with the noise in the offset and gain images. The dark image noise power just adds up to the noise power in the current image. The noise in the gain image alters the noise of the current image in a more intricate way. This is illustrated by simulations and experimental measurements. The conditions to safeguard the signal to noise ratio in the current image are detailed: There is an optimal number of dark and 'white' images to be averaged in order to keep their electronic and quantum noise negligible compared to that of the current image. Real conditions often force trade-offs between the desirable large number of offset/gain images to be averaged and the time effectively assigned to such acquisitions. Furthermore, the residual noise spectrum in the gain images is dependent on dose, uniformity of irradiation, temperature and detector spatial response. In the appropriate conditions, the intrinsic signal to noise ratio of an image can be preserved by offset and gain correction. Nevertheless, at high dose, the gain correction unavoidably introduces some high frequency proportional noise which degrades the DQE.

We seek to optimize a SPECT brain-imaging system for the task of detecting a small tumor located at random in the brain. To do so, we have created a computer model. The model includes three-dimensional, digital brain phantoms which can be quickly modified to simulate multiple patients. The phantoms are then projected geometrically through multiple pinholes. Our figure of merit is the Hotelling trace, a measure of detectability by the ideal linear observer. The Hotelling trace allows us to quantitatively measure a system's ability to perform a specific task. Because the Hotelling trace requires a large number of samples, we reduce the dimensionality of our images using Laguerre-Gauss functions as channels. To illustrate our method, we compare a system built from small high-resolution cameras to one utilizing larger, low-resolution cameras.

We show how high-resolution parallel-projection images can be obtained with a synthetic collimator, a device that consists of a multiple-pinhole aperture with detector arrays in several locations behind it. Simulation studies indicate that this system is superior to a parallel-hole collimator system for parallel-projection imaging. We also show how three- dimensional reconstructions can be created with the synthetic collimator and present evidence from simulations that lesion detection will be better in these reconstructions than in either of the parallel-projection imaging systems.

The objective imaging characteristics of a new high-speed mammographic screen-film combination is analyzed in detail. By combining a large reduction in film granularity with substantial improvements in screen technology, high-speed mammography can now be re-optimized to yield improvements in imaging system performance. Optimal use of the materials for reducing patient motion and for magnification techniques is potentially more effective than previous high-speed systems.

Since the video monitor is widely believed to be the weak link in the imaging chain, it is critical, to include it in the total image quality evaluation. Yet, most physical measurements of mammographic image quality are presently limited to making measurements on the digital matrix, not the displayed image. A method is described to quantitatively measure image quality of mammographic monitors using ACR phantom-based test patterns. The image of the test pattern is digitized using a charge coupled device (CCD) camera, and the resulting image file is analyzed by an existing phantom analysis method (Computer Analysis of Mammography Phantom Images, CAMPI). The new method is called CCD-CAMPI and it yields the Signal-to-Noise-Ratio (SNR) for an arbitrary target shape (e.g., speck, mass or fiber). In this work we show the feasibility of this idea for speck targets. Also performed were physical image quality characterization of the monitor (so-called Fourier measures) and analysis by another template matching method due to Tapiovaara and Wagner (TW) which is closely related to CAMPI. The methods were applied to a MegaScan monitor. Test patterns containing a complete speck group superposed on a noiseless background were displayed on the monitor and a series of CCD images were acquired. These images were subjected to CCD-CAMPI and TW analyses. It was found that the SNR values for the CCD-CAMPI method tracked those of the TW method, although the latter measurements were considerably less precise. The TW SNR measure was also about 25% larger than the CCD-CAMPI determination. These differences could be understood from the manner in which the two methods evaluate the noise. Overall accuracy of the CAMPI SNR determination was 4.1% for single images when expressed as a coefficient of variance. While the SNR measures are predictable from the Fourier measures the number of images and effort required is prohibitive and it is not suited to Quality Control (QC). Unlike the Fourier measures and the TW method, CCD-CAMPI is capable of yielding speck SNR on a single image. This is based on preliminary work and more complete testing is underway. Based on the early promising results, we expect that the CCD-CAMI method can be adapted to routine image QC of monitors using inexpensive equipment.

The detection characteristics of digital x-ray and film-screen mammography systems are different and thus current film-screen techniques are not ideal for digital mammography. Therefore optimum technical parameters required for digital mammography are likely to be different compared with film-screen mammography. The goal of this study is to evaluate the optimum technical parameters for full-field digital mammography by experimental and computer simulation methods. A General Electric Full Field Digital Mammography (FFDM) prototype unit using Cesium Iodide (CsI) on an amorphous Silicon photodiode array was used for the experimental measurements. Using breast equivalent phantoms, images were acquired for a set of x-ray target-filters for a range of peak kilovoltage, varying breast composition and thickness, with and without an anti-scatter grid. The signal-to-noise ratio (SNR) and figure-of-merit (FOM) were determined for simulated calcification and mass targets, independently by the two methods. The results for noise, contrast, SNR and FOM were compared and agree within 5% and 6% respectively. Combined results are presented for the case of 50% glandular - 50% adipose tissue breast composition using the grid and for the calcification target. Based on the FOM approach, preliminary results suggest that a Rhodium target-filter combination will be beneficial for higher breast thickness and for denser breasts.

The detection characteristics of digital x-ray and film-screen mammography systems are different and thus current film-screen techniques are not ideal for digital mammography. Therefore optimum technical parameters required for digital mammography are likely to be different compared with film-screen mammography. The goal of this study is to evaluate the optimum technical parameters for full-field digital mammography by experimental and computer simulation methods. A General Electric Full Field Digital Mammography (FFDM) prototype unit using Cesium Iodide (CsI) on an amorphous Silicon photodiode array was used for the experimental measurements. Using breast equivalent phantoms, images were acquired for a set of x-ray target-filters for a range of peak kilovoltage, varying breast composition and thickness, with and without an anti-scatter grid. The signal-to-noise ratio (SNR) and figure-of-merit (FOM) were determined for simulated calcification and mass targets, independently by the two methods. The results for noise, contrast, SNR and FOM were compared and agree within 5% and 6% respectively. Combined results are presented for the case of 50% glandular - 50% adipose tissue breast composition using the grid and for the calcification target. Based on the FOM approach, preliminary results suggest that a Rhodium target-filter combination will be beneficial for higher breast thickness and for denser breasts.

Monte Carlo calculations were performed to generate the point spread functions of x-ray photons absorbed in a CsI:Tl x-ray detector at the x-ray energies normally used in mammography (i.e., 20 keV to 50 keV). The corresponding modulation transfer functions [MTF(f)] for the CsI:Tl screen were also computed, taking into account the optical spread of light within the CsI:Tl crystals. The computed MTF(f)s were dominated by scintillation light lateral dispersion within the CsI:Tl screen. For the photon energy range encountered in digital mammography, the MTF(f) was a minimum at an x-ray photon energy just above the k-edge of Iodine (33 keV). Noise propagation theory for a cascaded imaging system was subsequently used to derive a theoretical expression for the detector DQE(f), including the dependence of DQE(f) on the spatial distribution of x-ray photon energy deposition. Detector performance was investigated as a function of x-ray exposure, CCD electronic noise, coupling efficiency of the fiber optical coupler, and the CCD quantum efficiency. Although most of the x-rays are absorbed via the photoelectric effect, the deposited x-ray energy spread within the CsI:Tl screen from the emission of characteristic x-rays can have a marked effect on detector performance, and the DQE(f) was found to decrease rapidly with photon energy just above the Iodine K-edge. X-ray exposure levels to the detector should be greater than or equal to 5 mR with a CCD electronic noise of approximately 20 electrons rms to ensure that DQE(f) performance is not significantly degraded at the spatial frequencies important in digital mammography (i.e., 0 to 10 lp/mm). Light dispersion within the CsI:Tl crystals was the major factor degrading imaging system DQE(f) at higher spatial frequencies. Optical coupling efficiency and CCD quantum efficiency are important system design parameters, which need to be maintained at a relatively high value. An optical coupling efficiency of approximately 0.7, and a CCD quantum efficiency of approximately 0.4, would still permit system DQE values greater than 60% at a spatial frequency of 5 lp/mm.

An x-ray camera for imaging with a spatial resolution in the micrometer and sub-micrometer range has been developed. The camera consists of a scintillator, light microscopy optics and a cooled charged-coupled device (CCD). A transparent scintillator converts the x-ray field into a visible light image which is projected onto the CCD by the light optics. A resolution of 0.8 micrometer fwhm was achieved using 12-keV x rays and a 5-micrometer thick commercially available Y₃Al₅O₁₂:Ce (YAG:Ce) scintillator. The detective quantum efficiency (DQE) of the camera is mainly limited by the low absorption of x rays in the thin layer of the scintillator. To increase the absorption, Lu₃Al₅O₁₂ (LAG) scintillators have been grown by liquid phase epitaxy (LPE). The characteristics of LAG activated with Eu or Tb have been investigated, in particular spectral emission, efficiency of x-ray to light conversion, and time response. Three-dimensional computed x-ray microtomography (3D-CMT) images of mouse bone samples have been recorded with this camera using a 5-micrometer thick YAG:Ce screen. The 3D-CMT system uses a parallel monochromatic x-ray beam extracted from the synchrotron radiation. A series of 2D x-ray projection images at different angles were recorded, and processed numerically to yield the 3D image of the bone structure with a voxel size of 1.8 micrometer. Image features smaller than 3 micrometer are clearly visible in the reconstructed tomogram. The quality of the images allows the analysis of the trabecular bone structure that is important for the understanding of the mechanisms of osteoporosis.

This paper presents a calibration and correction method for detector cell gain variations. To provide variable slice thickness capability in multislice volumetric scanners, while minimizing costs, it is necessary to combine the signals from several detector cells. The process of combining the output of several detector cells with non-uniform gains can introduce numerical errors when the impinging x-ray signal varies over the range of the combined cells. These scan dependent numerical errors can lead to artifacts in the reconstructed images, particularly when the numerical errors vary from channel-to-channel. A projection data correction algorithm has been developed to subtract the associated numerical errors. For effectiveness and data flow reasons, the algorithm works on a slice-by-slice basis. An initial error vector is calculated by applying a high-pass filter to the projection data. The essence of the algorithm is to correlate that initial error vector, with a calibration vector obtained by applying the same high-pass filter to various z-combinations of the cell gains. The solution to the least-square problem gives the coefficients of a polynomial expansion of the signal z-slope and curvature. From this information, and given the cell gains, the final error vector is calculated and subtracted from the projection data.

Recent development of large area flat panel solid state detector arrays indicates that flat panel image sensors have some common potential advantages: compactness, absence of geometric distortion and veiling glare with the benefits of high resolution, high DQE, high frame rate and high dynamic range, small image lag (less than 1%) and excellent linearity (approximately 1%). The advantages of the new flat-panel detector make it a promising candidate for cone beam volume CT imaging. The purpose of this study is to characterize an amorphous silicon thin film transistor (STFT) flat panel detector-based imaging system for cone beam volume CT applications. A prototype STFT detector-based cone beam volume CT imaging system has been designed and constructed based on the modification of a GE 8800 CT scanner. This system is evaluated using two phantoms: a low contrast phantom, and a GE CT quality assurance phantom. The results indicate that the flat panel detector-based volume cone beam can achieve a better spatial resolution than that of a standard fan beam CT and a low contrast resolution approaching that of a standard fan beam CT. The results also suggest that to use a flat panel detector for cone beam Volume CT, the design and construction of the detector should be optimized for tomographic acquisition and reconstruction.

