Robust Detection of Masses in Digitized Mammograms
Lihua Li · 2002
Abstract : This project is to develop a robust computer aided diagnosis (CAD) system for mass detection with high sensitivity and specificity in digitized mammograms. The research scope in past year is to improve and optimize detection performance and classification generalizability. Several major progresses have been made including (1). A novel graph-based algorithm was proposed to segment stellate masses in mammograms by separating the adjacent regions while keeping the spiculation of masses. It is helpful for the improvement of stellate late mass and distortion detection. (2). A hybrid hard-soft classification method was proposed, where the hard decision classifier is cascaded with a soft decision classification with the objective to reduce false-positives (FPs) in the cases with multiple FPs retained after the hard decision classification. It has a much better performance and generalizability of classification. (3!. A training database was generated for fine tuning the parameters of CAD system. An FROC curve of CAD mass detection using training database was obtained. It is expected that these processing will be very helpful in improving the robustness of the detection system.