XAI-Driven Mammogram Classification: A Comparative Study of SHAP and LIME in Deep Learning Models

Dr Sharada Guptha M N; Chaitra M S; Shilpa M; Divyashree B P; Dr. Babitha MN

doi:10.29284/ex54sw96

Authors

Dr Sharada Guptha M N Associate Professor, Department of Electronics and Communication Engineering, Sri Siddhartha Academy of Higher Education, Tumakuru, India
Chaitra M S Selection Grade lecturer, Electronics and Communication Department, Government Polytechnic, Malavalli, Mandya
Shilpa M Selection Grade lecturer, Electronics and Communication Department, Government Polytechnic, Malavalli, Mandya
Divyashree B P Assistant professor, Debarment of Electrical and Electronics Engineering, SJCE, JSS Science and Technology University, Mysuru, India
Dr. Babitha MN Associate Professor, Department of Electronics and Communication Engineering, Sri Siddhartha Academy of Higher Education, Tumakuru, India

DOI:

https://doi.org/10.29284/ex54sw96

Keywords:

Explainable Artificial Intelligence (XAI), Mammogram Classification, Deep Learning, SHAP and LIME, Breast Cancer Detection

Abstract

Breast cancer remains one of the leading causes of mortality among women worldwide, where early and accurate diagnosis plays a crucial role in improving survival rates. Mammography is the most widely used screening technique; however, its interpretation is often challenging due to subtle variations in tissue structures and high inter-observer variability among radiologists. In recent years, deep learning models, particularly Convolutional Neural Networks (CNNs), have demonstrated strong performance in automated mammogram classification. Despite their effectiveness, these models lack transparency, limiting their adoption in clinical decision-making due to their black-box nature. This study proposes an Explainable Artificial Intelligence (XAI)-driven framework for mammogram classification, focusing on a comparative analysis of SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations). A ResNet-based CNN model is trained on the CBIS-DDSM dataset to classify mammograms into benign and malignant categories. SHAP and LIME are integrated to provide both global and local interpretability of model predictions. Experimental results indicate that while both methods enhance model transparency, SHAP provides more stable and consistent feature attributions across samples, whereas LIME offers faster and more localized explanations. Quantitative evaluation using fidelity and stability metrics, along with qualitative visualization analysis, highlights key differences between the two approaches. The findings suggest that SHAP is more suitable for clinical validation and auditing, while LIME is effective for rapid case-based interpretation. This work contributes toward improving trust and reliability in AI-assisted breast cancer diagnosis.

References

1. Dhungel, N., Carneiro, G., & Bradley, A. P. (2017). Automated mass detection in mammograms using cascaded deep learning and random forests. Proceedings of the IEEE International Conference on Image Processing, 415–419. https://doi.org/10.1109/ICIP.2017.8296346

2. Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542(7639), 115–118. https://doi.org/10.1038/nature21056

3. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770–778. https://doi.org/10.1109/CVPR.2016.90

4. Holzinger, A., Carrington, A., & Müller, H. (2020). Measuring the quality of explanations: The system causability scale (SCS). KI - Künstliche Intelligenz, 34(2), 193–198. https://doi.org/10.1007/s13218-020-00636-z

5. Kim, T. K. (2015). T test as a parametric statistic. Korean Journal of Anesthesiology, 68(6), 540–546. https://doi.org/10.4097/kjae.2015.68.6.540

6. Kingma, D. P., & Ba, J. (2015). Adam: A method for stochastic optimization. International Conference on Learning Representations (ICLR).

https://arxiv.org/abs/1412.6980

7. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 25, 1097–1105.

8. Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., van der Laak, J. A. W. M., van Ginneken, B., & Sánchez, C. I. (2017). A survey on deep learning in medical image analysis. Medical Image Analysis, 42, 60–88.

https://doi.org/10.1016/j.media.2017.07.005

9. Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765–4774.

10. Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). “Why should I trust you?” Explaining the predictions of any classifier. Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 1135–1144.

https://doi.org/10.1145/2939672.2939778

11. Samek, W., Wiegand, T., & Müller, K.-R. (2019). Explainable artificial intelligence: Understanding, visualizing and interpreting deep learning models. IEEE Signal Processing Magazine, 36(3), 56–64. https://doi.org/10.1109/MSP.2018.2881525

12. Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. Proceedings of the IEEE International Conference on Computer Vision (ICCV), 618–626. https://doi.org/10.1109/ICCV.2017.74

13. Tjoa, E., & Guan, C. (2020). A survey on explainable artificial intelligence (XAI): Toward medical XAI. IEEE Transactions on Neural Networks and Learning Systems, 32(11), 4793–4813. https://doi.org/10.1109/TNNLS.2020.3027314

14. Yosinski, J., Clune, J., Bengio, Y., & Lipson, H. (2014). How transferable are features in deep neural networks? Advances in Neural Information Processing Systems, 27, 3320–3328.