To characterize the performance of a cone-beam computed tomography (CBCT) imaging system based upon an indirect- detection, amorphous silicon flat-panel imager (FPI). Tomographic images obtained using the FPI are presented, and the signal and noise characteristics of reconstructed images are quantified. Specifically, the spatial uniformity, CT linearity, contrast performance, noise characteristics, spatial resolution, and soft-tissue visualization are examined. Finally, the performance of the FPI-based CT system is discussed in relation to existing clinical technologies. A table-top measurements system was constructed to allow investigation of FPI performance in CBCT within a precisely controlled and reproducible geometry. The FPI incorporates a 512 X 512 active matrix array of a-Si:H thin-film transistors and photodiodes in combination with an overlying (133 mg/cm² Gd₂O₂S:Tb) phosphor. The commercially available prototype FPI has a pixel pitch of 400 micrometer, a fill factor of approximately 80%, can be read at a maximum frame rate of 5 fps, and provides 16 bit digitization. Mounted upon an optical bench are the x-ray tube (in a rigid support frame), the object to be imaged (upon a precision rotation/translation table), and the FPI (mounted upon a precision translation table). The entire setup is directed under computer control, and volumetric imaging is accomplished by rotating the object incrementally over 360 degrees, delivering a radiographic x-ray pulse (e.g., 100 - 130 kVp, approximately 0.1 - 10 mAs), and acquiring a projection image at each increment. Prior to reconstruction, dark and flood- field corrections are applied to account for stationary nonuniformities in detector response and dark current. Tomographic images are reconstructed from the projections using the Feldkamp filtered back-projection algorithm for CBCT. The linearity of the CBCT system was compared to that of a commercial scanner (Philips SR-7000) using materials ranging in CT number from approximately 900 to 1100. The contrast sensitivity of the CBCT system and the conventional scanner was compared using these same materials. Images of a uniform water bath were acquired for characterization of the response uniformity and the dependence of noise on exposure. The spatial frequency response characteristics of the system were measured using a steel wire, from which the point spread function and modulation transfer function were determined. Finally, the soft-tissue contrast and spatial resolution of the CBCT system was demonstrated in volumetric images of a euthanized rat. The image quality was compared to images of the same subject acquired with an equivalent technique on the commercial scanner. A table-top CBCT scanner based upon an a- Si:H FPI has been constructed, and a system for CBCT image acquisition, processing, and reconstruction has been implemented. This system is capable of producing high-quality volumetric images. Reconstructions were generated from 300 radiographs (100 kVp; 1 mAs per projection) obtained at 1.2 degree increments through 360 degrees. Image acquisition and reconstruction required approximately 30 min and approximately 2 h 20 min (250 MHz UltraSparc), respectively. The system has demonstrated signal and noise performance comparable to that of commercial CT scanners. The imaging performance of the prototype supports the hypothesis that FPIs can be employed in computed tomography applications.

In this paper, we report our investigation using a CsI(Tl) transparent scintillator x-ray detector together with the multi-level scheme algebraic reconstruction technique (MLS- ART) for megavoltage computed tomography (CT) reconstructions. The reconstructed CT images may be useful for positional verification in radiotherapy. The CsI(Tl) imaging system consists of a scintillator screen coupled to a liquid- nitrogen-cooled slow-scan CCD-TV camera. This system provides better contrast resolution than the standard electronic portal imaging system (EPID), which is especially useful given the low number of projections we are aiming at. The geometry of the imaging system has also been optimized to achieve high spatial resolution (1 lp/mm) in spite of the thickness of the screen. We present the images reconstructed using a pediatric head phantom using a total of 99 projections, and a combined phantom using 50 projections. Image reconstruction was carried out using the MLS-ART technique. We also present the CT images obtained using the back projection technique for comparison purposes.

Ultrasonography, a widely used imaging modality for the diagnosis and staging of many diseases, is an important cost- effective technique, however, technical improvements are necessary to realize its full potential. Two-dimensional viewing of 3D anatomy, using conventional ultrasonography, limits our ability to quantify and visualize most diseases, causing, in part, the reported variability in diagnosis and ultrasound guided therapy and surgery. This occurs because conventional ultrasound images are 2D, yet the anatomy is 3D; hence the diagnostician must integrate multiple images in his mind. This practice is inefficient, and may lead to operator variability and incorrect diagnoses. In addition, the 2D ultrasound image represents a single thin plane at some arbitrary angle in the body. It is difficult to localize and reproduce the image plane subsequently, making conventional ultrasonography unsatisfactory for follow-up studies and for monitoring therapy. Our efforts have focused on overcoming these deficiencies by developing 3D ultrasound imaging techniques that can acquire B-mode, color Doppler and power Doppler images. An inexpensive desktop computer is used to reconstruct the information in 3D, and then is also used for interactive viewing of the 3D images. We have used 3D ultrasound images for the diagnosis of prostate cancer, carotid disease, breast cancer and liver disease and for applications in obstetrics and gynecology. In addition, we have also used 3D ultrasonography for image-guided minimally invasive therapeutic applications of the prostate such as cryotherapy and brachytherapy.

Elasticity imaging is an emerging diagnostic modality whose development now is largely empirical. This paper outlines a framework for designing ultrasonic bioelasticity imaging systems based on a maximum-likelihood estimator of tissue motion. A principal goal for image formation is to maximize waveform coherence, which is to correctly match Fourier coefficients of the echo data recorded before deformation to those recorded after deformation. A crosstalk matrix is developed for strain imaging to obtain new insights into the physics of elasticity imaging, in particular resolution, and a simple figure of merit for evaluating system designs.

Ultrasound is an attractive modality for adjunctive characterization of certain breast lesions, but it is not considered specific for cancer and it is not recommended for screening. An imaging technique remarkably different from pulse-echo ultrasound, termed Optical Sonography^TM (Advanced Diagnostics, Inc.), uses the through-transmission signal. The method was applied to breast examinations in 41 asymptomatic and symptomatic women ranging in age from 18 to 83 years to evaluate this imaging modality for detection and characterization of breast disease and normal tissue. This approach uses coherent sound and coherent light to produce real-time, large field-of-view images with pronounced edge definition in soft tissues of the body. The system patient interface was modified to improve coupling to the breast and bring the chest wall to within 3 cm of the sound beam. System resolution (full width half maximum of the line-spread function) was 0.5 mm for a swept-frequency beam centered at 2.7 MHz. Resolution degrades slightly in the periphery of the very large 15.2-cm field of view. Dynamic range of the reconstructed 'raw' images (no post processing) was 3000:1. Included in the study population were women with dense parenchyma, palpable ductal carcinoma in situ with negative mammography, superficial and deep fibroadenomas, and calcifications. Successful breast imaging was performed in 40 of 41 women. These images were then compared with images generated using conventional X-ray mammography and pulse-echo ultrasound. Margins of lesions and internal textures were particularly well defined and provided substantial contrast to fatty and dense parenchyma. In two malignant lesions, Optical Sonography^TM appeared to approximate more closely tumor extent compared to mammography than pulse-echo sonography. These preliminary studies indicate the method has unique potential for detecting, differentiating, and guiding the biopsy of breast lesions using real-time acoustical holography.

An advanced Scanning-Beam Digital X-ray (SBDX) system for cardiac angiography has been constructed. The 15-kW source operates at 70 - 120 kVp and has an electron beam that is electromagnetically scanned across a 23-cm X 23-cm transmission target. The target is directly liquid cooled for continuous full-power operation and is located behind a focused source collimator. The collimator is a rectangular grid of 100 X 100 apertures whose axes are aligned with the center of the detector array. X-ray beam divergence through the collimator apertures is matched to the 5.4-cm X 5.4 cm detector, which is 150 cm from the source. The detector is a 48 X 48 element CdZnTe direct-conversion photon-counting detector. A narrow x-ray beam scans the full field of view at up to 30 frames per second. A custom digital processor simultaneously reconstructs sixteen 1,000² pixel tomographic images in real time. The slices are spaced 1.2 cm apart and cover the entire cardiac anatomy. The small detector area and large patient-detector distance result in negligible detected x-ray scatter. Image signal-to-noise ratio is calculated to be equal to conventional fluoroscopic systems at only 12% of the patient exposure and 25% of the staff exposure. Exposure reduction is achieved by elimination of detected scatter, elimination of the anti-scatter grid, increased detector DQE, and increased patient entrance area.

This investigation defines and tests a simple, nonlinear, task-specific method for rapid tomosynthetic reconstruction of radiographic images designed to allow an increase in specificity at the expense of sensitivity. Representative lumpectomy specimens containing cancer from human breasts were radiographed with a digital mammographic machine. Resulting projective data were processed to yield a series of tomosynthetic slices distributed throughout the breast. Five board-certified radiologists compared tomographic displays of these tissues processed both linearly (control) and nonlinearly (test) and ranked them in terms of their perceived interpretability. In another task, a different set of nine observers estimated the relative depths of six holes bored in a solid Lucite block as perceived when observed in three dimensions as a tomosynthesized series of test and control slices. All participants preferred the nonlinearly generated tomosynthetic mammograms to those produced conventionally, with or without subsequent deblurring by means of iterative deconvolution. The result was similar (p less than 0.015) when the hole-depth experiment was performed objectively. We therefore conclude for certain tasks that are unduly compromised by tomosynthetic blurring, the nonlinear tomosynthetic reconstruction method described here may improve diagnostic performance with a negligible increase in cost or complexity.

The paper investigates the variability of the geometrical deformation in digital X-ray image intensifier images as a function of acquisition parameters and time. We have increased our understanding of the deformation patterns and of how they are influenced by normal clinical use. Our goal is to reduce the frequency of calibration and the complexity of the protocol for diagnostic and interventional procedures so that geometrical distortion correction can become part of daily clinical practice.

Routine mammographic screening has proven to be the only reliable diagnostic method for early detection of breast cancers. The introduction of digital mammographic imaging to clinical practice should provide significant improvement in image quality and lead to better detection sensitivity. One limitation of some recently developed indirect digital imagers is a small fill factor in the photodetector. A small fill factor reduces the amount of light collected from individual x-ray interactions in the screen and can lead to an overall reduction in the number of x-ray quanta detected. At the SPIE meeting last year, we suggested using a focused microlens array between the florescent screen and photodetector as a method for improving x-ray utilization in indirect digital mammographic imaging systems. This year we have conducted a Monte Carlo simulation of a uniform field x-ray source and a digital imager. We have used this simulation tool to compare the proposed imager, which employed microlens focusing, with more standard imager configurations. The evaluation included light collection efficiency and x-ray utilization performance measure. The results indicate that including microlens focusing between the scintillation screen and the photodetector can improve light collection and increase the number of x-ray quanta detected.

This paper proposes a realistic joint energy-spatial description of the various phenomena encountered in low energy gamma imagery: Compton diffusion, photoelectric absorption, collimator and detector behavior, shot noise affecting the images. This physical model takes into account the full energy dimension of the scattered photons, and can construct a series of images indexed in energy. It is based on the last generation of SPECT (single photon emission computed tomography) medical gamma cameras which operate in list mode (both the coordinates and the energy of the (gamma) photons are measured), in the energy range [50 - 500 keV]. In the model proposed, two hypotheses are made: (1) only the first order scattering is considered, (2) the diffusing medium is homogeneous. This model is computationally light and could lead to the development of new enhancing techniques using the energy dimension.

