A Novel Approach for Enhanced Early Breast Cancer Detection

Abstract

Breast cancer is a leading cause of women's mortality globally, with early diagnosis crucial for survival. This study addresses diagnostic challenges including imbalanced, noisy datasets and irrelevant features using Wisconsin Diagnostic Breast Cancer (WDBC) and Wisconsin Breast Cancer Database (WBCD) datasets. The proposed approach integrates Custom Adaptive Teaching-Learning-Based Optimization (TLBO) for optimal feature selection and a novel Focal Long Short-Term Memory (Focal LSTM) network to handle imbalanced data effectively. Performance evaluation using accuracy, precision, sensitivity, specificity, F-score, and AUC metrics demonstrates significant improvements. This innovative machine learning approach successfully addresses dataset limitations, contributing robust and accessible diagnostic solutions for healthcare applications.