Automatic Detection of Benign Breast Tumors

Abstract

Breast cancer is the second most common cancer. Mostly it founds in females as compared to males. The objectives of this study are to show and prove the difference of the interpretation of mammographic images by radiologists and attempt to develop CAD techniques that aim to provide a second option diagnosis for radiologist. Samples of twenty digital mammograms obtained from the mini-MIAS (Mammographic Image Analysis Society) database were distributed to three different radiologists in order to verify the variability amongst them, the result of their diagnosis showed a marked variation in size and number of tumors. MIAS images have been taken to detect benign tumors using digital image processing techniques (haralick equation), the features: energy, contrast and homogeneity were extracted using GLCM (Gray Level Co-Occurrence Matrix). For each image a normal an abnormal frame was taken and its features were determined. Its scatter shows it’s not possible to locate exactly place of the tumor.