The purpose of this work is to explore image quantitation in small-animal positron emission tomography (PET) as a tool for following proliferation and malignant transformation of mammary intraepithelial neoplasia outgrowths in vivo in laboratory mice. The expected application of the work is preclinical evaluations of breast cancer therapeutics. For nonpalpable or prepalpable disease, current practice involves large cohorts of mice with groups sacrificed at each time point. Because of the substantial variability in tumor development, numerous mice are needed at each time point to obtain statistical power. In vivo imaging techniques have the ability to follow a single mouse over multiple time points in longitudinal studies, and therefore reduce the number of mice needed for evaluation. Longitudinal studies offer an additional increase in statistical power by being able to control for the condition of disease (e.g. tumor size) at onset of treatment. Critical to the success of this approach is the ability to extract meaningful quantitative markers of disease. The study reported here describes a computational approach to extracting quantitative markers of disease progression and proliferation in longitudinal PET studies, and an analysis of the increase in statistical power due to temporal correlations in the extracted markers.

The Variable Resolution X-ray (VRX) CT system, developed at the UTHSC, Memphis, has the potential for use in animal imaging. Animal models of tumor progression and pharmacological impact are becoming increasingly important in understanding the molecular and mechanistic basis of tumor development. In general, CT-imaging offers several advantages in animal research: a fast throughput of seconds to minutes reducing the physiological stress animals are exposed to, and it is an inexpensive modality affordable to many animal laboratories. We are developing the VRX CT scanner as a non-invasive imaging modality to measure tumor volume, progression, and metastasis. From the axial images taken by the VRX CT-scanner, tumor area was measured and the tumor volume was calculated. Animals were also imaged using an optical liquid nitrogen-cooled CCD camera to detect tumor fluorescence. A simple image fusion with a planner x-ray image was used to ascertain the position of the tumors, animals were then sacrificed the tumors excised, and the tumor volume calculated by physical measurements. Furthermore, using a specially designed phantom with three spheres of different volumes, we demonstrated that our system allowed us to estimate the volume with up to 10% accuracy; we expect this to increase dramatically in the next few months.

Photodynamic therapy (PDT) is a novel cancer treatment modality where the therapeutic action is controlled by light and the potency of the photosensitizer used. Development of new potent photosensitizers (PS) for clinical applications requires that the PDT effects are maximized while minimizing dark cytotoxicity. The dark toxicity of photosensitizers is generally confirmed using cell lines. Photososensitizers that appear promising from in vitro assays need further investigations under in vivo conditions. As in vivo MR methods have the potential to provide information on the tumor status, they can be very effective tools to study dark toxicity of tumors. The tumor produced on the mouse foot dorsum was tested on two newly synthesized photosensitizers along with Photofrin as a control. The MR studies consisted of serial ³¹P spectral measurements both before and after PS injection. The results show significant changes in the tumor metabolism with increased inorganic phosphate while using new photosensitizers. However these changes slowly approached control levels several hours later. The studies performed while using Photofrin did not show any significant changes indicating minimal or no dark cytotoxicity. Similar studies performed on normal tissue such as the muscle indicated that the energy metabolism was minimally compromised. Our studies demonstrate that the effects of dark cytotoxicity can be observed by ³¹P MR. The growth profiles of tumors treated with PS alone indicate that the metabolic changes are temporary and do not interfere with the tumor growth. The studies suggest that MR is a new method of monitoring the effect of PS administered toxicity in an in vivo model.

The objective is to compare power Doppler sonography (PD) and dynamic contrast-enhanced MRI (MR) and PET SUV in assessing the vascularity of benign and malignant breast lesions. Sixty two patients with 89 lesions (59 malignant lesions, 30 benign lesions) were evaluated by PD, MRI (MR) and PET SUV prior to surgery. Each imaging modality was evaluated independently. Lesion vascularity on PD was graded as avascular, intermediately vascular, or hypervascular. On MR, degree of maximal enhancement (minimal, moderate, or marked) and the kinetic pattern of enhancement (persistent, plateau, washout) were graded separately. For malignant lesions, PET SUV values were correlated with MRI kinetics. Gamma variable analysis was performed to assess the degree of correlation. Of the 89 lesions 44 were invasive ductal carcinoma, 2 were intraductal cancers, 6 were invasive lobular carcinoma, and 7 were invasive cancers with mixture of ductal and lobular features. There was a high degree of correlation between degree of maximal enhancement and enhancement kinetics on MRI (G=0.074, p<0.0001). The correlation between CDS and degree of maximal gadolinium enhancement was moderate (G=0.57, p=0.02). The correlation between PD vascularity and gadolinium enhancement kinetics was weak (G=0.37, p=0.12). Invasive malignancy demonstrated moderate correlation between SUV and MRI kinetics (G=0.64, p=0.14). There is a variable degree of correlation between various imaging modalities in assessing breast lesion vascularity. Further evaluation on the relationship between subjective reader assessment and objective quantitative image analysis is required to elucidate the differences in these measures of breast tumor physiology. This work was supported in part by the NIH grant P01CA085424-03.

Multimodality imaging has gained momentum in radiation therapy planning and image-guided treatment delivery. Specifically, computed tomography (CT) and magnetic resonance (MR) imaging are two complementary imaging modalities often utilized in radiation therapy for visualization of anatomical structures for tumour delineation and accurate registration of image data sets for volumetric dose calculation. The development of a multimodal contrast agent for CT and MR with prolonged in vivo residence time would provide long-lasting spatial and temporal correspondence of the anatomical features of interest, and therefore facilitate multimodal image registration, treatment planning and delivery. The multimodal contrast agent investigated consists of nano-sized stealth liposomes encapsulating conventional iodine and gadolinium-based contrast agents. The average loading achieved was 33.5 ± 7.1 mg/mL of iodine for iohexol and 9.8 ± 2.0 mg/mL of gadolinium for gadoteridol. The average liposome diameter was 46.2 ± 13.5 nm. The system was found to be stable in physiological buffer over a 15-day period, releasing 11.9 ± 1.1% and 11.2 ± 0.9% of the total amounts of iohexol and gadoteridol loaded, respectively. 200 minutes following in vivo administration, the contrast agent maintained a relative contrast enhancement of 81.4 ± 13.05 differential Hounsfield units (ΔHU) in CT (40% decrease from the peak signal value achieved 3 minutes post-injection) and 731.9 ± 144.2 differential signal intensity (ΔSI) in MR (46% decrease from the peak signal value achieved 3 minutes post-injection) in the blood (aorta), a relative contrast enhancement of 38.0 ± 5.1 ΔHU (42% decrease from the peak signal value achieved 3 minutes post-injection) and 178.6 ± 41.4 ΔSI (62% decrease from the peak signal value achieved 3 minutes post-injection) in the liver (parenchyma), a relative contrast enhancement of 9.1 ± 1.7 ΔHU (94% decrease from the peak signal value achieved 3 minutes post-injection) and 461.7 ± 78.1 ΔSI (60% decrease from the peak signal value achieved 5 minutes post-injection) in the kidney (cortex) of a New Zealand white rabbit. This multimodal contrast agent, with prolonged in vivo residence time and imaging efficacy, has the potential to bring about improvements in the fields of medical imaging and radiation therapy, particularly for image registration and guidance.

