Shape Based Breast Medical Image Retrieval System using Arc Difference Rate
Byungkwan Jung · Open PRAIRIE (South Dakota State University) · 2014
As Computer Aided Diagnosis (CAD) is being a regular routine in the medical field, processing distinctive features from the image aid physicians’ decisions. In case of breast cancer, beside from determining whether the tumor is benign or malignant, similar cases found in different patients provide medical doctors viral information before one’s final decision. Thus, detecting similar tumor cases can provide decisive information to the doctors. In this paper, shape based image retrieval method using arc difference ratio (ADR) has been introduced and compared with different shape based image retrieval systems using different shape features such as Contour to Centroid Triangulation (CTCT) method, and Sectored Contour to Centroid Triangulation (SCTCT) method. All image retrieval systems have been tested with 1,300 binary object images that have been through feature extraction using Support Vector Machine from actual breast mammograms. In order to inspect the matching rate of shape based image retrieval system, each system results were compared with sectored object matching method. 20 test sets were experimented and proposed shape based image retrieval system using ADR shows the highest matching rate compared to others.