Following our previous report concerning the development of a 127 micrometer resolution, 7.4 million pixel, 30 X 40 cm² active area, flat panel amorphous Silicon (a-Si) x-ray image sensor, this paper describes enhancements in image sensor performance in the areas of image lag, linearity, sensitivity, and electronic noise. New process improvements in fabricating a-Si thin film transistor (TFT)/photodiode arrays have reduced first-frame image lag to less than 2%, and uniformity in photoresponse to less than 5% over the entire 30 X 40 cm² active area. Detailed analysis of image lag vs. time and x-ray dose will be discussed. An improved charge amplifier has been introduced to suppress image cross-talk artifacts caused by charge amplifier saturation, and system linearity has been optimized to eliminate banding effects among charge amplifiers. Preliminary sensitivity improvements through the deposition of CsI(Tl) directly on the arrays are reported, as well as overall imaging characteristics of this improved image sensor.

A novel solid state detector has been developed for digital radiography. It is based on an assembly of CCD chips operated in TDI mode. The X-ray detection is achieved by a CsI scintillator layer which features a needle structure. It combines an excellent resolution with a high absorption efficiency. The detector is able to operate in 2 modes: a standard resolution mode and a high resolution mode. In the standard mode, the detector features a 440 mm long sensitive area and a 162 micrometer pixel size. The total number of pixels along detector direction is 2720. Along the scan direction, the detector features a 10.8 mm wide sensitive area. A TDI process is performed in the chip itself over 67 X- ray sensitive elements. A 440 mm X 440 mm sized image can be acquired in typically 1.2 second. The high resolution mode allows producing images with an 81 micrometer pixel size. The detection dynamic range is 10,000:1. The main performance characteristics of the detector are presented, such as DQE, MTF, dynamic range.

Advanced technical investigations, including DQE measurements and threshold contrast detail-detectability experiments, have been performed in order to demonstrate the superior image quality of an experimental flat dynamic X-ray image detector (FDXD) system. The dose efficiencies throughout a range of dose levels used in fluoroscopic and radiographic applications have been measured and are presented. Together with the results of a range of clinical patient examinations, the results of the technical investigations fully confirm earlier expectations in terms of increased image quality and improved dose efficiency with respect to current imaging modalities. Several mixed applications performed with the FDXD system are presented including those where subtraction techniques were used. The dynamic aspects of the FDXD system are discussed in detail. In the fluoroscopic mode, images have been acquired with a dose-rate as low as 5 nGy per image using a frame rate of approx. 25 fps. Low dose fluoroscopic images will be presented and it will be confirmed that low readout noise of the detection system facilitates the clinical acceptability of the images, even without applying any noise reduction algorithms. Post-processing algorithms for exposures will also be discussed. It can be concluded that the results of the technical measurements, together with the clinical examinations, prove that in areas regarding dose efficiency and image quality, this new detector technology is superior to the current X-ray modalities in many aspects.

A timing sequence for operation of an amorphous silicon detector within a digital chest radiography system is evaluated. Measurements of large-area spatial non-uniformities (i.e., artifacts) resulting from various implementations of the sequence are presented. Based on these measurements, a method for operation is determined which successfully eliminates large-area artifacts and which satisfies several important system latency requirements. In addition, a method for quantifying artifacts of the type observed in these experiments is presented.

Conventional radiography is based on absorption contrast and geometrical (ray) optics. After an outline of the relevant theory, this article reports results displaying both phase- and absorption-contrast, collected with a technique which utilizes a micro-focus x-ray source to achieve a high degree of spatial coherence, and relatively large object-to-image distances to enable (wave) interference effects (Fresnel diffraction) to occur and manifest themselves as phase contrast in the image plane. Both soft tissue (chicken knee) and hard tissue (finger bone) samples are investigated for a range of source sizes and object-to-image distances, encompassing conditions somewhat analogous to conventional radiography. Variation in image contrast and resolution as a function of these variables is observed and discussed.

We report the results of quantitative hard X-ray phase- contrast microscopy and tomography using synchrotron radiation, in-line imaging geometry and a non-interferometric phase retrieval technique based on the Transport of Intensity equation. This quantitative imaging method is fast, simple, robust, does not require sophisticated X-ray optical elements and can potentially provide submicron spatial resolution over a field of view of the order of centimeters. In the present experiment a spatial resolution of approximately 0.8 micron has been achieved in images of a polystyrene sphere using 19.6 keV X-rays. We demonstrate that appropriate processing of phase-contrast images obtained in the in-line geometry can reveal important new information about the internal structure of weakly absorbing organic samples. We present some preliminary results of a phase-contrast tomographic reconstruction with and without phase retrieval in each X-ray projection. We believe that this method of quantitative X-ray phase-contrast imaging will find applications in biology and medicine, particularly for high-contrast imaging of soft tissues.

Human tissues obtained from cancerous kidneys fixed in formalin were observed with phase-contrast X-ray computed tomography (CT) using 17.7-keV synchrotron X-rays. By measuring the distributions of the X-ray phase shift caused by samples using an X-ray interferometer, sectional images that map the distribution of the refractive index were reconstructed. Because of the high sensitivity of phase- contrast X-ray CT, a cancerous lesion was differentiated from normal tissue and a variety of other structures were revealed without the need for staining.

Optical tomography has recently demonstrated the potential for deep imaging of tissue oxygenation and blood volume, non- invasively, using relatively simple and inexpensive instrumentation. Prior demonstrations of this form of tomography have relied on scanning and data collecting methods limited to imaging bandwidths of Hz or slower. Here we report on an approach that significantly accelerates the imaging rate of optical tomographs based on time-domain methods to well beyond the kHz range. Such high bandwidths are critical for extending the capabilities of optical tomographs to include deep imaging of blood flow and neural activity.

Fluorescence lifetime imaging is a useful tool for quantifying site-dependent environmental conditions in tissue. Fluorophores exist with known lifetime dependencies on factors such as concentrations of O₂ and other specific molecules, as well as on temperature and pH. Extracting fluorophore lifetime for deeply embedded sites in turbid media such as tissue is made difficult by the multiple scattering of photons traveling through tissue. This scattering introduces photon arrival delays that have similar characteristics to the delays resulting from the excitation and subsequent emission of photons by fluorophores. Random walk theory (RWT) provides a framework in which the two sources of diffusion-like delays can be separated so that the part due to fluorescent lifetime can be quantified. We derive a closed-form solution that predicts time-resolved photon arrivals from a deeply embedded fluorophore site. The solution requires that an average absorption coefficient be used. However, it is shown that this assumption introduces only a small error. This RWT-derived solution is also shown to be valid for a range of geometries in which the fluorophore site is embedded at least 10 mean scattering lengths and in which the fluorophore lifetime is less than 1 ns.

A receiver operating characteristic (ROC) experiment was conducted to evaluate the effects of pixel size on the characterization of mammographic microcalcifications. Digital mammograms were obtained by digitizing screen-film mammograms with a laser film scanner. One hundred twelve two-view mammograms with biopsy-proven microcalcifications were digitized at a pixel size of 35 micrometer X 35 micrometer. A region of interest (ROI) containing the microcalcifications was extracted from each image. ROI images with pixel sizes of 70 micrometers, 105 micrometers, and 140 micrometers were derived from the ROI of 35 micrometer pixel size by averaging 2 X 2, 3 X 3, and 4 X 4 neighboring pixels, respectively. The ROI images were printed on film with a laser imager. Seven MQSA-approved radiologists participated as observers. The likelihood of malignancy of the microcalcifications was rated on a 10-point confidence rating scale and analyzed with ROC methodology. The classification accuracy was quantified by the area, A_z, under the ROC curve. The statistical significance of the differences in the A_z values for different pixel sizes was estimated with the Dorfman-Berbaum-Metz (DBM) method for multi-reader, multi-case ROC data. It was found that five of the seven radiologists demonstrated a higher classification accuracy with the 70 micrometer or 105 micrometer images. The average A_z also showed a higher classification accuracy in the range of 70 to 105 micrometer pixel size. However, the differences in A(subscript z/ between different pixel sizes did not achieve statistical significance. The low specificity of image features of microcalcifications an the large interobserver and intraobserver variabilities may have contributed to the relatively weak dependence of classification accuracy on pixel size.

A digital mammography system in which the x-ray sensitive device is a solid-state direct conversion detector is under development. This detector is a 1 mm thick silicon photodiode array hybridized to a CCD read-out, with a 50 micrometer pixel pitch. The detector is designed to be used in a slot-scanned system using time-delay integration (TDI) for signal acquisition. To handle the large signal generated in the photodiode, a novel read-out technique was used, in which charge was integrated 'on-chip' over a small number of rows, and the output of each of these sections was digitally summed 'off-chip' to produce the total integrated signal for each pixel in the image. This two-stage integration process not only allows easy acquisition of large signals, it effectively increases bit depth from 12 bits (for a single section) to approximately 16 (for the total integrated signal). The image quality of the device has been measured and compared to predictions based on cascaded linear systems theory. The resolution of the new detector was determined from the modulation transfer function (MTF) which was obtained from over-sampled edge spread functions (ESF). The ESF was measured in both the scan and slot directions from four repeated images of a tantalum edge. Noise power spectra (NPS) were determined from 40 repeated flat-field images at each of several x-ray exposures. By combining the MTF and NPS measurements, the detective quantum efficiency (DQE) was also determined. The MTF in the non-scanned direction was found to greater than 20% at 10 mm^-1 and slightly lower in the scanned direction (approximately equals 10% at 10 mm^-1). In all cases, the DQE was at least comparable to film-screen mammography receptors. The DQE at 120 mR detector exposure at zero spatial frequency ranged from 0.4 to 0.6 depending on the sample tested. Electronic noise was fairly low, contributing to less than plus or minus 7 ADU (out of a possible 98304 ADU). Future work will involve re-designing the prototype to use a photoconductor with higher density and atomic number to improve quantum interaction efficiency and reduce geometric constraints on image quality.

We present a refractive x-ray lens to be used in a scanned- slit detector system for mammography. The objective is to increase the x-ray flux on the detector and thus reduce scan- times. The lens consists of 600 diamond-cut parabolic cylinders with a radius of curvature of 100 micrometer cut from a 60 mm long slab of PMMA. The refractive effect from each void is very small, but the cummulative effect gives a focal length of approximately 200 mm. A 50 micrometer source was projected onto a 30 micrometer slit 970 mm away with a gain in flux of 4.3 for 27 keV. Since the index of refraction depends on the energy, the gain reaches a peak value for a certain energy, and by changing the geometry and the parameters of the lens, this peak can be tuned to coincide with the optimal energy for the imaging task. In mammography, the favorable energy region is quite narrow, and the spectrum shaping will improve the signal-to-noise ratio. Calculations show about 20% dose reduction with preserved image quality. Furthermore, less or even no filtering of the beam is needed, which corresponds to an additional gain of flux of about a factor of 2.

In this paper, we report measurements from a prototype 1024 X 1024 selenium-based flat panel detector suited for interventional digital mammography applications. This detector is based on an amorphous silicon TFT array, with a pixel pitch of 85 micrometer and a fill factor of 70%. A 200 micrometer layer of amorphous selenium is used to directly convert the incident x-rays into electrical charges. The detector electronics, TFT array, and selenium converter structure are designed to operate at a frame rate of 10 images per second. Experimentally, this detector yields an x-ray sensitivity of nearly 290 electrons/absorbed x-ray nearly 100% absorption of x-rays at a beam energy of 18 keV, a high spatial resolution (limited only by the pixel pitch up to the Nyquist limit), and quantum-noise limited operation down to the lowest exposures currently investigated. Images from the ACR phantom and contrast detail phantom reveal all embedded targets in the phantoms, which indicates the potential of this technology for digital mammography.