Micro-CT will have a profound influence on the accumulation of anatomical and physiological phenotypic changes in natural and transgenetic mouse models. Longitudinal studies will be greatly facilitated, allowing for a more complete and accurate description of events if in-vivo studies are accomplished. The purpose of the ongoing project is to establish a feasible and reproducible setup for in-vivo mouse lung micro-computed tomography (μCT). We seek to use in-vivo respiratory-gated μCT to follow mouse models of lung disease with subsequent recovery of the mouse. Methodologies for optimizing scanning parameters and gating for the in-vivo mouse lung are presented. A Scireq flexiVent ventilated the gas-anesthetized mice at 60 breaths/minute, 30 cm H20 PEEP, 30 ml/kg tidal volume and provided a respiratory signal to gate a Skyscan 1076 μCT. Physiologic monitoring allowed the control of vital functions and quality of anesthesia, e.g. via ECG monitoring. In contrary to longer exposure times with ex-vivo scans, scan times for in-vivo were reduced using 35μm pixel size, 158ms exposure time and 18μm pixel size, 316ms exposure time to reduce motion artifacts. Gating via spontaneous breathing was also tested. Optimal contrast resolution was achieved at 50kVp, 200μA, applying an aluminum filter (0.5mm). There were minimal non-cardiac related motion artifacts. Both 35μm and 1μm voxel size images were suitable for evaluation of the airway lumen and parenchymal density. Total scan times were 30 and 65 minutes respectively. The mice recovered following scanning protocols. In-vivo lung scanning with recovery of the mouse delivered reasonable image quality for longitudinal studies, e.g. mouse asthma models. After examining 10 mice, we conclude μCT is a feasible tool evaluating mouse models of lung pathology in longitudinal studies with increasing anatomic detail available for evaluation as one moves from in-vivo to ex-vivo studies. Further developments include automated bronchial tree segmentation and airway wall thickness measurement tools. Improvements in Hounsfield unit calibration have to be performed when the interest of the study lies in determining and quantifying parenchymal changes and rely on estimating partial volume contributions of underlying structures to voxel densities.

Mouse models are important for pulmonary research to gain insight into structure and function in normal and diseased states, thereby extending knowledge of human disease conditions. The flexibility of human disease induction into mice, due to their similar genome, along with their short gestation cycle makes mouse models highly suitable as investigative tools. Advancements in non-invasive imaging technology, with the development of micro-computed tomography (μ-CT), have aided representation of disease states in these small pulmonary system models. The generation ofμCT 3D airway reconstructions has to date provided a means to examine structural changes associated with disease. The degree of accuracy ofμCT is uncertain. Consequently, the reliability of quantitative measurements is questionable. We have developed a method of sectioning and imaging the whole mouse lung using the Large Image Microscope Array (LIMA) as the gold standard for comparison. Fixed normal mouse lungs were embedded in agarose and 250μm sections of tissue were removed while the remaining tissue block was imaged with a stereomicroscope. A complete dataset of the mouse lung was acquired in this fashion. Following planar image registration, the airways were manually segmented using an in-house built software program PASS. Amira was then used render the 3D isosurface from the segmentations. The resulting 3D model of the normal mouse airway tree developed from pathology images was then quantitatively assessed and used as the standard to compare the accuracy of structural measurements obtained from μ-CT.

In this study, we explored the potential of diffuse optical tomography for brain oximetry and describe our efforts towards imaging hemodynamic changes in rat brains during kainic-acid (KA) induced seizures. Using electrophysiological techniques we first showed that KA induces a pronounced transient hypotension in urethane anesthetized rats that is coincident with seizure activity beginning in ventral and spreading to dorsal hippocampus. We observed sustained increases in vagus and sympathetic activity during generalized limbic seizure activity, which alters blood pressure regulation and heart rhythms. Subsequently, we used optical tomographic methods to study KA induced seizures in anesthetized animals to better define the hemodynamic cerebral vascular response. We observed a lateralized increase in deoxyhemoglobin after KA injection at the time when the blood pressure (BP) was decreased. By contrast, injection of phenylephrine produced a symmetric global increase in total hemoglobin. These findings indicate that our instrument is sensitive to the local hemodynamics, both in response to a global increase in blood pressure (phenylephrine injection) and a lateralized decrease in oxyhemoglobin produced by an asymmetric response to KA; a response that may be critically important for severe autonomic nervous system alterations during seizures. The results of this study provide the impetus for combining complimentary modalities, imaging and electrophysiological, to ultimately gain a better understanding of the underlying physiology of seizure activity in the rat.

Given the nature of pulmonary embolism (PE), timely and accurate diagnosis is critical. Contrast enhanced high-resolution CT images allow physicians to accurately identify segmental and sub-segmental emboli. However, it is also important to assess the effect of such emboli on the blood flow in the lungs. Expanding upon previous research, we propose a method for 3D visualization of lung perfusion. The proposed method allows users to examine perfusion throughout the entire lung volume at a single glance, with areas of diminished perfusion highlighted so that they are visible independent of the viewing location. This may be particularly valuable for better accuracy in assessing the extent of hemodynamic alterations resulting from pulmonary emboli. The method also facilitates user interaction and may help identify small peripheral sub-segmental emboli otherwise overlooked. 19 patients referred for possible PE were evaluated by CT following the administration of IV contrast media. An experienced thoracic radiologist assessed the 19 datasets with 17 diagnosed as being positive for PE with multiple emboli. Since anomalies in lung perfusion due to PE can alter the distribution of parenchymal densities, we analyzed features collected from histograms of the computed perfusion maps and demonstrate their potential usefulness as a preliminary test to suggest the presence of PE. These histogram features also offer the possibility of distinguishing distinct patterns associated with chronic PE and may even be useful for further characterization of changes in perfusion or overall density resulting from associated conditions such as pneumonia or diffuse lung disease.

This study was aimed to evaluate a morphology-based approach for prediction of postoperative forced expiratory volume in one second (FEV1) after lung resection from preoperative CT scans. Fifteen Patients with surgically treated (lobectomy or pneumonectomy) bronchogenic carcinoma were enrolled in the study. A preoperative chest CT and pulmonary function tests before and after surgery were performed. CT scans were analyzed by prototype software: automated segmentation and volumetry of lung lobes was performed with minimal user interaction. Determined volumes of different lung lobes were used to predict postoperative FEV₁ as percentage of the preoperative values. Predicted FEV₁ values were compared to the observed postoperative values as standard of reference. Patients underwent lobectomy in twelve cases (6 upper lobes; 1 middle lobe; 5 lower lobes; 6 right side; 6 left side) and pneumonectomy in three cases. Automated calculation of predicted postoperative lung function was successful in all cases. Predicted FEV₁ ranged from 54% to 95% (mean 75% ± 11%) of the preoperative values. Two cases with obviously erroneous LFT were excluded from analysis. Mean error of predicted FEV₁ was 20 ± 160 ml, indicating absence of systematic error; mean absolute error was 7.4 ± 3.3% respective 137 ± 77 ml/s. The 200 ml reproducibility criterion for FEV₁ was met in 11 of 13 cases (85%). In conclusion, software-assisted prediction of postoperative lung function yielded a clinically acceptable agreement with the observed postoperative values. This method might add useful information for evaluation of functional operability of patients with lung cancer.

A computational model of soft tissue mechanics and air flow has been developed with the aim of linking computed tomography measures of ventilation distribution to subject-specific predictions in image-based geometric (finite element) models of the lung and airway tree. Computational techniques that can deal with anatomical detail and spatially-distributed non-linear material properties have been used to couple solution of parenchymal soft tissue mechanics in an anatomically-based model of the ovine lung to predictions of flow and pressure in an embedded model of the ovine airway tree. The lung is modeled as a homogeneous, compressible, non-linear elastic body. Using equations for large deformation mechanics, the change in geometry of the lung is simulated at static inflation pressures from 25 to 0 cmH₂O. Multi-detector row computed tomography imaging has been used to define the model geometry (lung and airway), to define the movement of the model lung surface during inflation, and for measurements of internal material point displacements for comparison with the predicted internal displacements of the model. This preliminary model predicts airway bifurcation point displacements that are generally in agreement with imaged displacements (total RMS error for all bifurcation points is < 4 mm from 25 to 0 cm H₂O). Further development of the model will provide a predictive link between subject-specific anatomical and functional information.

