MedDetect AI: A Deep Learning System for Multi-Class Medical Image Diagnosis with Grad-CAM Visualization

This paper presented MedDetect AI V2, a full-stack deep learning diagnostic system addressing the core limitations of its predecessor: from-scratch initializa-tion, class imbalance, overconfident predictions, and lack of interpretability. By adopting EfficientNet architectures with Im-ageNet pretraining, implementing two-phase fine-tuning, combining class weights with focal loss and targeted oversampling, and integrating Grad-CAM visualization, the system achieves 98.47% kidney clas-sification accuracy and 88%+ brain classification ac-curacy while providing clinically interpretable visual explanations. The publicly deployed system demonstrates that high-accuracy, interpretable medical AI can be built and deployed at zero cost using open datasets, open-source frameworks, and free cloud infrastructure, making it accessible for academic and clinical re-search contexts in resource-constrained settings.  

[1]     World Health Organization, “Global Health Workforce Statistics,” WHO, Geneva, 2023. [2]    R. Litjens et al., “A Survey on Deep Learning in Medical Image Analysis,” Medical Image Anal-ysis, vol. 42, pp. 60-88, 2017. [3]    N. Abiwinanda et al., “Brain Tumor Classifica-tion Using Convolutional Neural Network,” in Proc. World Congress on Medical Physics and Biomedical Engineering, 2019. [4]    J. Cheng et al., “Enhanced Performance of Brain Tumor Classification via Tumor Region Aug-mentation and Partition,” PLOS ONE, vol. 10, no. 10, 2015. [5]    M. N. Islam et al., “A Novel Deep Learning Based Multi-class Classification Method for Alzheimer’s Disease Detection Using Brain MRI,” in Proc. BRAIN Informatics, 2018. [6]    R. R. Selvaraju et al., “Grad-CAM: Visual Ex-planations from Deep Networks via Gradient-based Localization,” in Proc. IEEE ICCV, 2017, pp. 618-626. [7]     A. Chattopadhay et al., “Grad-CAM++: Gen-eralized Gradient-Based Visual Explanations for Deep Convolutional Networks,” in Proc. IEEE WACV, 2018. [8]    M. Buda, A. Maki, and M. A. Mazurowski, “A Systematic Study of the Class Imbalance Prob-lem in Convolutional Neural Networks,” Neural Networks, vol. 106, pp. 249-259, 2018. [9]    J. M. Johnson and T. M. Khoshgoftaar, “Survey on Deep Learning with Class Imbalance,” Jour-nal of Big Data, vol. 6, no. 1, pp. 1-54, 2019. [10]     M. Tan and Q. V. Le, “EfficientNet: Rethink-ing Model Scaling for Convolutional Neural Net-works,” in Proc. ICML, 2019, pp. 6105-6114.