Acoustic pressure waves are induced in soft tissue whenever time-varying radiation is absorbed. By recording these time- dependent pressure waves over a sufficient number of angles surrounding the tissue being imaged, it is possible to reconstruct the pattern of radiation absorption within the tissue in three dimensions with spatial resolution that is independent of the carrier frequency of the irradiating energy. We recently constructed the world's first thermoacoustic computed tomography (TACT) scanner, which exploits this physical interaction. Initial in vivo imaging of a human breast was performed using safe levels of 434 MHz radiation. Good soft tissue differentiation with 2 - 5 mm spatial resolution to a depth of 40 mm was achieved. The absorption properties of the breast and the irradiation pattern within the breast determined the TACT image contrast. The length of the RF pulse, the size of the transducers and their frequency response, the geometry of the detector array, and the reconstruction algorithm that was used determined the spatial resolution. We conclude that TACT imaging may have application to breast cancer detection.

The X-ray intensity pattern in a radiograph of the thorax was measured. The measurements were carried out in the detector plane with high spatial resolution. A special phantom prepared from the lung of a dead body in which all fine details were existing was used for these measurements. This phantom was exposed to X-rays applying the standard examination conditions for thoracic imaging according to the 'European Guidelines on Quality Criteria for Diagnostic Radiographic Images,' and the image was stored on industrial X-ray film normally used for non-destructive testing. These images were digitized with a high-resolution drum film scanner. It is shown that the image degradation by the industrial X-ray film and the scanner was very small in the range of the estimated spectrum of the intensity pattern and that no problems with aliasing due to the scanning process would arise. The characteristic curves of the film and the scanner were evaluated. The X-ray intensity pattern of the thorax was calculated on the basis of these curves. Different methods to evaluate the information content of the image in the frequency domain are presented. The results obtained by these methods have been used to propose requirements for digital detector systems.

A novel approach to patient dose and image quality optimization was developed and implemented for chest and lumbar spine radiography. A Monte Carlo model of the imaging chain, including an anthropomorphic voxel-phantom to simulate the patient, was utilized. Detector noise and system unsharpness were modeled and their influence on image quality considered. Image quality was quantified by the contrast ((Delta) OD) and the ideal observer signal-to-noise (SNR) for a number of relevant image details at various positions in the anatomy and measures of dynamic range (DR). Among systems evaluated in a clinical trial, a reference system, acknowledged to yield acceptable image quality, was selected. A large variety of other imaging conditions were simulated and compared to the reference system. Some of the simulated systems were found to give as good imaging performance but at substantially reduced patient doses: 35% and 50% reduction in the lumbar spine AP and the chest PA view, respectively. The model was also used to define a single-valued 'figure-of- merit,' the physical image quality score, PIQS, with the aim to make possible ranking of the imaging systems. By comparing the ranking according to PIQS with radiologists' ranking it was possible to analyze the features in the images which are clinically important.

We have developed a new detector for dual energy computed radiography that uses fast switching of the x-ray tube voltage. Voltage switching gives x-ray spectra with wider separation and lower average energy than conventional 'sandwich' detectors. When properly optimized, these better- conditioned spectra produce dual energy images with lower noise for the same dose. We constructed a prototype of the detector, optimized it, and used it to produce human subject images of ten human volunteers. We also compared the performance of the active detector with a commercial dual- energy, computed radiography system that uses a sandwich detector. The results are as follows: We were able to produce high quality images for all ten volunteers. There were slight artifacts at the edge of the heart in some of the images but these did not affect their diagnostic utility. The active detector required approximately 1/4 the entrance exposure of the commercial, dual-energy system to produce diagnostic quality images of an anthropomorphic chest phantom. A substantial decrease in radiation exposure can be obtained with the dual-energy active detector approach with a modest increase in the complexity of the acquisition equipment.

Twenty-five paired chest images (twenty Posterior-Anterior views and five Lateral views) were obtained under identical x- ray exposure conditions with both an amorphous silicon digital detector system and a thoracic film-screen system (InSight HC screen and InSight IT film; Kodak, Rochester, NY). The digital image data was transformed using non-linear algorithms to obtain a gray-scale appearance comparable to conventional film images. The digital images were printed at full resolution onto laser film (3M, 969). Each hard copy clinical image pair was evaluated by six radiologist experts in a blinded study. Two thirds of the radiologists' ratings showed the digital detector images 'Better' or 'Much Better' than screen-film images (p less than 0.05); Ninety-eight percent of the radiologists' ratings indicated the digital 'Equivalent,' 'Better' or 'Much Better' (p equals 0.0004) than analog film. The digital image system improvement can be explained by significantly higher DQE and an extended dynamic range. The study clearly demonstrates that clinically the new digital x- ray detector radiographic system can produce diagnostic images equivalent or superior to conventional screen film system.

A beam equalization system has been developed for a recently introduced digital slot scanner. Its equalization range is tailored to the dynamic range of digital camera's. Applied beam profiles are made available to be included in the image file for associated image processing. A scanning wheel in the primary beam plays a key role in the generation of beam profile signals and simplifies system alignment. Because of its simplicity, it can be added to already installed camera's Special measures have been taken to minimize X-ray tube loading.

An x-ray imaging detector designed for both radiographic and fluoroscopic medical applications has been developed. The requirement that the detector provide superior imaging in both fluoroscopic and radiographic operation put severe constraints on its design. User requirements and a translation of those requirements to detector performance parameters guided detector design. This paper reports on the performance of a 20.5 X 20.5 cm prototype detector which was a product of this design effort. The detector was tested using both physical measurements and clinical imaging trials. The frequency dependent DQE is used as a measure of contrast to noise performance. Measurements of DQE were made at both fluoroscopic and radiographic signal levels. In fluoroscopic operation, measurements of lag were also made. The detector performance is compared to that of existing and emerging technologies. Results of clinical studies in both radiographic mode (chest imaging) and fluoroscopic mode (cardiac imaging) are reported.

This paper describes a third-generation multi-mode x-ray imager whose applications include low-dose fluoroscopy, cine, spot films, and radiography. In addition, volumetric CT and applications whose environment includes a 2 tesla magnetic field are also in development. The VIP-9 is based on an amorphous silicon TFT/Photodiode array and x-ray conversion screen, which is optionally a deposited CsI(Tl) film or a removable Gd₂O₂S screen. There are three primary modes of operation: RAD for high resolution radiographs and spot films; Fluoro for video rate, low dose fluoroscopy as well as cine; Zoom for high resolution, limited field of view (FOV) fluoroscopy. Through improved electronics, the imager has greater sensitivity at low doses and far better rejection of correlated line noise than its predecessors. In addition, the VIP-9 incorporates many ease-of-use features absent from earlier prototype imagers. While previous reports have primarily focused on the imager construction and noise issues in large area sensing technology, in this paper the emphasis is on features which facilitate integration into a complete imaging system and measures of image quality.

Because of traps in the photodiodes, amorphous silicon detectors retain charge and release it slowly after exposure. One effect of this property is that after an exposure, a ghost image may appear in subsequent images. This is a particular problem when fluoroscopic imaging follows radiographic exposure. In this paper we propose a method for predicting the magnitude of the ghost image so that it can be eliminated. The method uses linear systems theory to model the time behavior of the signal. Then the model is used to predict the offset as a function of time so that offset correction can be performed.

We describe new amorphous silicon (a-Si:H) image sensor arrays which are the highest resolution imagers so far reported. The pixel sizes of 64 micrometer (resolution 8 lp/mm) and 75 micrometer (6.7 lp/mm) are made possible using a photodiode technology that enables high sensor fill factor even in very small pixels. This approach allows the a-Si:H imagers to satisfy high resolution requirements of digital mammography. Each array contains 512 X 512 pixels with matrix addressing provided by a-Si:H thin film transistors (TFT). The high fill factor structure contains a continuous a-Si:H photodiode layer grown on top of the TFT array, with contacts to each pixel through a patterned metal/n⁺ layer. X-ray detection is accomplished by use of a phosphor layer superimposed on the array. The continuous photodiode layer maximizes light absorption from the phosphor and provides high x-ray conversion efficiency. Since the photodiode forms a continuous layer, crosstalk between adjacent pixels due to the lack of isolation is a particular concern, and has been extensively studied. We find that the high fill factor structure can be made such that the lateral charge leakage is minimal in the dark or under moderate illumination, although small amount of charge spreading is observed under conditions of sensor saturation. The measured MTF for optical illumination exceeds 60% at the Nyquist frequency, even for long integration times.

A theoretical cascaded systems analysis of the performance limits of x-ray imagers based on thin-film, active matrix flat-panel technology is presented. This analysis specifically focuses upon an examination of the functional dependence of the detective quantum efficiency on exposure. While the DQE of AMFPI systems is relatively high at the large exposure levels associated with radiographic x-ray imaging, there is a significant decline in DQE with decreasing exposure over the medium and lower end of the exposure range associated with fluoroscopic imaging. This fall-off in DQE originates from the relatively large size of the additive noise of AMFPI systems compared to their overall system gain. Therefore, strategies to diminish additive noise and increase system gain should significantly improve performance. Potential strategies for noise reduction include the use of charge compensation lines while strategies for gain enhancement include continuous photodiodes, pixel amplification structures, or higher gain converters. The effect of the implementation of such strategies is examined for a variety for hypothetical imager configurations. Through the modeling of these configurations, such enhancements are shown to hold the potential of making low frequency DQE response large and essentially independent of exposure while greatly reducing the fall-off in DQE at higher spatial frequencies.

In MRI, gradient magnetic fields are used to obtain the spatial information by frequency modulation of the received signal. The gradient fields are generated by switching currents on the gradient coils, which generates acoustic noise due to Lorentzian force. In particular, fast imaging methods, which are usually used for fMRI, require fast switching of the gradient pulse, thereby generating large acoustic noise. The intensity of the acoustic noise depends on the imaging method and the pulse sequences. The acoustic noise induced by gradient pulsing may interfere for signal enhancement of brain areas with the presentation of auditory stimuli during fMRI. In this paper, the gradient pulsing effects on fMRI are analyzed for different combinations of gradients. The experimental results show that total activations by visual stimulation are slightly larger for a combination of Z readout and Y phase-encoding gradients than those for a combination of Y readout and Z phase-encoding gradients when sagittal-view fMRI is performed.

Mammography is a critical medical imaging procedure concerning resolution due to the features of the details of clinical interest for the diagnosis, such as specifically the microcalcifications. Nonisotropic mammographic systems usually present discrepant dimensions of the effective focal spot between the parallel and the perpendicular axes relative to the tube axis, and this may cause a significant difference in microcalcifications sharpness, depending on the location where they are positioned. We could previously verify that there are regions of the radiation field in nonisotropic systems where sharper images can be obtained, named as Optimum Region. This work is about the determination of the Optimum Region for nonisotropic mammographic systems, by means of developing a procedure which determines the modulation transfer function (MTF) due to the focal spot by simulation. This procedure allows to obtain the MTFs corresponding to any orientation and location of the field, and, by a quantitative analysis of the MTFs curves, the Optimum Region can be determined. Comparisons with phantom images obtained in actual mammographic systems has shown that the Optimum Regions found by the MTF method were in agreement with the subjective visual analysis of the microcalcifications sharpness.