Stable Xenon (Xe) gas has been used as an imaging agent for decades in its radioactive form, is chemically inert, and has been used as a ventilation tracer in its non radioactive form during computerized tomography (CT) imaging. Magnetic resonance imaging (MRI) using hyperpolarized Helium (He) gas and Xe has also emerged as a powerful tool to study regional lung structure and function. However, the present state of knowledge regarding intra-bronchial Xe and He transport properties is incomplete. As the use of these gases rapidly advances, it has become critically important to understand the nature of their transport properties and to, in the process, better understand the role of gas density in general in determining regional distribution of respiratory gases. In this paper, we applied the custom developed characteristic-Galerkin finite element method, which solves the three-dimensional (3D) incompressible variable-density Navier-Stokes equations, to study the transport of Xe and He in the CT-based human lung geometries, especially emulating the washin and washout processes. The realistic lung geometries are segmented and reconstructed from CT images as part of an effort to build a normative atlas (NIH HL-064368) documenting airway geometry over 4 decades of age in healthy and disease-state adult humans. The simulation results show that the gas transport process depends on the gas density and the body posture. The implications of these results on the difference between washin and washout time constants are discussed.

In this study we developed a new reconstruction kernel specifically for chest CT imaging. An experimental flat-panel CT scanner was used on large dogs to produce 'ground-truth’ reference chest CT images. These dogs were also examined using a clinical 16-slice CT scanner. We concluded from the dog images acquired on the clinical scanner that the loss of subtle lung structures was due mostly to the presence of the background noise texture when using currently available reconstruction kernels. This qualitative evaluation of the dog CT images prompted the design of a new recon kernel. This new kernel consisted of the combination of a low-pass and a high-pass kernel to produce a new reconstruction kernel, called the 'Hybrid’ kernel. The performance of this Hybrid kernel fell between the two kernels on which it was based, as expected. This Hybrid kernel was also applied to a set of 50 patient data sets; the analysis of these clinical images is underway. We are hopeful that this Hybrid kernel will produce clinical images with an acceptable tradeoff of lung detail, reliable HU, and image noise.

Though an approach to image the electrical activity of neurons directly by detecting phase shifts in MRI was first reported in 1991, results to-date remain equivocal due to the low signal-to-noise ratio. The objective of this work was to develop a stimulus-presentation and data acquisition strategy specially geared to detect phase-dispersion effects of neuronal currents within 10-100 ms following stimulation. The key feature is to set the repeated MR data acquisition time TR and the stimulus presentation interval (TI) slightly different from each other so that the time at which images are acquired shifts gradually from one acquisition to the next with respect to stimulus onset. For example, at TR=275ms and 4 Hz stimulus presentation (TI=250ms), initial synchronization of the stimulus onset and MR acquisition would result in the first image being acquired at a latency of 0± (temporal width of data acquisition window), second image at a latency of 25ms, third image at a latency of 50ms and so on up to a latency of 250ms, at which time the stimulus and data acquisition times would become re-synchronized to once again acquire an image at latency=0. Human data were acquired on a 1.5T GE EXCITE scanner from two 8mm thick contiguous slices bracketing the calcarine fissure during a checkerboard flashing at 4 Hz. Preliminary results show activity in the visual cortex at latencies consistent with EEG studies, suggesting the potential of this methodology to image neural activity directly.

In this paper, we have investigated local spatial couplings in the human brain by applying nonlinear dynamical techniques on fMRI data. We have recorded BOLD-contrast echo-planar fMRI data along with high-resolution T1-weighted anatomical images from the resting brain of healthy human subjects and performed physiological correction on the functional data. The corrected data from resting subjects is spatially embedded into its phase space and the largest Lyapunov exponent of the resulting attractor is calculated and whole slice maps are obtained. In addition, we segment the high-resolution anatomical image and obtain a down sampled mask corresponding to gray and white matter, which is used to obtain mean indices of the exponents for both the tissues separately. The results show the existence of local couplings, its tissue specificity (more local coupling in gray matter than white matter) and dependence on the size of the neighborhood (larger the neighborhood, lesser the coupling). We believe that these techniques capture the information of a nonlinear and evolving system like the brain that may not be evident from static linear methods. The results show that there is evidence of spatio-temporal chaos in the brain, which is a significant finding hitherto not reported in literature to the best of our knowledge. We try to interpret our results from healthy resting subjects based on our knowledge of the native low frequency fluctuations in the resting brain and obtain a better understanding of the local spatial behavior of fMRI. This exploratory study has demonstrated the utility of nonlinear dynamical techniques like spatial embedding in analyzing fMRI data to gain meaningful insights into the working of human brain.

A novel approach for noninvasively tracing brain white matter fiber tracts is presented using diffusion tensor magnetic resonance imaging (DT-MRI) data. This technique is based on performing anisotropic diffusion simulations over a series of overlapping three dimensional diffusion kernels that cover only a small portion of the human brain volume and are geometrically centered upon selected starting voxels where a seed is placed. The simulations conducted over diffusion kernels are initiated from those starting voxels and are utilized to construct diffusion fronts. The fiber pathways are determined by evaluating the distance and orientation from fronts to their corresponding diffusion seed voxels. Synthetic and real DT-MRI data are employed to demonstrate the tracking scheme. It is shown that the synthetic tracts can be accurately replicated, while several major white matter fiber pathways in the human brain can be reproduced noninvasively as well. Since the diffusion simulation makes use of the entire diffusion tensor data, including both the magnitude and orientation information, the proposed approach enhances robustness and reliability in DT-MRI based fiber reconstruction.

In the absence of ground truth, there are very few methods available to evaluate the accuracy of a specific tracking algorithm or the various data acquisition protocols for DTI-tractography. The objective of this work was to develop methodology, based on tract-length histograms, that could be used to evaluate whole-brain tractography with data acquired under different conditions for a given subject, for example six versus 25 gradient directions, or use of an 8-element phased array versus quadrature head-coil. Whole-brain DTI data were acquired from six healthy normal human volunteers on a 1.5 T GE scanner at TR=10.3s, field-of-view 26cm, 128x128 matrix, 28 contiguous 4mm thick slices from 25 isotropic gradient directions with b=1000s/mm², one b=0 acquisition, and number of excitations (NEX)=1 for a total acquisition time of 3min 53s. Similarly, four sets of data were acquired from 6 non-colinear directions and combined with two b=0 acquisitions to equalize the time for 25 and 6-directions acquisitions. The tract-length histograms clearly show that at equal acquisition time, there are more long tracts in the 25-direction acquisition than the 6-direction acquisition, suggesting better estimation of the tensor with 25 directions. Tract-counts above a threshold provide an objective index to evaluate tractography. Also a comparison of the two coils shows a higher tract-count for long tracts with the 8-element coil, consistent with the demonstrated higher sensitivity and higher signal-to-noise ratio for EPI acquisitions by the 8-element coil.

In conventional diffusion tensor imaging (DTI) based on magnetic resonance data, each voxel is assumed to contain a single component having diffusion properties that can be fully represented by a single tensor. In spite of its apparent lack of generality, this assumption has been widely used in clinical and research purpose. This resulted in situations where correct interpretation of data was hampered by mixing of components and/or tractography. Even though this assumption can be valid in some cases, the general case involves mixing of components resulting in significant deviation from the single tensor model. Hence, a strategy that allows the decomposition of data based on a mixture model has the potential of enhancing the diagnostic value of DTI. This work aims at developing a stable solution for the most general problem of multi-component modeling of diffusion tensor imaging data. This model does not include any assumptions about the nature or volume ratio of any of the components and utilizes a projection pursuit based strategy whereby a combination of exhaustive search and least-squares estimation is used to estimate 1D projections of the solution. Then, such solutions are combined to compute the multidimensional components in a fast and robust manner. The new method is demonstrated by both computer simulations and real diffusion-weighted data. The preliminary results indicate the success of the new method and its potential to enhance the interpretation of DTI data sets.

