Breast biopsy serves as the key diagnostic tool in the evaluation of breast masses for malignancy, yet the procedure affects patients physically and emotionally and may obscure results of future mammograms. Studies show that high quality ultrasound can distinguish a benign from malignant lesions with accuracy, however, it has proven difficult to teach and clinical results are highly variable. The purpose of this study is to develop a means to optimize an automated Computer Aided Imaging System (CAIS) to assess Level of Suspicion (LOS) of a breast mass. We examine the contribution of 15 object features to lesion classification by calculating the Wilcoxon area under the ROC curve, A<sub>W</sub>, for all combinations in a set of 146 masses with known findings. For each interval A, the frequency of appearance of each feature and its combinations with others was computed as a means to find an “optimum” feature vector. The original set of 15 was reduced to 6 (area, perimeter, diameter ferret Y, relief, homogeneity, average energy) with an improvement from A<sub>w</sub>=0.82∓0.04 for the original 15 to A<sub>w</sub>=0.93∓0.02 for the subset of 6, p=0.03. For comparison, two sub-specialty mammography radiologists also scored the images for LOS resulting in A<sub>z</sub> of 0.90 and 0.87. The CAIS performed significantly higher, p=0.02.
Research studies indicate that careful application of breast ultrasound is capable of reducing the number of unnecessary biopsies by 40% with potential cost savings of as much as $1 billion per year in the U.S. A well-defined rule-based system has been developed for scoring the Level of Suspicion (LOS) based on parameters describing the ultrasound appearance of breast lesion. Acceptance and utilization of LOS is increasing but it has proven difficult to teach the method and many radiologists have felt uncomfortable with the number of benign and malignant masses that overlap in appearance. In practice, the quality of breast ultrasound is highly operator dependent, it is often difficult to reproduce a finding and there is high variability of lesion description and assessment between radiologists. The goal of this research is to improve the uniformity and accuracy of applying the LOS scheme by automatically detecting, analyzing and comparing breast masses using sophisticated software developed for satellite imagery applications. The aim is to reduce biopsies on the masses with lower levels of suspicion, rather that increasing the accuracy of diagnosis of cancers, which will require biopsy anyway. In this paper we present our approach to develop a system to process, segment, analyze and classify medical images based on information content. A feasibility study was completed in a digital database of biopsy-proven image files from 46 women retrieved chronologically from our image library. Segmentation and classification were sufficiently accurate to correctly group all benign cystic masses, all benign solid masses and all solid malignant masses. The image analysis, computer-aided detection and image classification software system Image Companion developed by Almen Laboratories, Inc. was used to achieve the presented results.
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<SUP>TM</SUP> (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<SUP>TM</SUP> 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.
This paper describes a novel area detector for direct conversion and readout of the x-ray energy that eliminates multiple conversions and coupling stages which degrade performance. The pixel array and readout electronics are fabricated on the same piece of silicon. The detector consists of a uniform layer (approximately 300 micrometers) of amorphous selenium alloy vapor-deposited on an electronic readout array fabricated using conventional complementary metal oxide semiconductor (CMOS). The CMOS array features 66 micrometer pixels in a 1024 X 832 array providing a 5.5 X 6.75 cm image area. Each pixel has active circuitry including signal amplification, pixel selection and reset, while peripheral circuitry on one end of the array provides shift registers, sample and hold and multiplexing. The CMOS readout array was fabricated at a standard facility on a 10-cm diameter silicon wafer using 2 micrometer CMOS process. Fifteen separate image sensors were assembled for evaluation in a 3 X 5 format to provide a 20 X 27 cm composite field of view. Missing data between sensors is recovered by acquiring three sub-exposures, between which the array is translated diagonally approximately 2 mm. Total exposure time for an average breast is less than one second. Conversion efficiency was found to be approximately 120 electrons per absorbed x-ray (19 keV average). Electronic readout noise was measured to be 2.4 ADU corresponding to approximately 500 electrons. Detective quantum efficiency was found to be 0.65 at low spatial frequency (0.25 lp/mm) and at 0.2 at high spatial frequency (8 lp/mm) for x-ray fluence ranging from 5 - 35 mR. Images of an ACR phantom show visualization of all of the fibers, specks and masses when displayed with a linear lookup table on a high-resolution monitor. These studies demonstrated that there is a slight but measurable image retention evident as 'ghost' images. The two most effective means to reduce this effect are flushing the sensors with infrared light or x-rays between exposures and reversing the applied voltage on the selenium layer. A number of improvements designed to increase sensitivity and reduce noise also have been identified and are being implemented. Sample images were acquired from four volunteer human subjects at exposure factors identical to their film-screen mammograms. The results suggest that the detector performance is suitable for further clinical investigation.
