Deep Parallel CNN and LSTM for Enhanced Breast Cancer Classification

Breast cancer is a leading cause of mortality among women, necessitating early detection for effective treatment. Accurate classification remains challenging due to the limited ability of models to capture both micro and macro features in mammogram data. To address this, the study proposes a Parallel CNN-LSTM (PCNN-LSTM) model, designed to enhance classification accuracy by integrating micro and macro feature analysis. The proposed PCNN architecture extracts features by analyzing both micro and macro features using two separate window sizes. In the preprocessing stage, the images are resized to 60x60, followed by image sharpening and contrast enhancement to improve feature clarity. Batch normalization is applied during feature extraction to reduce overfitting. The classification performance is evaluated using three classifiers: K-Nearest Neighbors (KNN), Random Forest (RF), and LSTM. The PCNN-LSTM model achieves a highest accuracy of 98.50% in categorizing breast cancer images into three types—benign, malignant, and normal—demonstrating its potential for improving diagnostic accuracy and efficiency.