MRI Neuroimaging provides a rich source of image content including structural (MRI, Diffusion DTI), functional (fMRI, Perfusion ASL), and metabolic (MRS) information. Today MRI capabilities allow to acquire these imaging techniques in one session in most cases. In order to be of diagnostic value, the immense and diverse data needs to be (i) automatically post-processed to extract the relevant information, e.g. 3D brain maps from 4D fMRI, and to be (ii) fused and visualized to correlate the voxel-based findings. The purpose of this study is to demonstrate the feasibility of automatic relevant information retrieval and fusion of MRI, fMRI, DTI, ASL, and MRS data of a pediatric population into a single semantic data representation. By using advanced imaging, we may able to detect a larger spectrum of abnormalities in the neonatal brain. Each imaging application, provides unique information about the physiology (fMRI, ASL), the anatomy (DTI), and the biochemistry (MRS) of the newborn brain in relation to normal development and brain injury. By being able to integrate this technology, we will be able to combine biochemical, physiologic and anatomic information which can provide unique insight about not only the normal development of the brain, but also injury of the neonatal brain.

Purpose: To validate a computer algorithm for measuring myocardial infarct size on gadolinium enhanced MR images. The results of computer infarct sizing are studied on phase-sensitive and magnitude imaging against a histopathology reference. Materials and Methods: Validations were performed in 9 canine myocardial infarctions determined by triphenyltetrazolium chloride (TTC). The algorithm analyzed the pixel intensity distribution within manually traced myocardial regions. Pixels darker than an automatically determined threshold were first excluded from further analysis. Selected image features were used to remove false positive regions. A threshold 50% between bright and dark regions was then used to minimize partial volume errors. Post-processing steps were applied to identify microvascular obstruction. Both phase sensitive and magnitude reconstructed MR images were measured by the computer algorithm in units of % of the left ventricle (LV) infarction and compared to TTC. Results: Correlations of MR and TTC infarct size were 0.96 for both phase sensitive and magnitude imaging. Bland Altman analysis showed no consistent bias as a function of infarct size. The average error of computer infarct sizing was less than 2% of the LV for both reconstructions. Fixed intensity thresholding was less accurate compared to the computer algorithm. Conclusions: MR can accurately depict myocardial infarction. The proposed computer algorithm accurately measures infarct size on contrast-enhanced MR images against the histopathology reference. It is effective for both phase-sensitive and magnitude imaging.

Myocardial microcirculation disturbances often precede angiographically visible of narrowing large epicardial coronary arteries and associated symptoms. Clinical tomographic imaging cannot resolve the microcirculation, hence an indirect method of quantitating microvascular disturbances in those images must be developed. We propose that such an indirect method can be based on the characterization of the spatial heterogeneity of myocardial intravascular blood volume. We evaluated the relationship of multi-resolution, nested multi Region-of-Interest (ROI) analysis of EBCT images to the actual intravascular volume of microvascular branches as measured directly with micro-CT images in the same myocardial regions. We selectively altered the intravascular volume of vessels by injecting 30, 100, 200 or 300μm diameter microspheres into anesthetized pigs’ LAD coronary arteries prior to EBCT scanning during contrast injection. The heart was then harvested and the LAD coronary artery was infused with Microfil polymer. An approximately 2cm³ transmural “biopsy” of the same ROI within the myocardium analyzed in the EBCT images was scanned by micro-CT resulting in a 3D image of 20μm cubic voxels. Myocardial opacification was measured in both the EBCT and micro-CT images. The EBCT and micro-CT images were analyzed with the nested multi ROI method which provides an index of spatial heterogeneity of intramyocardial blood volume in terms of the linear relationship between the logarithms of the coefficient of variation within the data obtained at any one size of the ROI, and the logarithm of the volume of that selected ROI. The minimum ROI volume in the EBCT analysis was 8.96 mm³ and for the micro-CT it was 0.07 mm³. There is linear correlation when EBCT and micro-CT image CT gray-scale numbers are plotted as Log (standard deviation/mean) against Log (Volume of ROI). The results show that the slopes and offsets of the EBCT-based and micro-CT-based regression lines were indistinguishable. Moreover, when a fraction of microvessels of selected diameter was embolized, the change in the resulting regression line was characteristic for that diameter. In summary, the EBCT-based analysis spatial heterogeneity of myocardial blood volume can be extrapolated to describe the spatial distribution of the microcirculatory branching geometry in terms of intra segmental lumen volume.

We conduct experiments to look at the in-vivo cardiac motion during systole, to visualize heart contraction, and to examine the clinical usefulness. Our model-based technique incorporates subject-specific modeling, motion analysis and the extraction of clinically relevant parameters within one framework. Previous bi-ventricular model based method could only handle up to the mid-ventricles and have a few test-subjects. Our parameterized model includes the LV, RV and up to the basal area for full ventricular motion study. Finite element methods capture cardiac motion by tracking the material points from tagged Magnetic Resonance (MR) images. A number of experiments from ten subjects are evaluated and analyzed. We tested subject several times and compared the resulting parameters to ensure the reproducibility and deviations. The resulting parameters can be used to describe the cardiac motion of normal subjects. The patterns of normal subjects were derived from experiments. While significant shape and motion variations were apparent in normal subjects, the quantitative analysis show typical patterns. Generally, the basal area moves downwards and the apical area contracts towards the cavity. The principal strain analysis describes the directions and magnitudes of maximum shortening, and maximum thickening.

We present current research in which both left and right ventricular deformation is estimated from tagged cardiac magnetic resonance imaging using volumetric deformable models constructed from nonuniform rational B-splines (NURBS). The four model types considered include Cartesian-based NURBS models with both a cylindrical and prolate-spheroidal parameterization, prolate spheroidal-based NURBS models with a prolate-spheroidal parameterization, and cylindrical-based NURBS models with a cylindrical parameterization. For each frame subsequent to end-diastole, a NURBS model is constructed by fitting two surfaces with the same parameterization to the corresponding set of epicardial and endocardial contours from which a volumetric model is created. Using normal displacements of the three sets of orthogonal tag planes as well as displacements of contour/tag line intersection points and tag plane intersection points, one can solve for the optimal homogeneous coordinates, in a weighted least squares sense, of the control points of the deformed NURBS model at end-diastole using quadratic programming. This allows for subsequent forward displacement fitting from end-diastole to all later time frames. After fitting to all time points of data, lofting the NURBS model at each time point creates a comprehensive 4-D NURBS model. From this model, we can extract 3-D myocardial deformation fields and corresponding strain maps which are local measures of non-rigid deformation. The results show that, in the case of simulated data, the quadratic Cartesian-based NURBS model outperformed its counterparts in predicting normal strain. This model was used to then calculate normal Lagrangian and Eulerian strains in canine data.

Three-Dimensional Coronary Angiography (3D-CA) is a novel tool that allows clinicians to view and analyze coronary arteries in three-dimensional format. This will help to find accurate length estimates and to find the optimal viewing angles of a lesion based on the three-dimensional vessel orientation. Various advanced algorithms are incorporated in this 3D processing utility including 3D-RA calibration, ECG phase selection, 2D vessel extraction, and 3D vessel modeling into a utility with optimized workflow and ease-of-use features, which is fully integrated in the environment of the x-ray catheterization lab. After the 3D processing, the 3D vessels can be viewed and manipulated interactively inside the operating room. The TrueView map provides a quick overview of gantry angles with optimal visualization of a single or bifurcation lesion. Vessel length measurements can be performed without risk of underestimating a vessel segment due to foreshortening. Vessel cross sectional diameters can also be measured. Unlike traditional, projection-based quantitative coronary analysis, the additional process of catheter calibration is not needed for diameter measurements. Validation studies show a high reproducibility of the measurements, with little user dependency.

