An effective CAD system using Tchebichef features and Fisher score-based distributed filter for digital mammogram classification

An effective way to control the high mortality and morbidity of breast cancer among women is to prioritize the early and correct diagnosis. Therefore, inclusion of a computer-aided diagnosis (CAD) system as a second reader assists the radiologists to detect and diagnose the disease. This paper aims at designing an automated CAD system that can proficiently categorize the mammograms as normal, benign, and malignant. The present paper applies DTT as an efficient feature extractor. Further, a distributed filter approach utilizing Fisher score filter is proposed to meticulously select the most important features followed by PNN classifier for performing the classification task. The assessment of the current model is carried out on two distinct datasets, i.e. MIAS and DDSM. The highest accuracy for MIAS and DDSM dataset comes out to be 99.586% and 99.09%, respectively.