Content based mammogram retrieval using Gray Level Aura Matrix
Severin Wiesmüller, D. Abraham Chandy · 2010
Diagnosis of breast cancer in mammograms is also for specialists a difficult and error-prone task. A good opportunity to support radiologists in their decision is to find similar mammograms out of a database to compare the current case with past cases. In this work a complete content based image retrieval (CBIR) system for mass and calcification class mammograms has been implemented under usage of MATLAB ® . The necessary feature extraction is realized on the basis of Gray Level Aura Matrix (GLAM). The normalized Gray Level Aura Matrices for the feature extraction is completely independent from shape and size of the region of interest (ROI) which gives a huge freedom for the user of the content based image retrieval system. Foundation for the built up database are 420 mammograms from the Digital Database for Screening Mammography (DDSM) database. The functionality of texture comparison using gray level aura matrix is demonstrated using the average precision of five retrieved images whereby the best result with 82.2% is reached with the nearest neighbourhood system and the quantization with eight allowed gray levels. Under these conditions the average precision of one retrieved image is 95%. Both results are better than comparable works. Index Terms—Content based image retrieval, mammogram, gray level aura matrix, image processing