Application of Active Learning Technique with CNN for the Classification of Microscopic Breast Cancer Images

Detecting diseases, especially cancer, early is crucial for better healthcare outcomes. Current diagnostic methods often fall short, leading to higher mortality rates. Breast cancer is a prime example where early detection can lead to complete recovery. Digital mammography provides high-resolution images with key indicators like Architectural-Distortion, Mass, Bilateral Asymmetry, and Micro-calcification. However, the process can be tiring for radiologists, leading to errors. Histopathology analysis, the gold standard for cancer diagnosis, has limitations in research due to reliance on experts and time-consuming procedures. The medical field needs innovative solutions, and deep learning has improved histopathology analysis by enhancing accuracy, efficiency, and automation. However, challenges like dataset acquisition, model transparency, overfitting, computational demands, ethics, and standardization persist. A proposed system tackles these issues through data preprocessing, active learning, deep learning classification, and evaluation metrics. Preliminary results show an impressive accuracy of 98.78%. This comprehensive approach aims to boost breast cancer diagnosis accuracy and efficiency, making a significant contribution to patient care and medical research.