A Monte Carlo program has been developed to model X-ray imaging systems. It incorporates an adult voxel phantom and includes anti-scatter grid, radiographic screen and film. The program can calculate contrast and noise for a series of anatomical details. The use of measured H and D curves allows the absolute calculation of the patient entrance air kerma for a given film optical density (or vice versa). Effective dose can also be estimated. In an initial validation, the program was used to predict the optical density for exposures with plastic slabs of various thicknesses. The agreement between measurement and calculation was on average within 5%. In a second validation, a comparison was made between computer simulations and measurements for chest and lumbar spine patient radiographs. The predictions of entrance air kerma mostly fell within the range of measured values (e.g. chest PA calculated 0.15 mGy, measured 0.12 - 0.17 mGy). Good agreement was also obtained for the calculated and measured contrasts for selected anatomical details and acceptable agreement for dynamic range. It is concluded that the program provides a realistic model of the patient and imaging system. It can thus form the basis of a detailed study and optimization of X-ray imaging systems.

An edge-enhanced imaging technique using the X-ray refraction effect for high contrast outline imaging have been investigated using a third generation of synchrotron radiation source. This technique can be applied to imaging of soft tissues that can not be imaged by conventional absorption- contrast imaging. We have attempted to apply this method to accurate diagnosis of the lung cancer with reduced absorbed dose. In preliminary experiments, we took images of a glass capillary tube and a nude mouse with a long object-to-detector distance using monochromatized X-ray and a high-spatial- resolution CCD-based image detector. Compared to conventional absorption-contrast images, the image contrast is enhanced at an interface between two materials by the X-ray refraction. In a chest image of the mouse, outline images of the alveoli, the bronchi and the trachea were visualized with higher contrast than that of the ribs. It may be effective for early detection of small lung cancer lesions.

This paper describes an approach to further improving the image quality of computed radiography (CR) by use of a both- side reading method and an Imaging Plate (IP) with a transparent support. The most important factor for determining the image quality in an X-ray imaging system is X-ray utilization efficiency. Accordingly, we have proposed an imaging plate that consists of a transparent support and a thicker photostimulable phosphor layer, and a reading method that permits detection of emissions from both sides of the IP, so that an improvement in image quality can be achieved by adding image data derived from both sides of the IP. We have also found that the optimal addition ratio is dependent upon spatial frequency, and proposed a spatial filter-based addition process method for achieving optimal addition over the entire spatial frequency. The system based on the proposed reading method and addition method provided 30 - 40 % increase in detective quantum efficiency (DQE), as compared to single- side reading method. Furthermore, it has been confirmed that the DQE of the addition image is dependent upon the correlation between the data of the two images detected from both sides of the imaging plate. The reading method proposed in this paper can provide increased DQE, so that enhancement in diagnostic performance can be anticipated.

The many applications of Monte Carlo modeling in PET arouse to increase the accuracy and computational speed of Monte Carlo codes. The accuracy of Monte Carlo simulations strongly depends on cross-section libraries used for photon transport calculations. Furthermore, large amounts of CPU time are required to obtain meaningful simulated data. We present a comparison of different photon cross-section libraries together with the parallel implementation of the 3D PET Monte Carlo simulator, Eidolon on a MIMD parallel architecture. Different photon cross-section libraries and parametrizations show quite large variations as compared to the most recent EPDL97 nuclear data files for energies from 1 keV to 1 MeV. Together with the optimization of the computing time performances of the Monte Carlo software, photon transport in 3D PET could be efficiently modeled to better understand scatter correction techniques. In implementing Eidolon on a parallel platform, a linear speedup factor was achieved with the number of computing nodes.

Scattered radiation reduces image contrast and is probably the biggest factor contributing to poor diagnostic quality in radiography. In conventional radiography, X-ray grids are widely used to improve the diagnostic quality of radiographs by absorbing the great part of scattered radiation. In digital radiography the decoupling of the image acquisition and display allows potential image processing solutions to remove the effects of scatter. In this work, we present novel scatter and grid models that are constructed to aid in finding potential solutions for scatter reduction that do not entail the degree of tradeoff between scatter removal and dose as in conventional radiography.

Image quality characteristics of spiral CT are similar to conventional CT except for image noise and slice sensitivity profile. While these factors have been studied for high contrast, low contrast detectability in spiral CT has not yet been fully explored. Low contrast simulations that have been performed demonstrate the behavior of contrast sensitivity as related to the parameters of pitch, collimation, size of the object, alignment between the object and reconstructed slice, and reconstruction algorithm. The effect of spiral CT parameters on contrast to noise ratio (CNR), one component of low contrast detectability, is the focus of this study. A 9.5 mm diameter object was scanned using different collimations and pitches and reconstructed with different intervals. Results showed that low contrast detectability is optimized with collimations 7 mm and 10 mm at a pitch of 1 and the lowest CNR values were at a 1 mm collimation. A pitch of 1 had the highest CNR values while a pitch of 3 had the lowest CNR values. For similar table speeds selection of a lower pitch will improve CNR. The effect of object-slice alignment on CNR indicates no significant difference at a pitch of 3, but it had the lowest CNR values. For a pitch of 1 or 2, a small reconstruction interval is preferred over a large reconstruction interval.

The Algebraic Reconstruction Technique (ART) reconstructs a 2D or 3D object from its projections. It has, in certain scenarios, many advantages over the more popular Filtered Backprojection approaches and has also recently been shown to perform well for 3D cone-beam reconstruction. However, so far, ART's slow speed has prohibited its routine use in clinical applications. Currently, a software implementation requires several hours for a 3D reconstruction, even on modest reconstruction grid sizes. Although one solution to combat these problems would be the time-consuming design of expensive custom accelerator boards, we would rather like to resort to existing and widely available hardware for our purposes. In this sense, we find that ART's main operations, i.e., volume projections and image backprojections, can be performed very rapidly on standard 2D texture mapping hardware, resident in many graphics workstations and PC graphics boards. In this paper, we discuss the use of this hardware in two volume decomposition modes: voxel and slice. Although we find that the speedups obtained in the voxel mode are respectable, the speedups obtained in the slice-mode are tremendous. Here, a quality cone-beam reconstruction on a 128³ grid can be obtained in less than 2 minutes, which corresponds to a speedup of over 70. Since our rapid ART reconstruction algorithm can be run on the same workstations that are typically used for the viewing of clinical datasets, it is immediately available for routine parallel- and cone-beam CT.

The flat panel detector (FPD) has become a highly promising candidate for a wide variety of applications. A prototype selenium thin film transistor (STFT) array-based volume tomographic angiography (VTA) imaging system has been constructed for the feasibility study. This experimental set- up uses a 14' X 17' STFT detector with a 2560 X 3072 array of 14 bit pixels. While an STFT detector offers high resolution digital images, there will always be some defects on the detector. These defects will result in severe streaks and ring artifacts, which have been found in reconstructed images of preliminary phantom studies. It is obvious that the stationary noise sources of the FPD are enhanced by the reconstruction procedure. In this paper, an accurate and efficient FPD calibration method for the VTA imaging system is proposed to reduce the artifacts. An improved gain map and a bad pixel detection method with an adaptive threshold are introduced based on statistical models of the FPD. A more efficient localized and sensitive bad pixel detection ability is obtained by sub-dividing the detector array into sub- arrays, classifying bad pixels as different regional patterns, and then optimizing an interpolation scheme for each pattern. The real-time background correction, gain correction, bad- pixel correction, and methods to generate calibration maps are described in detail. The calibration technique is examined through phantom studies and evaluated by comparing the artifacts and noise in reconstructed images. Improvement of image quality is obtained utilizing the calibration technique. It has been clearly verified that the streaks and ring artifacts in reconstructed VTA images are significantly reduced. Finally, the advantages of our method and future works are also discussed.

The Kinestatic Charge Detector (KCD) digital radiography system has proven itself experimentally to be comparable with or superior to other x-ray imaging systems in the production of quality images at the same dose. The prototype large-field detector design has obtained images that have relatively high spatial and contrast resolution with low scatter and low quantum noise compared with current commercially available clinical x-ray systems. The NIH has approved a grant to develop and construct an advanced clinical KCD digital radiography system. The goals of this project are to design the gantry and clinically evaluate the new system. This system will allow for improved diagnosis, reduced patient dose, and provide other features unique to a digital radiography system.

We developed the automatically controlled X-rays compensating filter system (Automatic Filter System) to prevent the halation area that catheter's operation of the doctor in the angiography is obstructed. The algorithm which we have developed consists of the following five steps. 'Capture:' We take in a fluoroscopic image. 'Correction:' We correct the amount of decline by the X-rays compensating filter (filter) if the filter is already inserted. 'Detection:' We detect the halation area. 'Calculation:' We calculate an insertion position and rotate angle. 'Control:' We control the filter. We were able to get the following effects by the development of the Automatic Filter System. First, it was possible that the fluoroscopic image was displayed without losing contrast. Second, we were able to establish the most suitable X-rays conditions, such as tube voltage and tube current, for the region of interest (ROI), for example catheter and cardiovascular. And third, as the most important effect, we were able to decrease the exposure dose during the fluoroscopic procedure. Moreover, to confirm the validity of the Automatic Filter System, we applied this system to the angiography. As a result, we made sure that the operator could conduct the fluoroscopic procedure without manually control.

We have developed avalanche photodiode (APD) arrays of the beveled edge type with high responsivity in the ultraviolet (UV) region. A 3 X 3 array with pixel size 3 X 3 mm² was made, in which the segmentation was done using selective diffusion in the front surface. This technique is an improvement over previous avalanche photodiodes, where the segmentation was done cutting channels in the back of the detector. After the die was cut from the wafer, beveling and passivation was performed to avoid lateral surface breakdown. The responsivity from 200 to 400 nm is close to 1.1 A/W., which makes this detector suitable for florescence applications. The gain and dark current coming from all pixels connected together, is the same as a single element detector with the same area, which indicates that the pixelization process does not reduce the performance compared to a detector without segmentation. We measured high crosstalk between adjacent pixels (30%) which indicates that the resistance between them is too low.

Geometric distortions are one of the most important degrading factors in MRI. They usually do not greatly affect the clinical relevance of images, but their correction is indispensable for lesion volume measurements, radiotherapy and surgical planning. In this work, the main sources of geometric distortion in Cuban low-field MRI systems are studied. Geometric distortion models and correction algorithms are tested by means of computer simulation using theoretical distributions of the magnetic fields. The real distributions are determined from images of a grid phantom. Calculated static field distributions showed that the system magnetic center is shifted, relative to magnet geometric center. Quantitative measurements provided inhomogeneity values (93 ppm in a spherical region of diameter 256 mm) larger than calibration data (65 ppm) obtained 15 months ago. The shim settings must be readjusted. The temporal behavior of static field was also studied. The magnet heating produces a slow time variation in static field intensity, but field error distribution proved to be stable. In the images, geometric distortions increase with increasing distance from image center and ranged from -6 to 7 mm. The implemented correction procedure reduced distortions from maximally 7 mm to the order of pixel resolution (0.8 - 1 mm).

Although scattered radiation is generally regarded as a nuisance in radiological imaging, many innovative imaging concepts that use the scattered field have been demonstrated. A systematic approach, however, for analyzing the medical applications of x-ray scatter imaging has been lacking. We have therefore formulated a simple semi-analytic model that consists of imaging a target object against a background material of the same dimensions when both are situated within a water phantom. The target and background objects have small cross-sectional areas (1.0 mm²) to allow the omission of self-attenuation and multiple scatter within the objects. The incident energy fluence is kept constant so that similar doses are delivered by the various photon beams. For imaging white brain matter versus gray brain matter in a 15 cm thick water phantom, the maximum signal-to-noise ratio, over all photon energies, for images obtained with the forward scatter between 2 degrees - 12 degrees exceeds that of primary images for all object thicknesses less than or equal to 40 mm. The penalty in dose as a result of spectral blur is generally moderate. For example, using an 80 kV beam for the previous imaging task would require approximately a 24% dose increase relative to using a monoenergetic beam. A high-precision experimental apparatus has been assembled to validate our predictions.