With the recent, rapid development of multidetector computed tomography (MDCT), excitement has built around the possibility of noninvasively imaging the coronary arteries. While the development of hardware and reconstruction technologies have advanced significantly, current image analysis techniques are dominated by manual interpretation using maximum intensity projections and volume rendering. If MDCT is to become the tool that it aims to be, objective, quantitative methods of image analysis will be necessary - not only to facilitate the study of atherosclerosis and coronary heart disease, but also for the accurate and timely interpretation of clinical data. This study focuses on the interobserver variability associated with the analysis of coronary MDCT images and a method for automatic segmentation of the same images. In the study of interobserver variability, six independent experts manually traced the luminal border in 60 randomly selected vascular cross sections (5 cross section each from: 4 LAD, 4 LCX, and 4 RCA). The images were acquired with an Mx8000 IDT 16-slice MDCT scanner. The mean unsigned difference for all observers was 0.38 ± 0.26 mm, with an average maximum difference of 1.32 mm. Using the expertly identified luminal borders, an independent standard was created by averaging the six sets of contours. This standard was then used to validate a prototypical automated segmentation system that uses dynamic programming and a knowledge-based cost function to optimally segment the luminal border. The resulting border positioning error was 0.17 ± 0.12 mm.

The mechanisms of plaque development in coronary arteries are not yet completely understood. Vessel geometry influences the local hemodynamics within a vessel, and the resulting wall shear stress in turn influences plaque development. Previously, we showed in-vivo that plaque tends to accumulate more on the inner curvature of a vessel than on its outer curvature. While vessel curvature is preserved during plaque progression, the local wall shear stresses change with lumen narrowing. The aim of this study was to test how the hypothesis that locations of low wall shear stress coincide with circumferentially larger plaque accumulation depends on vascular remodeling with or without lumen narrowing. We have analyzed 39 in-vivo intravascular-ultrasound pullbacks, for which geometrically accurate 3-D models were obtained by fusion with x-ray angiography. Distorting subsegments (branches, calcifications, stents) were discarded, and the relative number of vessel locations was determined within a 10-40% area-stenosis range. This range corresponds to compensatory enlargement (outward or positive vessel remodeling), but not yet lumen narrowing, and these vessel segments were a focus of our study. For each segment, we determined the relative number of vessel locations for which circumferentially low wall shear stress coincided with larger plaque thickness and vice versa. The inverse association between wall shear stress and plaque thickness was significantly more pronounced (p<0.005) in vessel cross sections exhibiting compensatory enlargement without luminal narrowing than when the full spectrum of vessel stenosis severity was considered. Thus, the hypothesis is supported more in subsegments with less developed disease.

Previous patient-specific computational fluid dynamics (CFD) models of cerebral aneurysms constructed from 3D rotational angiography have been limited to aneurysms with a single route of blood flow. However, there are numerous aneurysms that accept blood flow from more than one avenue of flow such as aneurysms in the anterior communicating artery. Although the anatomy of these aneurysms could be visualized with other modalities such as CTA and MRA, cerebral rotational angiography has the highest resolution, and is therefore the preferred modality for vascular CFD modeling. The purpose of this paper is to present a novel methodology to construct personalized CFD models of cerebral aneurysms with multiple feeding vessels from multiple rotational angiography images. The methodology is illustrated with two examples: a model of an anterior communicating artery aneurysm constructed from bilateral rotational angiography images, and a model of the complete circle of Willis of a patient with five cerebral aneurysms. In addition, a sensitivity analysis of the intraaneurysmal flow patterns with respect to mean flow balance in the feeding vessels was performed. It was found that the flow patterns strongly depend on the geometry of the aneurysms and the connected vessels, but less on the changes in the flow balance. These types of models are important for studying the hemodynamics of cerebral aneurysms and further our understanding of the disease progression and rupture, as well as for simulating the effect of surgical and endovascular interventions.

Although the natural history of cerebral aneurysms remains unknown, hemodynamics is thought to play an important role in their initiation, growth and rupture. This paper describes a pilot clinical study of the association between intraaneurysmal hemodynamic characteristics and the rupture of cerebral aneurysms. A total of 62 patient-specific models of cerebral aneurysms were constructed from 3D angiography images. Computational fluid dynamics simulations were performed under pulsatile flow conditions. The aneurysms were classified into different categories depending on the complexity and stability of the flow pattern, the location and size of the flow impingement region, and the size of the inflow jet. These features were analyzed for associations with history of rupture. A large variety of flow patterns was observed. It was found that 72% of ruptured aneurysms had complex or unstable flow patterns, 80% had small impingement regions and 76% had small jet sizes. Conversely, unruptured aneurysms accounted for 73%, 82% and 75% of aneurysms with simple stable flow patterns, large impingement regions and large jet sizes, respectively.

A computational model of blood flow through the human pulmonary arterial tree has been developed to investigate the relative influence of branching structure and gravity on blood flow distribution in the human lung. A geometric model of the largest arterial vessels and definitions of the lobar boundaries were first derived using multi-detector row x-ray computed tomography (MDCT) scans from the Lung Atlas. Further accompanying arterial vessels were generated from the MDCT vessel end points into the lobar volumes using a volume filling branching algorithm. A reduced form of the Navier-Stokes equations were solved within the geometric model to simulate pressure, velocity and vessel radius throughout the network. Blood flow results in the anatomically-based model, with and without gravity, and in a symmetric arterial model were compared in order to investigate their relative contributions to blood flow heterogeneity. Results showed a persistent blood flow gradient and flow heterogeneity in the absence of gravitational forces in the anatomically-based model. Results revealed that the asymmetric branching structure of the model was largely responsible for producing this heterogeneity. Analysis of average results in different slice thicknesses illustrated a clear flow gradient due to gravity in 'lower-resolution’ data (thicker slices), but on examination of higher resolution data a trend was less obvious. Results suggest that while gravity does influence flow distribution, the influence of the tree branching structure is also a dominant factor. These results are consistent with high-resolution experimental studies that have demonstrated gravity to be only a minor determinant of blood flow distribution.

Multi-spectral imaging of the ocular fundus suffers from three main problems: the image must be taken through an aperture (the pupil), meaning that the absolute light intensity at the fundus cannot be known; long acquisition times are not feasible due to patient discomfort; patient movement can lead to loss of image quality. These difficulties have meant that multi-spectral imaging of the fundus has not yet seen wide application. We have developed a new method for optimizing the multi-spectral imaging process which also allows us to derive semi-quantitative information about the structure and properties of the fundus. We acquire images in six visible spectral bands and use these to deduce the concentration and distribution of the known absorbing compounds in the fundus: blood haemoglobins in the retina and choroid, choroidal melanin, RPE melanin and xanthophyll. The optimisation process and parameter recovery uses a Monte Carlo model of the spectral reflectance of the fundus, parameterised by the concentrations of the absorbing compounds. The model is used to compute the accuracy with which the values of the model parameters can be deduced from an image. Filters are selected to minimise the error in the parameter recovery process. Theoretical investigations suggest that parameters can be recovered with RMS errors of less than 10%. When applied to images of normal subjects, the technique was able to successfully deduce the distribution of xanthophyll in the fundus. Further improvement of the model is required to allow the deduction of other model parameters from images.

