An Efficient Image Based Mammogram Classification Framework Using Depth Wise Convolutional Neural Network

Abstract

Architectural Distortion is the third most concerning sign of abnormal areas on a mammogram. It's challenging to diagnose architectural distortion (AD) using mammograms because of the condition's delicacy, fluctuating imbalance on mammary mass, and small size. The use of computer algorithms for the early identification of aberrant ADs areas in mammography might aid radiologists and clinicians. Classification performance is negatively impacted due to star-shaped structural defects in ROI recognition, noise reduction, and object localization. This method uses computer vision to automatically filter out background noise and pinpoint the precise placement of items inside complex patterns. This study used computer vision techniques to investigate the potential for identifying mammography with geometric deformation inside ROIs. The researcher proposed a computer-aided diagnostic approach that utilizes machine training to analyze architectural deformation in digital mammography for the purpose of identifying breast cancer. Image preprocessing, enhancement, and pixel-by-pixel segmentation are only some of the four components of the proposed mammography classification system. Architecture-based distorted region-of-interest (ROI) identification, deep learning and machine learning network training for malignant/benign ROI classification in Alzheimer's disease.