Authors:
Abeer A. Amer, Ghada Elkhayat, Bayan Nader Alkawas
Addresses:
Department of Computer Science and Information Systems, Faculty of Management Science, Sadat Academy for Management and Sciences, Cairo, Cairo Governorate, Egypt. Department of Information Systems and Computers, Faculty of Business, Alexandria University, Alexandria, Alexandria Governorate, Egypt.
Abstract:
Breast cancer continues to be a primary cause of mortality among women worldwide, highlighting the necessity for early and precise diagnostic techniques. Deep learning techniques have demonstrated potential in medical image analysis; yet, many existing models struggle to balance high predictive accuracy and interpretability, both of which are crucial for clinical adoption. This paper presents a hybrid convolutional neural network architecture that combines the strengths of ResNet50 and VGG16 to improve classification performance on the CBIS-DDSM mammography dataset. To ensure the evaluation framework is robust, the dataset will be split into 80% for training and 20% for testing. To help doctors make better decisions, the study uses explainable artificial intelligence (XAI) approaches, including Gradient-Weighted Class Activation Mapping (Grad-CAM). This method will enable visualisation of the important areas in mammograms that affect the model's predictions, thereby making the process more transparent and trustworthy. The suggested hybrid model should do better than the separate baseline models because it uses deeper feature extraction and better representation. The integrated XAI component will also give interpretable insights that can help radiologists understand how the model works. This could help them find and diagnose breast cancer earlier. In general, this method aims to connect high-performance AI systems with interpretable models useful in the clinic.
Keywords: Breast Cancer Detection; Mammogram Classification; Hybrid CNN; Interpretable Model; Precise Diagnostic Techniques; Deep Learning; Explainable Artificial Intelligence (XAI).
Received: 09/12/2024, Revised: 28/03/2025, Accepted: 03/07/2025, Published: 09/12/2025
DOI: 10.64091/ATIHL.2025.000208
AVE Trends in Intelligent Health Letters, 2025 Vol. 2 No. 4 , Pages: 208-215