Trabecular bone (TB) consists of a network of interconnected struts and plates occurring near the joints of long bones and in the axial skeleton. In response to mechanical stresses it remodels such that trabeculae are aligned with the major stress lines, thus leading to a highly anisotropic network. Beside bone volume fraction, anisotropy and topological indices are known to be strong predictor of the TB mechanical competence. In osteoporosis, the most common bone disorder, the remodeling balance is perturbed due to increased resorption, resulting in net bone loss accompanied by architectural deterioration, leading to fragile bone and increased fracture risk. In vertebral osteoporosis, preferential loss of transverse trabeculae leads to increased anisotropy and change in topology, hence exact measurements of these parameters are of paramount interest. Current in vivo imaging yields voxel size comparable to TB thickness, thus resulting in inherently fuzzy representations. The commonly used methods for anisotropy require binarization which is difficult to achieve in the limited spatial resolution regime where the intensity histogram is mono-modal. Here, we present a new tensor scale (t-scale) based TB architectural measures that (1) obviates binarization, and (2) yields localized measures. We evaluate the performance of this method on micro-CT images of vertebral bone and test the hypothesis that the method, along with BMD and other structural parameters, allows prediction of TB’s mechanical competence. Toward this goal, we estimate Young’s modulus (YM) of (13mm)³ vertebral TB samples under uniaxial loading and examine linear correlation of different t-scale parameters computed via micro-CT imaging .

The spatial autocorrelation analysis method represents a powerful, new approach to quantitative characterization of structurally quasi-periodic anisotropic materials such as trabecular bone (TB). The method is applicable to grayscale images and thus does not require any preprocessing, such as segmentation which is difficult to achieve in the limited resolution regime of in vivo imaging. The 3D autocorrelation function (ACF) can be efficiently calculated using the Fourier transform. The resulting trabecular thickness and spacing measurements are robust to the presence of noise and produce values within the expected range as determined by other methods from μCT and μMRI datasets. TB features found from the ACF are shown to correlate well with those determined by the Fuzzy Distance transform (FDT) in the transverse plane, i.e. the plane orthogonal to bone’s major axis. The method is further shown to be applicable to in-vivo μMRI data. Using the ACF, we examine data acquired in a previous study aimed at evaluating the structural implications of male hypogonadism characterized by testosterone deficiency and reduced bone mass. Specifically, we consider the hypothesis that eugonadal and hypogonadal men differ in the anisotropy of their trabecular networks. The analysis indicates a significant difference in trabecular bone thickness and longitudinal spacing between the control group and the testosterone deficient group. We conclude that spatial autocorrelation analysis is able to characterize the 3D structure and anisotropy of trabecular bone and provides new insight into the structural changes associated with osteoporotic trabecular bone loss.

The understanding of the mastication system motion is essential to maxillofacial surgeons and dentists in the procedures concerning jaw and teeth corrections. The temporomandibular joint (TMJ), despite its complexity, is one of the most frequently used joints of the human body. The incidence of a great number of injuries in this joint is influenced not only by its regular use during the mastication, but also by the strong forces applied by the muscles and the wide range of movements it is capable to perform. In this work, we propose the development of a jaw simulator capable of reproducing the complete mastication movement. Our jaw simulator is basically composed by three triangle meshes representing the 3D model of the cranium, mandible and teeth; and an anatomically-based joint model conceived to represent the TMJ motion. The polygonal meshes describing the bones and teeth are obtained from CT images and the jaw motion is simulated using the joint model guided by a 3D motion curve obtained from the composition of the standard 2D curves available in the medical literature. The scale, height and width of these original curves are modified to simulate different kind and size of food and to represent the movements’ variability depending on patient morphology (teeth, bones, joints and muscles). The evaluation of preliminary results involved the comparison of a dynamic MRI of a healthy person with the respective simulation.

Microscopic magnetic resonance elastography (μMRE) is a phase contrast based imaging technique that is capable of mapping the acoustic shear waves resulting from low amplitude cyclic displacement in tissue-like materials. This new technique has proven successful in imaging gel phantoms mimicking soft biological tissues with shear moduli ranging from 0.7 to 40 kPa. The 4-dimensional (4D) spatial-temporal shear wave vector can be measured, which in turn can be used to identify material properties with high spatial resolution. Experiments were conducted using 5 and 10 mm RF saddle coils in the 10 mm vertical imaging bore of an 11.74 Tesla magnet. The field-of-view ranged from 4 to 14 mm, with in plane resolution up to 34 μm x 34 μm and slice thickness up to 100 μm using shear wave excitation of 550 to 580 Hz. In this study, the capability and constraints of μMRE are investigated. The constraints include the range of measured shear moduli, excitation frequency, and minimum physical sample volume. Applications investigated include: 1) late-stage frog oocytes with typical diameter from 1 to 1.5 mm; and 2) tissue engineered constructs at different growth stages. Mesenchymal stem cells (MSCs) extracted from bone marrow can serve as progenitor cells that differentiate into specific types of tissues such as bone, adipose tissue, cartilage and muscle. μMRE can monitor the growth of such tissues and evaluate their mechanical properties. Also, a silicon-based tissue phantom material (CF-11-2188, Nusil Technologies) is tested in order to address challenges associated with excitation frequency and the dispersive nature of the media.

Recent efforts in medical imaging have shown that mechanical stimulation of tissues and a suitable imaging modality can be used to interrogate elastic properties of human tissues. Malignant tissues can have elastic properties that allow the physician to separate them from benign counterparts or plaque in arteries can be characterized in regard to its age by measuring its elastic properties. Our system consists of: (1) an acoustic source to induce tissue displacement, (2) a tissue mimicking phantom, and (3) MRI as a method for imaging and measuring the induced shear wave in the phantom. Agar was used to construct a tissue mimicking phantom. A modified spin echo sequence was written to trigger the acoustic system and phase encode the displacement information with magnetic field gradients. A series of images was obtained from the modified multi-slice, spin-echo sequence. Images showed z-axis displacement created by the radiation force. Additional experiments recorded the x and y displacement and allowed for a full 3D vector reconstruction of shear wave propagation. MRI provides a method to record displacements created by radiation force. Acoustical sources can be used to induce shear waves, which in turn can be imaged with MRI methods to quantify and display this wave in a 3D fashion.

Virtual endoscopy needs some precomputation of the data (segmentation, path finding) before the diagnostic process can take place. We propose a method that precomputes multinode cubic panorama movies using Quick-Time-VR. This technique allows almost the same navigation and visualization capabilities as a real endoscopic procedure, a significant reduction of interaction input is achieved and the movie represents a document of the procedure.

This paper presents a method for virtually generating unfolded views of the stomach using volumetric image deformation. When we observe an organ with a large cavity in it, such as the stomach or the colon, by using a virtual endoscopy system, many changes of viewpoint and view direction are required. If virtually unfolded views of a target organ could be generated, doctors could easily diagnose the organ's inner walls only by one or a several views. In the proposed method, we extract a stomach wall region from a 3-D abdominal CT images and the obtained region is shrunken. For every voxel of the shrunken image, we allocate a hexahedron. In the deformation process, nodes and springs are allocated on the vertices, edges, and diagonals of each hexahedron. Neighboring hexahedrons share nodes and springs, except for the hexahedrons on the cutting line that a user specifies. The hexahedrons are deformed by adding forces that direct the nodes to the stretching plane to the nodes existing on the cutting line. The hexahedrons are deformed using iterative deformation calculation. By using the geometrical relations between hexahedrons before and after deformation, a volumetric image in which the stomach region is unfolded. Finally, the unfolded views are obtained by visualizing the reconstructed volume can be constructed. We applied the proposed method to eleven cases of 3-D abdominal CT images. The results show that the proposed method can accurately reproduce folds and lesions on the stomach.

