Weakly-Supervised Classification and Localization of Retinal Macular Pathologies in Optical Coherence Tomography Using DenseNet-121 with Explainable Gradient Activation Mapping

In this study, we have formulated and evaluated a comprehensive, weakly-supervised deep learning framework for the automated detection, spatial localization, and clinical uncertainty quantification of retinal macular pathologies from optical coherence tomography B-scans. Utilizing a lightweight DenseNet-121 backbone fine-tuned exclusively on image-level diagnostic labels, the framework achieves an overall macro-averaged accuracy of 99.40%, a sensitivity of 99.40%, and an AUC-ROC of 0.9995 across the rigorous 1,000-image Kermany benchmark test cohort. By coupling deep dense feature extraction with terminal-layer Grad-CAM backpropagation, the system autonomously extracts high-fidelity spatial lesion localizations and bounding boxes without requiring expensive, labor-intensive pixel-level manual segmentation masks. Furthermore, the integration of an inference-time Monte Carlo Dropout Bayesian approximation provides calibrated epistemic uncertainty metrics (predictive standard deviation and Shannon entropy) that trigger automated clinical triage handoffs for borderline scans, while a dual-stage chromatic screening gate completely eliminates diagnostic hallucinations on non-OCT imagery. Future research will focus on extending the architectural framework to 3D volumetric multi-slice spatio-temporal representations, integrating multi-modal fusion with ultra-widefield color fundus photography, and deploying the model within distributed edge computing kiosks to support community-level diabetic retinopathy and macular disease screening programs.

[1] D. S. Kermany, M. Goldbaum, W. Cai, C. C. Valentim, H. Liang, S. L. Baxter, et al., "Identifying medical diagnoses and treatable diseases by image-based deep learning," Cell, vol. 172, no. 5, pp. 1122-1131, Feb. 2018. [2] G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, "Densely connected convolutional networks," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Honolulu, HI, USA, 2017, pp. 4700-4708. [3] R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, "Grad-CAM: Visual explanations from deep networks via gradient-based localization," in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), Venice, Italy, 2017, pp. 618-626. [4] Y. Gal and Z. Ghahramani, "Dropout as a Bayesian approximation: Representing model uncertainty in deep learning," in Proc. 33rd Int. Conf. Mach. Learn. (ICML), New York, NY, USA, 2016, pp. 1050-1059. [5] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, NV, USA, 2016, pp. 770-778. [6] K. Simonyan and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," in Proc. 3rd Int. Conf. Learn. Represent. (ICLR), San Diego, CA, USA, 2015, pp. 1-14. [7] J. De Fauw, J. R. Ledsam, B. Romera-Paredes, S. Nikolov, N. Tomasev, S. Blackwell, et al., "Clinically applicable deep learning for diagnosis and referral in retinal disease," Nature Medicine, vol. 24, no. 9, pp. 1342-1350, Sep. 2018. [8] C. S. Lee, D. M. Baughman, and A. Y. Lee, "Deep learning is effective for classifying normal versus age-related macular degeneration and diabetic macular edema in optical coherence tomography images," Ophthalmology Retina, vol. 1, no. 4, pp. 322-327, Jul. 2017. [9] O. Ronneberger, P. Fischer, and T. Brox, "U-Net: Convolutional networks for biomedical image segmentation," in Medical Image Computing and Computer-Assisted Intervention (MICCAI), Munich, Germany: Springer, 2015, pp. 234-241. [10] B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, and A. Torralba, "Learning deep features for discriminative localization," in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Las Vegas, NV, USA, 2016, pp. 2921-2929.