A new prototype imaging system based on ultrasound transmission through the object of interest -- acoustical holography -- was developed which incorporates significant improvements in acoustical and optical design. This system is being evaluated for potential clinical application in the musculoskeletal system, interventional radiology, pediatrics, monitoring of tumor ablation, vascular imaging and breast imaging. System limiting resolution was estimated using a line-pair target with decreasing line thickness and equal separation. For a swept frequency beam from 2.6 - 3.0 MHz, the minimum resolution was 0.5 lp/mm. Apatite crystals were suspended in castor oil to approximate breast microcalcifications. Crystals from 0.425 - 1.18 mm in diameter were well resolved in the acoustic zoom mode. Needle visibility was examined with both a 14-gauge biopsy needle and a 0.6 mm needle. The needle tip was clearly visible throughout the dynamic imaging sequence as it was slowly inserted into a RMI tissue-equivalent breast biopsy phantom. A selection of human images was acquired in several volunteers: a 25 year-old female volunteer with normal breast tissue, a lateral view of the elbow joint showing muscle fascia and tendon insertions, and the superficial vessels in the forearm. Real-time video images of these studies will be presented. In all of these studies, conventional sonography was used for comparison. These preliminary investigations with the new prototype acoustical holography system showed favorable results in comparison to state-of-the-art pulse-echo ultrasound and demonstrate it to be suitable for further clinical study. The new patient interfaces will facilitate orthopedic soft tissue evaluation, study of superficial vascular structures and potentially breast imaging.
Breast ultrasound is a valuable adjunct to mammography but is limited by a very small field of view, particularly with high-resolution transducers necessary for breast diagnosis. We have been developing an ultrasound system based on a diffraction tomography method that provides slices through the breast on a large 20-cm diameter circular field of view. Eight to fifteen images are typically produced in sequential coronal planes from the nipple to the chest wall with either 0.25 or 0.5 mm pixels. As a means to simplify the interpretation of this large set of images, we report experience with 3D life-sized displays of the entire breast of human volunteers using a digital holographic technique. The compound 3D holographic images are produced from the digital image matrix, recorded on 14 X 17 inch transparency and projected on a special white-light viewbox. Holographic visualization of the entire breast has proved to be the preferred method for 3D display of ultrasound computed tomography images. It provides a unique perspective on breast anatomy and may prove useful for biopsy guidance and surgical planning.
Tungsten (W) target x-rays tubes are being studied for use in digital mammography to improve x-ray flux, reduce noise and increase tube heat capacity. A parametric model was developed for digital mammography to evaluate optimization of x-ray spectra for a particular sensor. The model computes spectra and mean glandular doses (MGD) for combinations of W target, beam filters, kVp, breast type and thickness. Two figures of merit were defined: (signal/noise)<SUP>2</SUP>/MGD and spectral quantum efficiency; these were computed as a means to approach optimization of object contrast. The model is derived from a combination of classic equations, XCOM from NBS, and published data. X-ray spectra were calculated and measured for filters of Al, Sn, Rh, Mo and Ag on a Eureka tube. (Signal/noise)<SUP>2</SUP>/MGD was measured for a filtered W target tube and a digital camera employing CsI scintillator optically coupled to a CCD for which the detective quantum efficiency (DQE) was known. A 3-mm thick acrylic disk was imaged on thickness of 3-8 cm of acrylic and the results were compared to the predictions of the model. The relative error between predicted and measured spectra was +/- 2 percent from 24 to 34 kVp. Calculated MGD as a function of breast thickness, half-value layer and beam filter compares very well to published data. Best performance was found for the following combinations: Mo filter with 30 mm breast, Ag filter with 45 mm, Sn filter for 60 mm, and Al filter for 75 mm thick breast. The parametric model agrees well with measurement and provides a means to explore optimum combinations of kVp and beam filter. For a particular detector, this data may be used with the DQE to estimate total system signal-to-noise ratio for a particular imaging task.