The recent introduction of ultrathin bronchoscopy offers considerable promise for diagnosing even small peripheral lung nodules previously considered inaccessible for routine flexible bronchoscopy. However this requires obtaining an accurate roadmap prior to endoscopy. Although virtual bronchoscopy (VB) has proved to be a useful tool for planning transbronchial interventions involving the central airways, to date, VB has received little attention for providing roadmaps to peripheral lesions. This may be especially problematic, as ultrathin bronchoscopes can now access airways not visualized on routine high-resolution CT scanners. We propose to extend the reach of virtual bronchoscopy by using peripheral arteries as surrogates for peripheral bronchi that cannot be identified even with high-resolution CT technique. Since every bronchus is accompanied by an artery, it should hypothetically be possible to substitute one for another and derive useful navigational roadmaps. This paper presents a preliminary investigation of this concept, using a combination of virtual endoscopic techniques. Virtual angioscopic and bronchoscopic flythroughs are created and transition points are selected at points that can be easily identified on CT images as corresponding structures. The proximal bronchial path and the distal arterial path are then combined and presented as a single continuous flythrough. Our preliminary investigations show that as expected, the local geometry of the airway and corresponding artery are similar. In addition to visual inspection, we use the segmentation of the arterial and bronchial trees and their tree models. Selected paths from each tree model are compared by various similarity measures in order to demonstrate their correspondence. We anticipate that this technique for bronchoscopy planning will enable bronchoscopic evaluation of previously unreachable peripheral lung nodules.

A system for planning transbronchial needle aspiration (TBNA) based on high-resolution chest CT is presented, comprising 2D axial, coronal or sagittal views and a 3D perspective intra-luminal view of the airways. The biopsy site can be defined interactively on the 2D views, and is displayed as 3D object across the translucent bronchial wall. Reference points can be placed on anatomical landmarks like the carina, which allows measuring 3D distances to viewpoints or to other landmarks. Orientation of the targets can be estimated based on a consistent orientation of the virtual endoscopic view. The system can be used as a pre-interventional planning tool, or simultaneously during the biopsy, in order to select the optimal needle insertion points. The system does not provide registration between the virtual and the real images, and does not require special hardware for tracking or any modifications of the bronchoscope. A phantom study comprising three bronchoscopists with different levels of experience showed a significant increase in yield compared to the traditional procedure based on axial CT images alone.

We propose a new quantitative method for detailed analysis of the major airways. Using a 3D MDCT chest image as input, the method involves three major steps: (1) segmentation of the airway tree, (2) extraction of the central-axis structure of the major airways, and (3) a novel improvement on the standard full-width half-maximum approach for airway-wall delineation. The method produces measurements for all defined tree branches. These measurements include various airway diameters and cross-sectional area values. To facilitate the examination of these measurements, we also demonstrate an integrated virtual-bronchoscopic analysis system that enables flexible interrogation of the airways. Of particular note are techniques for unraveling and viewing the topography of selected airways. A large series of phantom and human tests confirm the efficacy of our methods.

This paper presents a method to optimize the SVM committee used in a colonic polyp CAD system to achieve high detection performance and efficiency. In our CAD system, characteristic features of polyp candidates are fed into a committee of SVMs to determine if one detection is a true polyp. The committee consists of M different SVMs, and each of them is established by an N-feature vector. A progressive feature vector selection scheme was proposed to select a population of feature vectors, in which N-feature vectors are composed progressively in N stages. To optimize the SVM committee configuration, two-way ANOVA is performed to analyze the effect of committee-member-number (M) and feature-vector-length (N). The area under the ROC curve (AUC) in a ten-fold cross validation is used as the performance metric. Pairwise Tukey’s tests are performed to reveal if the performance differences between two configurations are statistically significant. The experiments were tested on 29 patients with 53 polyps. The committee configuration in comparison are N=1 to 7 and M=1, 3, 5, 7, or 9. ANOVA showed that N = 3 has statistically significant performance improvement over N=1 and 2, but is statistically equivalent with N= 4 to 7. It also showed that there is statistical improvement from M = 1 to 7, while M = 7 and 9 are statistically equivalent. Based on the result, we chose a committee configuration with N = 3 and M = 7 since it is the most efficient committee with statistically best performance.

Colonic polyps are growths on the inner wall of the colon. They appear like elliptical protrusions which can be detected by curvature-derived shape discriminators. For reasons of computation efficiency, much of the past work in computer-aided diagnostic CT colonography adopted kernel-based convolution methods in curvature estimation. However, kernel methods can yield erroneous results at thin structures where the gradient diminishes. In this paper, we investigate three surface patch fitting methods: Cubic B-spline, paraboloid, and quadratic polynomials. This "patch" approach is based on the fact that a surface can be re-oriented such that it can be approximated by a bivariate function locally. These patch methods are evaluated by synthesized data with various orientations and sampling sizes. We find that the cubic spline method performs best regardless of large orientation variances. Cubic spline and quadratic polynomial methods perform equally well for large samples while the latter performs better for small ones. Based on the performance evaluation, we propose a new, two-stage curvature estimation method. The cubic spline fitting is performed first for its insensitivity to orientation. If the spline fitting errs by more than a preset value (indicating high surface tortuosity), a small data sample is fitted by a quadratic function. The evaluation is performed on 29 patients (58 data sets). With 88.7% sensitivity, the average number of false positives per data set is reduced by 44.5% from 33.5 (kernel method) to 18.6 (new method).

This paper proposes a method for detecting colonic-polyp candidates from 3-D abdominal CT images based on curvatures calculated from curve fitting results. The proposed method can detect polyp candidates with a very low false-positive rate. Development of a computer aided detection (CAD) system for colonic polyps is expected to be continue due to significant increase of colonic cancers in Japan. Many research groups have reported methods for detecting colonic polyp candidates based on curvatures on colonic walls. However, because they approximated the first and second-order derivatives, which are required for computing curvatures, by using intensity differentiation, detection results were significantly influenced by noise and included many false-positives. To reduce false-positives, we propose a method for geometrically calculating curvatures based on curve-fitting to iso-intensity points. First the colonic wall region is segmented by using a region-growing method from original images. For each point on the colonic wall (target point), we find iso-intensity points around the given processing area by using linear-interpolation. Curve fitting to the obtained points is performed by using a least squared error method, and the size of the processing area is automatically adjusted during the curve fitting process. The curvature of the target point is calculated from the first and second derivatives of the obtained curve, after which the shape index and curvedness are calculated. We extract points whose shapes are classified as convex and within the predefined curvedness. Colonic polyp candidates are obtained by performing connected component analysis including small component elimination. The proposed method was applied to a noisy artificial-figure image and abdominal CT images. Experimental results indicate that the proposed method detected a few false-positives while maintaining 100% true positive rate, whereas the previous method generated 10 FPs under the same experimental conditions.

Objective: To investigate the feasibility of laxative-free bowel preparation to relieve the patient stress in colon cleansing for virtual colonoscopy. Materials and Methods: Three different bowel-preparation protocols were investigated by 60 study cases from 35 healthy male volunteers. All the protocols utilize low-residue diet for two days and differ in diet for the third day - the day just prior to image acquisition in the fourth day morning. Protocol Diet-1 utilizes fluid or liquid diet in the third day, Diet-2 utilizes a food kit, and Diet-3 remains the low-residue diet. Oral contrast of barium sulfate (2.1%, 250 ml) was added respectively to the dinner in the second day and the three meals in the third day. Two doses of MD-Gastroview (60 ml) were ingested each in the evening of the third day and in the morning before image acquisition. Images were acquired by a single-slice detector spiral CT (computed tomography) scanner with 5 mm collimation, 1 mm reconstruction, 1.5-2.0:1.0 pitch, 100-150 mA, and 120 kVp after the colons were inflated by CO₂. The contrasted colonic residue materials were electronically removed from the CT images by specialized computer-segmentation algorithms. Results: By assumptions that the healthy young volunteers have no polyp and the image resolution is approximately 4 mm, a successful electronic cleansing is defined as “no more than five false positives and no removal of a colon fold part greater than 4 mm” for each study case. The successful rate is 100% for protocol Diet-1, 77% for Diet-2 and 57% for Diet-3. Conclusion: A laxative-free bowel preparation is feasible for virtual colonoscopy.

