Breast Cancer Detection using Improved MHSA-Residual Network and CLAHE Image Enhancement

Breast Cancer disease are typically severe and spread quickly. Experts in health care departments, especially in developing countries, surface challenges in detecting breast cancer since a lack in diagnosing accuracy. Henceforth, this work put forward a breast cancer detection method using an Improved Attention Residual Net (ResNet) for diagnose early-stage by analyzing Breast Ultrasound images. Also this work integrates the preprocessing step utilizes a Contrast Local Adaptive Histogram Equalization (CLAHE), segmentation step using K-means clustering identifies Regions of Interest (ROI) for further analysis of Breast cancer. The Scale-Invariant Feature Transform (SIFT) is utilized to capture important characteristics of the tumor images. The proposed model, developed using Python software and trained on a Breast Ultra Sound (BUS) dataset from kaggle repository, validates high classification of 94 % accuracy on the test dataset. This method enhances accurate detection and significantly decreases the requirement for expert intervention by providing an effective automated solution for accurate and efficient Breast cancer identification.