Index of content:
Volume 31, Issue 5, May 2004
- PH. D. THESES ABSTRACTS
31(2004); http://dx.doi.org/10.1118/1.1688038View Description Hide Description
The research presented in this dissertation represents an innovative application of computer-aided diagnosis and signal detection theory to the specific task of early detection of breast cancer in the context of screening mammography. A number of automated schemes have been developed in our laboratory to detect masses and clustered microcalcifications in digitized mammograms, on the one hand, and to classify known lesions as malignant or benign, on the other. The development of fully automated classification schemes is difficult, because the output of a detection scheme will contain false-positive detections in addition to detected malignant and benign lesions, resulting in a three-class classification task. Researchers have so far been unable to extend successful tools for analyzing two-class classification tasks, such as receiver operating characteristic (ROC) analysis, to three-class classification tasks. The goals of our research were to use Bayesian artificial neural networks to estimate ideal observer decision variables to both detect and classify clustered microcalcifications and mass lesions in mammograms, and to derive substantial theoretical results indicating potential avenues of approach toward the three-class classification task. Specifically, we have shown that an ideal observer in an N-class classification task achieves an optimal ROC hypersurface, just as the two-class ideal observer achieves an optimal ROC curve; and that an obvious generalization of a well-known two-class performance metric, the area under the ROC curve, is not useful as a performance metric in classification tasks with more than two classes. This work is significant for three reasons. First, it involves the explicit estimation of feature-based (as opposed to image-based) ideal observer decision variables in the tasks of detecting and classifying mammographic lesions. Second, it directly addresses the three-class classification task of distinguishing malignant lesions, benign lesions, and false-positive computer detections. Finally, it develops important theoretical results for N-class classification tasks that should prove of value in the development of a three-class extension to ROC analysis methods.
31(2004); http://dx.doi.org/10.1118/1.1701953View Description Hide Description
The passage from conventional radiographic film-based to image-based high dose rate (HDR) brachytherapy for prostate cancer has significantly improved our ability to define the targets and organs at risk (OAR). Three-dimensional (3D) anatomical information are obtained from Computed Tomography(CT) or Magnetic Resonance Imaging(MRI), and cancer areas are validated using functional imaging with magnetic resonance spectroscopy imaging(MRSI). Together with inverse planning optimization, image-based HDR brachytherapy can deliver a highly conformal dose distribution to the target while sparing OAR. However, uncertainties inherent to the use of the new imaging modalities may impact on the dose distribution. In this work, we have selected the major uncertainties and studied their impact within the context of the clinical procedures of ultrasound guided HDR prostate brachytherapy delivered in two fractions. The feasibility of using functional imaging to guide the delivery of higher dose to dominant intraprostatic lesions (DIL) within the prostate was also investigated. The average cranio-caudal displacement of catheters between fractions was 2.7 mm and 5.4 mm for bony anatomy and gold seed marker measurement, respectively. Either increase or reduction of prostate volume was observed with an average of 7.8% and a maximum of 17% between fractions, resulting in minimal dose changes. The dose uncertainty due to the planning CT slice thickness showed a relative error of 1% on average for current 3 mm planning CT, independent of the transversal region of the prostate. A retrospective study taking advantage of MRI/MRSI for the dose escalation of the DIL demonstrated that high dose areas could be redistributed to boost the DIL by 120% without any additional dose delivered to OAR compared to a reference plan. On average, the rigid endorectal probe was shown to rotate the prostate anteriorly by to expand it in the antero–posterior direction by 1.2 mm, and compress it in lateral direction with 1.5 mm. Finally, MRI scans with and without the probe could be registered with a 2 mm precision using translation only. The use of 3D anatomical and functional information combined with inverse planned HDR brachytherapy is a precise procedure for the treatment of localized prostate cancer with potential for dose escalation.