CT colonography (CTC) is being extensively studied for its potential value in colon examinations, since it offers many advantages such as lower risk and less patient discomfort. However, CTC, like all other types of full structural colorectal examinations to date, requires complete bowel preparation. The inconvenience and discomfort associated with this preparation is an important obstacle to compliance with currently recommended colorectal screening guidelines. To maximize compliance, CTC would ideally be performed on an unprepared colon. However, in an unprepared colon residual stool and fluid can mimic soft tissue density and thus confound the identification of polyps. An alternative is to tag the stool with an opacifying agent so that it is brighter than soft tissue and thus easily recognized automatically and then reset to air values. However, such electronic stool subtraction in a totally unprepared colon is difficult to perform accurately for several reasons, including poorly labeled areas of stool, the need to accurately quantify partial volume effects, and noise. In this study the performance of a novel stool subtraction algorithm was assessed in unprepared CT colonography exams of 25 consecutive volunteers who had undergone colonoscopy with positive findings. Results showed 81% sensitivity to clinically relevant lesions > 1 cm with 0.52 false positives per patient compared to colonoscopy findings. Although further study and refinement of the stool subtraction process is required, CT colonography of the unprepared colon with electronic stool subtraction is feasible at detection levels comparable to the prepared colon.

We investigated the application of optimized ranking schemes in the analysis of polyp candidates detected from CT colonography datasets by our CAD scheme. CT colonography was performed for 28 patients in supine and prone positions with a standard pre-colonoscopy preparation and air distension. There were 42 colonoscopy-confirmed polyps 5-25 mm in size. The colons were extracted by use of a centerline-based colon segmentation technique. Polyp candidates were detected from the extracted region of colonic wall by use of geometric features sensitive to polypoid shapes. The by-polyp detection sensitivity was 98%. The detected polyp candidates were ranked based on a ranking function constructed from a linear combination of polyp features. The feature weights were optimized by a downhill simplex method by maximizing the rank of the lowest-ranking true-positive polyp candidate. We considered two types of ranking: by-dataset ranking and global ranking. The most effective ranking functions included multiple features and could reduce the amount of initially detected polyp candidates by 56% without compromising the high by-polyp detection sensitivity. The ranking schemes may be useful in optimizing the performance of CAD schemes for the detection of polyps in CT colonography.

Studying complex thorax breating motion is an important research topic for accurate fusion of functional and anatomical data, radiotherapy planning or reduction of breathing motion artifacts. We investigate segmented CT lung, airway and diaphragm surfaces at several different breathing states between Functional Residual and Total Lung Capacity. In general, it is hard to robustly derive corresponding shape features like curvature maxima from lung and diaphragm surfaces since diaphragm and rib cage muscles tend to deform the elastic lung tissue such that e.g. ridges might disappear. A novel registration method based on the shape context approach for shape matching is presented where we extend shape context to 3D surfaces. The shape context approach was reported as a promising method for matching 2D shapes without relying on extracted shape features. We use the point correspondences for a non-rigid thin-plate-spline registration to get deformation fields that describe the movement of lung and diaphragm. Our validation consists of experiments on phantom and real sheep thorax data sets. Phantom experiments make use of shapes that are manipulated with known transformations that simulate breathing behaviour. Real thorax data experiments use a data set showing lungs and diaphragm at 5 distinct breathing states, where we compare subsets of the data sets and qualitatively and quantitatively asses the registration performance by using manually identified corresponding landmarks.

A method is described to estimate regional lung expansion and related biomechanical parameters using multiple CT images of the lungs, acquired at different inflation levels. In this study, the lungs of two sheep were imaged utilizing a multi-detector row CT at different lung inflations in the prone and supine positions. Using the lung surfaces and the airway branch points for guidance, a 3D inverse consistent image registration procedure was used to match different lung volumes at each orientation. The registration was validated using a set of implanted metal markers. After registration, the Jacobian of the deformation field was computed to express regional expansion or contraction. The regional lung expansion at different pressures and different orientations are compared.

Airway mucosal color changes occur in response to the development of bronchial diseases including lung cancer, cystic fibrosis, chronic bronchitis, emphysema and asthma. These associated changes are often visualized using standard macro-optical bronchoscopy techniques. A limitation to this form of assessment is that the subtle changes that indicate early stages in disease development may often be missed as a result of this highly subjective assessment, especially in inexperienced bronchoscopists. Tri-chromatic CCD chip bronchoscopes allow for digital color analysis of the pulmonary airway mucosa. This form of analysis may facilitate a greater understanding of airway disease response. A 2-step image classification approach is employed: the first step is to distinguish between healthy and diseased bronchoscope images and the second is to classify the detected abnormal images into 1 of 4 possible disease categories. A database of airway mucosal color constructed from healthy human volunteers is used as a standard against which statistical comparisons are made from mucosa with known apparent airway abnormalities. This approach demonstrates great promise as an effective detection and diagnosis tool to highlight potentially abnormal airway mucosa identifying a region possibly suited to further analysis via airway forceps biopsy, or newly developed micro-optical biopsy strategies. Following the identification of abnormal airway images a neural network is used to distinguish between the different disease classes. We have shown that classification of potentially diseased airway mucosa is possible through comparative color analysis of digital bronchoscope images. The combination of the two strategies appears to increase the classification accuracy in addition to greatly decreasing the computational time.

Lung parenchyma evaluation via multidetector-row CT (MDCT), has significantly altered clinical practice in the early detection of lung disease. Our goal is to enhance our texture-based tissue classification ability to differentiate early pathologic processes by extending our 2-D Adaptive Multiple Feature Method (AMFM) to 3-D AMFM. We performed MDCT on 34 human volunteers in five categories: emphysema in severe Chronic Obstructive Pulmonary Disease (COPD) as EC, emphysema in mild COPD (MC), normal appearing lung in COPD (NC), non-smokers with normal lung function (NN), smokers with normal function (NS). We volumetrically excluded the airway and vessel regions, calculated 24 volumetric texture features for each Volume of Interest (VOI); and used Bayesian rules for discrimination. Leave-one-out and half-half methods were used for testing. Sensitivity, specificity and accuracy were calculated. The accuracy of the leave-one-out method for the four-class classification in the form of 3-D/2-D is: EC: 84.9%/70.7%, MC: 89.8%/82.7%; NC: 87.5.0%/49.6%; NN: 100.0%/60.0%. The accuracy of the leave-one-out method for the two-class classification in the form of 3-D/2-D is: NN: 99.3%/71.6%; NS: 99.7%/74.5%. We conclude that 3-D AMFM analysis of the lung parenchyma improves discrimination compared to 2-D analysis of the same images.

For differential diagnosis of pulmonary nodules, assessment of contrast enhancement at chest CT scans after administration of contrast agent has been suggested. Likelihood of malignancy is considered very low if the contrast enhancement is lower than a certain threshold (10-20 HU). Automated average density measurement methods have been developed for that purpose. However, a certain fraction of malignant nodules does not exhibit significant enhancement when averaged over the whole nodule volume. The purpose of this paper is to test a new method for reduction of false negative results. We have investigated a method of showing not only a single averaged contrast enhancement number, but a more detailed enhancement curve for each nodule, showing the enhancement as a function of distance to boundary. A test set consisting of 11 malignant and 11 benign pulmonary lesions was used for validation, with diagnoses known from biopsy or follow-up for more than 24 months. For each nodule dynamic CT scans were available: the unenhanced native scan and scans after 60, 120, 180 and 240 seconds after onset of contrast injection (1 - 4 mm reconstructed slice thickness). The suggested method for measurement and visualization of contrast enhancement as radially resolved curves has reduced false negative results (apparently unenhancing but truly malignant nodules), and thus improved sensitivity. It proved to be a valuable tool for differential diagnosis between malignant and benign lesions using dynamic CT.