A System for Computer Aided Early Diagnosis of Breast Cancer based on Microcalcifications Analysis

We propose that a system for medical diagnosis, in order to be reliable and useful, beyond systematic evaluation must follow also some basic guidelines: (a) it must be designed as open and self-explanatory giving the opportunity to the physician to understand the method and the rules comprised in the system, (b) it must be properly designed to provide maximum effectiveness in physician’s diagnosis instead of being fully automated aiming solely in its own maximum effectiveness of diagnosis, (c) it must be easily adapted to the physician’s expertise through interactive feedback. The main aim must be the optimization in physician’s diagnosis through the use of the system instead of the optimization of system’s diagnosis alone. We present a system for computer aided early diagnosis of breast cancer that is physician oriented instead of being fully automated. Clustered microcalcifications have been considered as important indicators of the presence of breast cancer. The system is at present based on the detection and analysis of these objects in order to help the physician in the early diagnosis of breast cancer. The basic concepts of the proposed system are: (a) optimized visual examination of certain cancer indices, (b) critical feature quantification and classification, (c) adaptation to doctor’s expertise through interactive feedback, (d) real time explanation of the suggested diagnosis flowchart, (e) patient oriented monitoring of clinical data. The performance of the system has been evaluated through laboratory tests on digitized and annotated mammograms from the Nijmegen Digital Mammogram Database as well as on mammograms from the well documented archive of mammographic images of “Prolipsis” Diagnostic Breast Center.