In a previous paper a portal imaging system was described that used a 200 mm diameter, 12.7 mm thick CsI(Tl) crystal transparent screen coupled with a Nikkor 35 mm f1.4 lens to an Astromed liquid nitrogen cooled CCD TV camera system. The whole 200 mm circular field of the crystal was imaged at a 0.53 mm pixel size. The geometry of the imaging system was optimized to achieve high spatial resolution in spite of the thickness of the screen. Since the last paper was written, an Angenieux 25 mm f0.95 Super 16 mm movie camera lens has been purchased which has higher optical quality and slightly better depth of focus than the Nikkor 35 mm f1.4 lens, but gives a pixel size with the Astromed CCD system of 0.8 mm. The Angenieux lens has been used with the Astromed CCD and the 200 mm diameter CsI(Tl) screen to acquire portal images of human subjects. Images of patients being treated in the prostate region have been acquired. These CsI(Tl) images have been compared with portal images produced using conventional portal film. The CsI(Tl) images are of higher quality than the film images and and be acquired at lower dose.

The aim of this study was to quantify the effects of inhomogeneities on magnetocardiography (MCG) forward solutions. A numerical model of a human torso was used which construction included geometry for major anatomical structures such as subcutaneous fat, skeletal muscle, lungs, major arteries and veins, and the bones. Simulations were done with a single current dipole placed at different sites of heart. The boundary element method (BEM) was utilized for numerical treatment of magnetic field calculations. Comparisons of the effects of different conductivity on MCG forward solution followed one of two basic schemes: (1) consider the difference between the magnetic fields of the homogeneous torso model and the same model with one inhomogeneity of a single organ or tissue added; (2) consider the difference between the magnetic fields of the full inhomogeneous model and the same model with one inhomogeneity of individual organ or tissue removed. The results of this study suggested that accurate representation of tissue inhomogeneity has a significant effect on the accuracy of the MCG forward solution. Generally lungs, subcutaneous fat, skeletal muscle play a larger role than other tissues. Our results showed that the inclusion of the boundaries also had effects on the topology of the magnetic fields and on the MCG inverse solution accuracy.

By changing the design paradigm for radiographic detectors to optimize imaging of a region of interest (ROI) for endovascular interventions, a new class of micro-angiographic detectors is proposed. Such ROI imagers optimized for high spatial resolution over a fraction of the conventional FOV accept compromises in x-ray absorption so that the desired high frequency DQE is achieved. A prototype demonstration system based upon a CsI(Tl) phosphor coupled by a fiber taper to a CCD is compared with an image-intensifier-based digital angiographic (DA) system for imaging typical neuro-vascular pathologies (stenoses, aneurysms, arterio-venous malformations (AVMs) and a variety of stents with wire diameter down to 50 micrometer. Although the zero frequency DQE of the DA unit exceeded that of the prototype with its thin phosphor layer, the prototype excelled in the desired frequency range of 3 - 10 lp/mm. Using an artery block phantom, the smallest 1 mm diameter stenoses and aneurysms were clearly visualized only with the prototype. For imaging stents, details of wires and struts were only visible with the prototype. ROI images of an AVM pig rete model showed more detailed angio-architecture compared to blurred-appearing DA images. It is expected that future such ROI cameras should allow improved clinical interventions.

The Scanning-Beam Digital X-ray (SBDX) system promises low- dose cardiac fluoroscopy and angiography with excellent image quality. The system demands a detector capable of high count rates and excellent detection efficiency. Cadmium zinc telluride (CdZnTe) is well suited to these requirements. The SBDX detector comprises sixteen 3-mm-thick, 13.5 mm X 13.5 mm tiles arranged in a 4 X 4 array. Each tile has 144 imaging elements. Thus, the entire detector measures 54.0 mm X 54.0 mm and includes 2,304 imaging elements on a 1.125 mm pitch. Because the SBDX system has a geometric magnification of 3.3, the imaging-element size is consistent with a system spatial-resolution of 2.2 lp/mm. The 3-mm thickness is chosen to guarantee a stopping efficiency of more than 90% at 120 kVp. Each detector tile is flip-chip mounted to a custom- designed integrated circuit (IC) using indium bump bonding techniques. Fabricated in a 1.2-micrometer CMOS process, the IC includes high-speed photon-counting circuitry that operates at rates up to 5 X 10⁶ counts/s(DOT)mm². The circuitry is designed both to maximize the achievable count- rate and to minimize false double counts due to charge sharing between elements. Testing confirms that the detector performs with minimum cross talk between elements at count rates in excess of 2 X 10⁶ counts/s(DOT)mm². Measurements of the detective quantum efficiency (DQE) are presented. The relationship between material properties and detector performance is also discussed. The circuit design and device fabrication techniques are applicable to a variety of imaging applications.

We investigated the influence of a scatter reduction grid on the spatial resolution properties of a digital X-ray imaging system (Philips Medical Systems, Thoravision). The spatial resolution using a grid was nearly independent of the X-ray energy as the modulation transfer function (MTF) showed. Maximum spatial resolution was reduced because of the grid in case of a constant exposure dose. The amount of absorbed photons raised and Wiener spectra were increased by the factor of ten. The scattered radiation was reduced by the grid at higher spatial frequencies and the Wiener spectra of different exposure parameters reached approximately the same value. The measurement of the detective quantum efficiency (DQE) showed the influence of absorption characteristics of the detector (aSe) on the resulting image quality. Detector absorption efficiency increased at a lower X-ray energy. This resulted in the opportunity to compensate the effects of low signal dynamics because of higher grid absorption. Inserting a filter in the X-ray beam altered the shape of the photon energy spectrum and influences image quality by changing the contrast. For this reason, we investigated the spectral composition of the X-rays, too. There were advantages using additional filtration opposite the scatter reduction grid. However, the scattered radiation was not reduced by the filtration and therefore this did not improve the signal-to- noise characteristic. The X-ray scatter reduction grid increased the detail detectability in high absorption image areas. However, without greater exposure doses, the use of a scatter reduction grid did not have as much advantage as expected. A detector material with higher quantum efficiency could solve this problem.

A laser produced plasma (LPP) x-ray source utilizing ultra- fast laser was investigated in the context of its utility for dual energy subtraction angiography. Experiments were performed with a Table Top Terawatt (TTT) laser using BaF₂ and rare-earth metal targets including La, Ce, Nd and Gd. The laser was operated in a single-pulse or in a dual pulse mode with surface power density in the 10¹⁸ - 10¹⁹ Wcm^-2 range with pulse duration of 150 or 450 fs. Infrared and/or green beams were utilized. Hot electrons' temperature was in the 35 - 50 keV range. The obtained LPP x- ray spectra were comprised of a continuous bremsstrahlung component as well as discrete characteristic lines. The bremsstrahlung extended to high energies with no evident cutoff energy below 100 keV. Its shape was best described by exp(-E/kT_e), where T_e is the hot electron temperature. The overall efficiency was equal to approximately 9 X 10^-4 for 450 fs pulse and approximately 6 X 10^-4 for 150 fs pulses. The x-ray focal spot size was in the range 13 to 50 microns. We have found that the LPP x- ray source with BaF₂ and rare-earth targets provide x-ray spectra that might be suitable for DESA.

We describe a 19 X 28 cm digital x-ray detector for use in clinical screening mammography. The detector design uses 6 CCD/taper modules tiled together to provide a full breast image without any image gaps at the tapers interfaces. We have attempted to limit the cost and mechanical complexity of the detector while maintaining a high DQE and a resolution cutoff of 10 lp/mm. The camera design employs a Gd₂O₂S phosphor x-ray converter, 3.3:1 demagnifying fiber optic tapers, and front-illuminated Thomson THX 7899 CCDs (2048 X 2048, 14 micrometer pixels) coupled to the tapers with optical epoxy. The effective pixel size in the 4000 X 6000 pixel full image is 47 micrometer. The signal for each pixel is digitized to 16-bits and the effective dynamic range is approximately 9000. Readout time for the 48 Mbyte image is 8s. Each CCD is cooled to -3 degrees Celsius by a thermoelectric (TE) cooling module.

We propose a small field-of-view color image acquisition system for the imaging and measurement of skin lesion and its properties in dermatology. The system consists of a 3 chip CCD camera, a frame grabber, a high-quality halogen annular light source and a pentium PC. The output images are in a standard device-dependent color space called sRGB or ITU-R BT.709 which has a known relation to the device-independent CIE XYZ color space and provides a fairly realistic view on a modern CRT- based monitor. In order to transform the images from the unknown and variable input RGB color space of the acquisition system to the sRGB space a profile of the acquisition system is determined based on 24 color targets with known properties. Determination of this profile is simple and quick, and it remains valid for many hours of operation (weeks or even months of normal use). Precision or reproducibility of the system is very good, both short-term (consecutive measurements) <(Delta) E*_ab> equals 0.04 with (Delta) E*_ab less than 0.1, medium-term (measurements under one profile but on different warm-up cycles) <(Delta) E*_ab> equals 0.34 with (Delta) E*_ab less than 1.2. Long-term precision (measurements under different profiles) is of the same order. Accuracy was evaluated for profiles based on different RGB to sRGB polynomial transforms computed both by linear least-squares in the sRGB space and by non-linear optimization in CIE L*a*b* color space. Results show that, using a set of test targets consisting of 15 paper color targets and 12 real measurements of human skin, the simple linear transform outperforms higher order polynomials and has <(Delta) E*_ab> equals 6.53, with (Delta) E*_ab less than 11.21. A small study of the pigmentation of the human skin after UV-radiation shows that when measuring areas of at least a few hundred pixels differences of more than 2 - 3 dE units are statistically significant.

This paper reports the validation work concerning the Monte Carlo simulator we developed for Nuclear Medicine imaging. First we compared the simulated and acquired data of a Data Spectrum Thorax phantom. The two data sets agree fairly well but significant differences can be found at the pixel level. They are probably due to slightly different experimental conditions which are very difficult to control. Second we compared the simulated data obtained using three different interaction models. No difference could be found at the pixel level but image-wide energy spectra slightly differ.

A pixel profile in the conventional burst sequence is so poor that excited area by burst sequence is a small portion of a pixel. This causes poor signal to noise ratio in burst image. In this paper, a frequency sweeping of RF pulse for burst imaging sequence is proposed to improve pixel profile. A burst pulse train is shaped by linear or nonlinear frequency sweeping function so that all the spins within a pixel are excited, thereby improving the signal to noise ratio. It also shows that the pixel profiles are dependent on how the frequency sweep is made. Computer simulations with Bloch equation and experimental results obtained using a 1.0 T NMR imaging system are presented.

In megavoltage imaging, current commercial electronic portal imaging devices (EPIDs), despite having the advantage of immediate digital imaging over film, suffer from poor image contrast and spatial resolution. The feasibility of using a kinestatic charge detector (KCD) as an EPID to provide superior image contrast and spatial resolution for portal imaging has already been demonstrated in a previous paper. The KCD system had the additional advantage of requiring an extremely low dose per acquired image, allowing for superior imaging to be reconstructed form a single linac pulse per image pixel. The KCD based images utilized a dose of two orders of magnitude less that for EPIDs and film. Compared with the current commercial EPIDs and film, the prototype KCD system exhibited promising image qualities, despite being handicapped by the use of a relatively simple image calibration technique, and the performance limits of medical linacs on the maximum linac pulse frequency and energy flux per pulse delivered. This image calibration technique fixed relative image pixel values based on a linear interpolation of extrema provided by an air-water calibration, and accounted only for channel-to-channel variations. The counterpart of this for area detectors is the standard flat fielding method. A comprehensive calibration protocol has been developed. The new technique additionally corrects for geometric distortions due to variations in the scan velocity, and timing artifacts caused by mis-synchronization between the linear accelerator and the data acquisition system (DAS). The role of variations in energy flux (2 - 3%) on imaging is demonstrated to be not significant for the images considered. The methodology is presented, and the results are discussed for simulated images. It also allows for significant improvements in the signal-to- noise ratio (SNR) by increasing the dose using multiple images without having to increase the linac pulse frequency or energy flux per pulse. The application of this protocol to a KCD system under construction is expected shortly.

