Hybrid CNN and LSTM Model for Early Detection of Breast Cancer Using Histopathology Images

Breast cancer is one of the leading causes of cancer related deaths in women around the globe, and its timely and accurate detection has become a topmost priority for optimizing patient survival. Recent advancements in deep learning techniques using medical imaging have optimized the accuracy and early detection of breast cancer. This study proposed a lightweight deep learning-based hybrid model, CNN-LSTM, specifically tailored for binary classification of breast cancer histopathology images while maintaining low architectural complexity. In this model, there are four convolutional blocks for feature extraction, and an LSTM of 128 units for sequential learning is deployed. The model was trained and evaluated by a public histopathology dataset, BreakHis, across four magnification levels. The experimental results showed the model achieved a balanced performance across accuracy, precision, recall, and F−1 score with 0.8883,0.9176, 0.9209, and 0.9192, respectively, at a magnification level of 200X. The results highlight the effectiveness of combining CNN and LSTM for feature extraction and sequential learning in breast histopathology images. Thus, the proposed model can significantly enhance the diagnostic accuracy and efficiency of early breast cancer detection, resulting in improved patient outcomes. However, the model generalization to diverse datasets and its sensitivity to class imbalance is still needed in future studies.