A novel approach for classification of abnormalities in digitized mammograms

Abstract

Feature extraction is an important process for the overall system performance in classification. The objective of this article is to reveal the effectiveness of texture feature analysis for detecting the abnormalities in digitized mammograms using Self Adaptive Resource Allocation Network (SRAN) classifier. Thus, we proposed a feature set based on Gabor filters, fractal analysis, multiscale surrounding region dependence method (MSRDM) to identify the most common appearance of breast cancer namely microcalcification, masses and architectural distortion. The results of the experiments indicate that the proposed features with SRAN classifier can improve the classification performance. The SRAN classifier produces the classification accuracy of 98.44% for the proposed features with 192 images from MIAS dataset.