In megavoltage imaging, current commercial electronic portal imaging devices (EPIDs), despite having the advantage of immediate digital imaging over film, suffer from poor image contrast and spatial resolution. In a previous paper, a prototype megavoltage portal imaging system was described that utilized a 3 mm thick 100 mm field of view CsI (Tl) transparent scintillating crystal (corresponding to a radiological thickness of 1350 mg/cm²) coupled to a liquid nitrogen cooled slow-scan CCD camera with a combination of two camera lenses to yield a 42 mm f1.0 macro lens and a 5:1 demagnification. The imaging display significantly superior contrast and spatial resolutions (1 lp/mm at 20% MTF) to that available from the commercial EPIDs, which typically consist of a CCD camera coupled to a relatively thin gadolinium oxysulfide screen (with a radiological thickness of 400 mg/cm²). However it required significantly higher dose than portal film. Subsequent effort has focused on optimization of the optics and scintillator thickness in order to reduce the required imaging dose, while still providing superior image and contrast resolutions to that of the commercial EPIDs. Improved images were acquired using a two- camera lens combination yielding a 50 mm f1.1 macro lens with a 7:1 demagnification. Subsequently, portal imaging with an even thicker 13 mm CsI(Tl) scintillator (corresponding to a radiological thickness of 5850 mg/cm²) was carried out. An increase in scintillator thickness was accompanied by only a small loss in spatial resolution (1 lp/mm at 17% MTF) by optimizing the optical geometry. The image quality was significantly superior to that of the commercial EPIDs (Elekta SRI-100 and Siemens BEAMVIEW), and comparable to that for portal film, while requiring an imaging dose that was less than or comparable to that for film or the EPIDs. The purpose of this research is to investigate the effect of spectral shifting and buildup material or imaging for this prototype system. The use of clear thick single crystal scintillators is relatively new in portal imaging. Early work on optimization of CCD based EPIDs dealt primarily with amorphous nontransparent scintillators, and the use of thick scintillators was abandoned due to a clinically unacceptable associated loss in spatial resolution. Optimization of CCD based EPIDs has been implicitly based on the use of thin scintillators. This recent imaging success of the CsI(Tl) scintillator CCD camera based system utilizing a relatively thick scintillator offers a possibly superior alternative to the current CCD based systems. This superior imaging was accomplished in the absence of any optimization dealing with the choice of buildup material or thickness. Such optimization presents the potential for further gains in imaging quality. Experimental results dealing with optimization of scintillator thickness and buildup plate thickness and material are presented. The effect on image quality due to a spectral shift in a 6 MV photon beam in the presence of phantom scatter is discussed.

The application of the newly developed flat panel x-ray imaging detector in cone beam volume CT has attracted increasing interest recently. Due to an imperfect solid state array manufacturing process, however, defective elements, gain non-uniformity and offset image unavoidably exist in all kinds of flat panel x-ray imaging detectors, which will cause severe streak and ring artifacts in a cone beam reconstruction image and severely degrade image quality. A calibration technique, in which the artifacts resulting from the defective elements, gain non-uniformity and offset image can be reduced significantly, is presented in this paper. The detection of defective elements is distinctively based upon two-dimensional (2D) wavelet analysis. Because of its inherent localizability in recognizing singularities or discontinuities, wavelet analysis possesses the capability of detecting defective elements over a rather large x-ray exposure range, e.g., 20% to approximately 60% of the dynamic range of the detector used. Three-dimensional (3D) images of a low-contrast CT phantom have been reconstructed from projection images acquired by a flat panel x-ray imaging detector with and without calibration process applied. The artifacts caused individually by defective elements, gain non-uniformity and offset image have been separated and investigated in detail, and the correlation with each other have also been exposed explicitly. The investigation is enforced by quantitative analysis of the signal to noise ratio (SNR) and the image uniformity of the cone beam reconstruction image. It has been demonstrated that the ring and streak artifacts resulting from the imperfect performance of a flat panel x-ray imaging detector can be reduced dramatically, and then the image qualities of a cone beam reconstruction image, such as contrast resolution and image uniformity are improved significantly. Furthermore, with little modification, the calibration technique presented here is also applicable to the flat panel x-ray imaging detector in digital radiography (DR), fluoroscopy (DF) and mammography.

A linear array silicon pixel detector has been developed to perform digital radiology with synchrotron radiation: in this communication we present the first results obtained using a three layer prototype at the Elettra Synchrotron Radiation Facility (Trieste, Italy). High efficiency (about 85% at 20 keV) is achieved by means of an Si strip detector irradiated in an 'edge-on' geometry: the pixel dimensions are 200 X 300 micrometer². To obtain a sensitive area of about 50 X 1 mm² stack of three layers has been assembled. Images are obtained by means of scanning techniques and their spatial resolution is determined by the scanning step. Detectors are read-out by dedicated, low noise VLSI CMOS chips in single photon counting mode. The distance between each detector layer has been measured to be 115 plus or minus 10 micrometer and the layers are parallel within a maximum tilt angle of 0.08 degrees. Cross-talk effects were found to be always smaller than 2% of the counts of the neighboring layers. The MTF in the scan direction has been evaluated. The acquisition time needed to acquire an image of a mammographic test object is about 6 minutes. The mean glandular dose is about 30% of the dose delivered at the conventional mammographic unit for the same test object while the detail visibility is comparable. These preliminary results indicate that a large area linear array silicon pixel detector can be developed: using this detector and a monochromatic synchrotron radiation beam the delivered mean glandular dose is significantly reduced when compared to conventional mammographic examinations.

An improved approach to the current fixed sampling area design of automatic brightness control (ABC) systems installed in most fluoroscopy units is described. Using a binary image to define the sampling and non-sampling areas, the digital image data and statistical information within the sampling area of the image can be extracted in real time for feedback control of the ABC system. The operator can select the binary image so the sampling area in that binary image matches the feature of interest wherever it is in the field of view. This design allows greater control and flexibility in the selection of sampling area for ABC systems in digital fluoroscopy, and therefore overcomes the problems inherent in current fixed sampling area systems. It is applicable to current image intensifiers and new flat panel detectors.

This paper addresses the problem of Magnetic Resonance (MR) imaging of moving targets using spiral scan echo planar data collection. A system model is developed to interpret the readouts of repeated spiral excitations via two separate time variables, the slow-time and fast-time. This mathematical model is used to construct an inversion for forming the time progression of the target image. A method for increasing the repetition rate of the spiral data collection is presented. Results are provided.

This paper presents a modification of the z-slope algorithm introduced in the accompanying paper 'An algorithm to correct for z-detector gain variations in multislice volumetric CT.' The x-expansion coefficients found by SVD are plugged into the equation describing the final estimate of the error vector. It is then found that the algorithm can be expressed in vector form as a matrix equation relating the final error estimate to the initial error estimate, where the final error vector estimate e is subtracted from the projection data for correction. In this form, the matrix coefficients, but not the size, are dependent on the correction order. Also, the final correction matrix MXF (size Nchan by N, where Nchan is the number of channels corrected) is calculated channel-by- channel, by evaluating MX as above for a segment of N channels (with N odd), and extracting the central row to define the corresponding row of MXF. The segment of N channels is then slid by one channel, and the process is iterated till completion of MXF. Although the number of calculations required to obtain the correction matrix is greatly increased, as MXF depends only on the detector cell gains, these calculations can be performed off-line in calibrations. The scan data-dependent calculations are greatly streamlined.

A semi-automated, quantitative film digitizer quality control program that is based on the computer analysis of the image data from a single digitized test film was developed. This program includes measurements of the geometric accuracy, optical density performance, signal to noise ratio, and presampled modulation transfer function. The variability of the measurements was less than plus or minus 5%. Measurements were made on a group of two clinical and two laboratory laser film digitizers during a trial period of approximately four months. Quality control limits were established based on clinical necessity, vendor specifications and digitizer performance. During the trial period, one of the digitizers failed the performance requirements and was corrected by calibration.

Though most objections to the use of selenium are largely unfounded (lag and ghosting effects, low DQE), the high bias voltage associated with the thick layer of selenium required to have an acceptable x-ray absorption in radiography and fluoroscopy applications, may have some practical inconvenience. The purpose of this study was to evaluate the pertinence of a solution using a thin coplanar selenium layer, as a photosensitive converter requiring only a few tens of volts of bias, associated with a thick columnar coating of sodium doped cesium iodide scintillator. It will be shown that CsI(Na) can be evaporated with a very uniform needle-like morphology on amorphous selenium structures, the later showing no evidence of thermal recrystallization. Photoluminescence characterization of this scintillator material shows a light emission peak centered at 420 nm as expected, which matches the sensitivity spectrum of selenium. Preliminary sensitivity measurements give a signal in the range of 2000 pC/cm²/mR for 400 (mu) -CsI, with no reflector present. The thin selenium layers deposited display low dark currents of less than 130 pA/cm² at an electric field of 10 volts per micron. Work in progress will be presented including the scintillator (x- ray absorption, sensitivity and emission), the thin selenium photosensor as well as the coupled structure characteristics.

This work incorporates development of a model that relates 3-D arrangement of breast tissue and its appearance in a mammogram. The primary contribution is modeling breast structures based on their anatomic properties, obtained from the literature and histologic slice images. Additionally, breast compression deformation is estimated using tissue elastic properties and a deformation model. Synthetic mammograms obtained by the proposed approach can be used in analysis of positioning and compression effects, and for testing computer algorithms for detection of abnormalities. The paper discusses three major model components: (1) modeling breast anatomic structures, (2) modeling breast tissue compression, and (3) modeling X-ray image acquisition.

Computer Analysis of Mammography Phantom Images (CAMPI) is a method for making quantitative measurements of image quality. This paper reports on further applications of the method to a prototype full-field digital mammography (FFDM) machine. The specific aim was to investigate the effect on speck Signal-to- Noise-Ratio (SNR) of grid vs. non-grid techniques and different target-filter imaging conditions, for 4-cm thick phantoms. Images of a 50-50 composition, 4-cm thick phantom containing a mammography accreditation phantom insert plate, were acquired on a General Electric FFDM machine under conditions of constant pixel value and constant mean glandular dose. They were obtained under conditions of grid and no-grid with the Mo-Mo target/filter. Also acquired were 4-cm phantom grid images at constant dose using different targets-filter combinations (Mo/Mo, Mo/Rh, Rh/Rh) and different phantom material glandular-fat compositions (percentages: 30-70, 50- 50, 70-30). Analyses of the images yielded signal-to-noise- ratio (SNR) for the specks and a non-uniformity measure. The SNR was converted to a Figure of Merit (FOM) by dividing by the square root of the mean glandular dose. For the grid-non- grid study, the FOM plots showed a broad maximum at about 26 - 28 kVp, meaning that this range is optimal in dose efficiency for imaging a 4-cm breast of 50-50 composition. The non-grid FOM values were greater than the grid values, meaning that the former was more dose-efficient. For the Target/Filter study the FOM also showed broad maxima as a function of kVp. The overall trends were as follows: (1) the Mo-Rh combination was superior to the Mo-Mo combination for all tissue compositions and kVps, and was generally superior to Rh-Rh except for kVp greater than 30. (2) The optimal kVp moves towards higher values for more glandular (dense) breast equivalent material. (3) At the optimal kVp, Mo-Rh is the superior combination, outperforming both Mo-Mo and Rh-Rh for imaging a 4-cm breast. The non-uniformity was higher by 60% for the non-grid condition, and this negated its dose efficiency advantage. In the non-grid mode the non-uniformity is quite dependent on the phantom dimensions and position, implying that scatter is a significant contributing factor. The Rh-Rh combination also yielded a higher non-uniformity than did Mo-Mo and Mo-Rh. The Mo-Rh target filter combination appears to be the best overall choice for imaging the 4-cm breast of arbitrary composition. In-spite of its superior dose efficiency, we recommend against using the non-grid technique until a better uniformity correction procedure is adopted.

