Explainable AI based hybrid DRM-Net transfer learning model for breast cancer detection and classification using
Shahid Mohammad Ganie1, Majid Bashir Malik2, Mir Aadil3
1Department of Health Information Management and Technology, College of Applied Medical Sciences, King Faisal University, 31982, Al-Ahsa, Saudi Arabia. sganie@kfu.edu.sa.
Scientific Reports
|December 19, 2025
Summary
This study introduces DRM-Net, a novel hybrid deep transfer learning model for enhanced breast cancer diagnosis from ultrasound images. DRM-Net significantly improves accuracy, aiding clinical decisions.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Breast cancer is a leading cause of death in women globally.
- Early diagnosis is critical for effective treatment and improved patient outcomes.
- Deep transfer learning shows promise in medical image analysis.
Purpose of the Study:
- To develop and evaluate a novel hybrid deep transfer learning model for breast cancer diagnosis using ultrasound images.
- To compare the performance of the proposed model against existing transfer learning models.
- To enhance the accuracy and reliability of breast cancer detection.
Main Methods:
- Trained six transfer learning models on a breast ultrasound dataset.
- Developed a hybrid model (DRM-Net) by combining top-performing transfer learning models.
- Employed image preprocessing, masking, data augmentation, and hyperparameter tuning.
- Interpreted model decisions using explainable AI (XAI) class activation mapping (CAM).
Main Results:
- The proposed DRM-Net model achieved superior performance compared to six other transfer learning models.
- DRM-Net demonstrated high accuracy (96.71%), precision (96%), recall (97%), F1-score (97%), and AUC (99%).
- The model outperformed existing state-of-the-art studies in breast cancer diagnosis.
Conclusions:
- The hybrid DRM-Net model shows significant potential for improving breast cancer diagnosis accuracy.
- This AI-driven approach can support informed clinical decision-making.
- The model's applicability extends to other ultrasound-based diagnostic tasks.
