3 D IMAGE PROCESSING FOR AUTOMATED DIGITAL MAMMOGRAPHY

点击翻译按钮获取中文摘要

We propose a program of res e arch focused ont he development of numerical models and new algorithms for digital stereomammography. The goal of digital stereomammography is two-fold: (i) to develop algorithms that will provide better separation of mammograms in normal, benign, and abnormal suspicious study sets by the use of additional information (from two views), and (ii) to allow localization, in 3-D, of suspect regions via direct computation from standard stereotactic views. We specifically plan: To develop and optimize numerical models and new algorithms for digital stereomammography. To study the efficacy of multiple-view (stereoscopic) breast imaging and the degree to which it enhances the dimensionality and information level of mammographic image data. To demonstrate the suspect tissues can be validated/de-emphasized based on automatic and/or semiautomatic stereo-comparison. To show that athe digital stereomammographic approach has the potential to greatly reduce sensitivity to occlusions and the projection problem (dimension loss). To occlusions and the projection problem (dimension loss). To develop methods for automatically calculating the 3-D morphology and placement of potential suspect lesions via automated stereo triangulation. This innovative development study will utilize high resolution (42 mu) stereotactic mammogram data from a Fisher Imaging Corporation PASS (Prototype Arizona Slot Scanner) obtained by agreement from Col. Ramikant Shah, Chief Mammography section, Brooke Arm y Medical Center (BAMC), Fort Sam Houston, San Antonio. A diversity of computational stereo vision approaches will be applied to the real data sets, including sparse feature- based algorithms and dense computation-intensive algorithms. In addition, significant experiments with a synthetic breast (phantom) supplied by BAMC will be conducted in an effort to explore the optimal separation angle between the stereotactic views. The successful development of digital stereomammography as part of a semi- automated computer-aided detection (CAD) system has significant potential to increase the efficiency of the screening processes, thus decreasing the average cost per study, reducing the manpower required for routine screening mammography, and possibly increasing the sensitivity and specificity of the screening process, allowing a greater percentage of time to be devoted t the analysis of abnormal images.