Purpose: To provide a mathematically rigorous basis for the study of data sampling issues in fast Magnetic Resonance Imaging (MRI). Methods: The problem of reconstructing MRI data that have been sampled along arbitrary curves or trajectories in k-space was formulated in the language of distribution theory. With this formulation the nature of the associated point-spread function could then be related to conventional band-pass operators. Results: The definition of band-pass operators can be extended to curve-pass and sample-pass operators in a rigorous way. Each of these operators can be associated with point-spread functions that are distributional Fourier transforms of generalized functions defined by the given sampling region in k-space. For band-pass operators, the generalized function is related to the functional of line integration. For sample-pass operators, the generalized function is a sum of weighted Dirac delta functions. Breakthroughs: The problem of reconstructing an approximation for a two-dimensional function (an image) from two-, one- and zero-dimensional spatial frequency data sets has been shown to be well-posed. A rigorous relationship, exposing the nature of the approximation in terms of distributional Fourier transforms, between the different approaches to reconstruction has been found. Conclusions: The principle unsolved problem in reconstructing MRI data collected on non-Cartesian grids in k- space is the formulation of a sampling theorem similar to the Nyquist sampling theorem. The distributional formulation clearly reveals the connections between continuous and discrete distributional Fourier transforms and promises to be a useful tool for the formulation of a sampling theorem.

Numerical simulations are a valuable tool in the development of complex systems. They provide the ability to determine the effects of individual parameters on system functionality, and in the case of electronic systems, the ability to examine the system without the limitations introduced by electronic noise. The Thermoacoustic Computed Tomography (TACT) system under development was a natural candidate for numerical analysis. Early versions of the system exhibited exceptional promise, but final image quality was limited by a variety of confounding geometrical and electronic factors. The simulations described in this paper were used to generate the transducer signals that would theoretically be collected by the actual TACT imaging system when a sample was exposed to a pulse of electromagnetic radiation. The simulated data streams were then fed into the actual image reconstruction software to provide images of the 'virtual' phantoms. These images were analyzed and quantified to provide a measure of the system parameters responsible for the image blurs that limit system spatial resolution.

In computed tomography (CT), the elapsed time of a complete organ coverage is an important parameter for many clinical applications. This requirement has let to the development of both sub-second and multi-slice CT scanners. Examples of the multi-slice scanners include twin beam CT and quad CT (QCT). Although they are different in terms of number of detector rows employed in the design, they share the common property that the spacing between adjacent detector rows is essentially identical to the detector cell size, and the spacing remains constant regardless of the scanning conditions. This restriction has led to a sub-optimal sampling along the patient axis. In this paper, we present an optimized detector sampling scheme. In the proposed scheme, the detector spacing is dynamically adjusted based on the helical pitch. This allows an improved sampling pattern along the patient axis. In conjunction with the proposed sampling scheme, an helical interpolation reconstruction algorithm is proposed. Detailed analysis and computer simulations are performed to validate our approach.

Three dimensional shape, volume and depth of penetration of a skin lesion are significant factors for early diagnosis and prognosis of melanoma. An optical imaging instrument, Nevoscope is pursued in this work to image and reconstruct pigmented lesions, in three dimensions. The Nevoscope provides a set of planar projections of the pigmented inhomogeneity using transillumination. This paper presents a novel and simple algorithm to reconstruct the volume of the skin lesion from the optical projections acquired using the Nevoscope. The annular ring source of the Nevoscope injects light in the visible spectrum into the skin area surrounding the skin lesion. Light in the visible spectrum undergoes absorption and multiple scattering in the skin. Light photons, which are not extinct, are back scattered and re-emerge carrying information of the structure of the skin-lesion. The transilluminated photons are detected by a set of mirrors functioning as detectors to form 2D projections of the skin-lesion. The multiple-scattering phenomenon renders the inverse problem of solving for the volume of lesion non-linear. A diffusion- theory based approach along with the physics of light propagation in superficial layers of the skin results in a proposition of a hybrid model for solving the forward problem. An iterative non-linear inversion method is pursued to solve the inverse problem. Reconstruction of the lesion volume based on iterative algebraic reconstruction technique involves computation of 'weights' (contribution of a given voxel for a given photon path between a source and a detector) to calculate the forward and inverse solution for every iteration. A previously proposed model computes these weights as a product of two fluences. The first is the fluence calculated at a given voxel due to the annular ring source (forward fluence) and the second is the fluence calculated at the same voxel due to an imaginary point source at the detector (adjoint fluence). A diffusion theory based solution for calculation of the weights results in an under-estimation of the volume. This is because, diffusion theory is not accurate for calculation of fluence very near the source. Conventional X-ray CT like approach over-estimates the volume of reconstruction as it assumes an imaginary light source emanating photons in straight-line paths between source and detector from within the volume of the medium, and ignores scattering. The proposed hybrid method uses both the solutions from a diffusion theory based approach and the X-ray CT like approach to solve the forward and the inverse problems. Milk- gelatin phantoms and a skin mole were used to validate the algorithm. The problem is solved using three different approaches; a straight line X-ray CT like approach, a Diffusion Theory approach and finally the proposed hybrid approach. Quantitative results show the hybrid model reconstructs the phantoms with less error as opposed to the other two.

Three-dimensional (3D) visualization of tomosynthetically generated focal planes computed with conventional techniques is limited by distortions caused from nonuniform sampling and projection magnification. These distortions are inherent to diagnostic systems with fixed, off-axis sampling geometry using a proximal source of radiation. This paper describes a technique for significantly reducing these distortions by merging independently generated sets of orthogonally oriented tomosynthetic slices. Precise registration of corresponding points of the slice volumes is required by the merging process. This is achieved by compensating projective transformations of the respective slice stacks to correct for slice-specific variation in image magnification. The result is a relatively distortion-free 3D image comprised of isotropic (cubic) voxels.

In this paper we present an idea of combing data acquired from different scan geometry such as two mutually tilted circles and the circle-and-line orbits. The focus is on patching the shadow zone, a region where Radon data are not available from the single circular orbit, with data acquired from another (nonplanary) circular or linear orbits. Conventional two-step approaches are sophisticated and computationally intensive due to the fact that the entire cone beam projections are transformed into the Radon space (or some intermediate form) for patching, followed by an inverse transformation process. In contrast, by substituting the shadow zone with a shadow cone, we show that the patching process can be done in the three-dimensional (3D) Fourier space. Furthermore, we show that patching data in 2D Fourier space is possible for the case when it involves the circle-and-line orbit.

We have developed The Local Scan technology, which limited the exposure field to restrain X-ray, as a low exposure technology of the X-ray CT scanner. The system contains the high speed collimator to cover X-ray, control device of high speed collimator, correction and image re-construction from the projection data by local scan technique. We have evaluated the image quality and the effect of decreasing exposure dose. Our result is as follows. Image quality of inside limited exposure field is almost the same as conventional scan image. That is, even if local scan is used, image quality does not become worse. At least, 20% of the surface exposure dose of patient decreases. We have established the basic technology of local scan from these results.

Recent developments in CMOS image detectors are changing the way digital imaging is performed for many applications. The replacement of charge coupled devices (CCDs), with CMOS detectors is a desirable paradigm shift that will depend on the ability to match the high performance characteristics of CCDs. Digital X-ray imaging applications (chest X-ray, mammography) would benefit greatly from this shift because CMOS detectors have the following inherent characteristics: (1) Low operating power (5 - 10 times lower than CCD/processing electronics). (2) Standard CMOS manufacturing process (CCD requires special manufacturing). (3) On-chip integration of analog/digital processing functions (difficult with CCD). (4) Low Cost (5 - 10 times lower cost than CCD). The achievement of both low cost and low power is highly desirable for portable applications as well as situations where large, expensive X-ray imaging machines are not feasible (small hospitals and clinics, emergency medical vehicles, remote sites). Achieving this goal using commercially available components would allow rapid development of such digital X-ray systems as compared with the development difficulties incurred through specialized direct detectors and systems. The focus of this paper is to evaluate a CMOS image detector for medical X-ray applications and to demonstrate the results obtained from a prototype CMOS digital X-ray camera. Results from the images collected from this optically-coupled camera are presented for a particular lens, X-ray conversion screen, and demagnification factor. Further, an overview of the overall power consumption and cost of a multi-sensor CMOS mosaic compared to its CCD counterpart are also reported.

In this work, we propose a method for scatter compensation in SPECT imaging, by which we can estimate the scatter components in projections in high speed with a good accuracy. The method is that, at first, we estimate the scatter components in projections based on scatter response kernels by one time of OS-EM iteration, and then, subtract the estimated scatter components from the projections and complete the reconstruction by FBP method. The principle is that, the image corresponding to the scatter components in projections consist of almost low-frequency components of the activity distribution and the low-frequency components will converge faster than the high ones during iterative reconstruction. Therefore, we can estimate the low-frequency component image before the image converges with high-frequency ones and estimate the scatter components by re-projecting the low- frequency component image with scatter response kernels. The effects of the method were compared with dual- and triple- energy window methods using experimental measurements. The results show a good accuracy in estimated scatter components, a good uniformity of subtraction at both the center and side spheres and a good noise property can be acquired by proposed method compared with the dual- and triple-energy window methods.

Novel detector systems based on large area thin film electronics have the potential for an excellent image quality. However, the raw images derived from such detectors show various artifacts that have to be removed in real-time before the image is visualized to the user. Hence, superior flexibility in the applicable algorithms, sufficient system performance scalability, and a short processing delay are the key factors for the choice of a well suited processing platform due to the nature of the artifacts and due to the requirements of the medical applications. Further the very rapid progress in the available processing hardware has to be taken into account. A system comprising multiple nodes of a modern digital signal processor family can properly fulfill the key demands: high processing performance, strong data handling, full programmability and a processor family roadmap linked to state-of-the-art chip technology. Architecture principles, implementation aspects and first results derived from our new demonstrator system based on the modern SHARC DSP family are presented. The given first generation of the multiprocessor system corrects the basic artifacts at pixel rates of 50 Mpixel/s in 32-bit floating point arithmetic at a processing delay in the millisecond range.

Some aspects of the extremely complex structure of the pulses utilized in medical ultrasound pulse-echo imaging devices are investigated. The following features are examined, in many instances from a fresh point of view: diffraction, Fourier representation, directivity spectrum, propagation of pulse projections, causality, superluminality, and header wave component. Analytical results are underpinned by experimental findings with (single-element) wideband ultrasound transducers. The discussion is conducted within the context of linear fields.