Automated Detection of Spicules in Mass Lesions on Mammograms

概要

論文の詳細を見る
We developed a new automated detection algorithm for spicules on digital mammograms. After the background trend around the detected mass was removed by curve fitting technique, highfrequency components were enhanced by using an unsharp-mask filter. The detection was performed by using a newly developed "pendulum filter" in three steps for examining the existance of spicules, searching the thick and/or thin spicules and finding the inner components in the mass. The pendulum filter detected most of the spicules that pointed out by a radiologist. The sensitivity and specificity of our algorithm for detecting the existance of spicules in the mass were 86.7% and 86.7%, respectively, in our database of 30 mammograms.
Medical Imaging and Information Sciencesの論文