Optimizing Clinical Decision-Making through the Integration of MobileNetV2 and Explainable AI in a Diabetic Retinopathy Screening System

Muhammad Faiz Abror; Nanang Hoesen Hidroes Abbrori

doi:10.36456/qj9j1m42

Authors

Muhammad Faiz Abror Digital Business Study Program, Universitas Negeri Surabaya
Nanang Hoesen Hidroes Abbrori Digital Business Study Program, Universitas Negeri Surabaya

DOI:

https://doi.org/10.36456/qj9j1m42

Keywords:

Clinical Decision Support System, Explainable AI, Grad-CAM, MobileNetV2, Retinopathy Diabetic

Abstract

Diabetic Retinopathy is a severe diabetic complication and a leading cause of global blindness. While recent studies predominantly rely on computationally heavy deep learning models that evaluate images in isolation without considering patient clinical histories, their implementation in primary healthcare is frequently hindered by massive hardware demands and the inherent opacity of the algorithms. Addressing this research gap, this study aims to develop an efficient and interpretable hybrid Clinical Decision Support System (CDSS) for Diabetic Retinopathy detection. The proposed methodology integrates a lightweight Convolutional Neural Network architecture, MobileNetV2, to extract pathological features from retinal fundus images. To mitigate the limitations of purely image-centric models, the network's probability outputs are dynamically calibrated utilizing a Rule-Based algorithm that processes real-world patient medical parameters, specifically age and the duration of the diabetic history. Furthermore, the system incorporates Grad-CAM-based Explainable Artificial Intelligence (XAI) to generate real-time spatial mappings of retinal damage. The results demonstrate that MobileNetV2 achieves stable computational convergence while maintaining a highly optimal Sensitivity (Recall) rate. The hybrid mechanism successfully improves diagnostic decisions by contextually adjusting risk weights according to the patient's medical history. Additionally, the XAI evaluation through spatial heatmap projections accurately localizes retinal lesions, offering valid anatomical justification for clinical practitioners. In conclusion, the comprehensive integration of lightweight computational efficiency, clinical parametric logic, and analytical transparency produces a robust, hardware-friendly screening platform that is highly applicable for accelerating patient triage in primary healthcare facilities.

References

S. Webber, International Diabetes Federation, vol. 102, no. 2. 2021. doi: 10.1016/j.diabres.2013.10.013.

[2] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet Classification with Deep Convolutional Neural Networks,” pp. 60-1-60–15, 2012, doi: 10.1201/9781420010749.ch60.

[3] V. Gulshan et al., “Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs,” JAMA - Journal of the American Medical Association, vol. 316, no. 22, pp. 2402–2410, 2016, doi: 10.1001/jama.2016.17216.

[4] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” 3rd International Conference on Learning Representations, ICLR 2015 - Conference Track Proceedings, pp. 1–14, 2015.

[5] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition,” ChemInform, vol. 37, no. 50, pp. 1–9, 2006, doi: 10.1002/chin.200650130.

[6] E. Decencière et al., “Feedback on a publicly distributed image database: The Messidor database,” Image Analysis and Stereology, vol. 33, no. 3, pp. 231–234, 2014, doi: 10.5566/ias.1155.

[7] I. Kandel and M. Castelli, “Transfer learning with convolutional neural networks for diabetic retinopathy image classification. A review,” Applied Sciences (Switzerland), vol. 10, no. 6, 2020, doi: 10.3390/app10062021.

[8] H. Pratt, F. Coenen, D. M. Broadbent, S. P. Harding, and Y. Zheng, “Convolutional neural networks for diabetic retinopathy,” Ingenius, vol. 00, no. 33, pp. 91–101, 2025, doi: 10.17163/ings.n33.2025.08.

[9] D. A. Suryaningrum, A. Rasyid, and A. Wiratama, “Klasifikasi retinopati diabetik pada citra retina menggunakan resnet50 dengan perbandingan vgg16 dan mobilenetv2,” vol. 11, no. 2, pp. 1627–1637, 2026.

[10] K. Waruwu and S. Arlis, “IMPLEMENTASI MODEL CONVOLUTIONAL NEURAL NETWORK BERBASIS ARSITEKTUR RESNET50 UNTUK KLASIFIKASI PADA,” vol. 3, no. 1, pp. 1–7, 2026.

[11] J. Zhou, C. Hu, J. Chen, J. Liu, and X. Li, “Artificial Intelligence in Screening and Grading Diabetic Eye Diseases: A Systematic Review From Algorithms to Clinic,” Diabetes Obes. Metab., pp. 4286–4302, 2026, doi: 10.1111/dom.70621.

[12] G. Rajarajeshwari and G. C. Selvi, “Application of artificial intelligence for Classificatioin, Segmentation, Early Detection, Early Diagnonsis, adn Grading of Diabetic Retinopathy From Fundus Retinal Images : A Comprehesive Review,” 1988. doi: 10.1145/43857.43869.

[13] A. M. Fahmy, “Assessing the Efficacy of AI Models in Medical Imaging Analysis for the Early Identification of Diabetic Foot Ulcers and Gangrene: A Review,” Premier Journal of Science, vol. 12, pp. 1–14, 2025, doi: 10.70389/PJS.100093.

[14] A. G. Howard et al., “MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications,” 2017, [Online]. Available: http://arxiv.org/abs/1704.04861

[15] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L. C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 4510–4520, 2018, doi: 10.1109/CVPR.2018.00474.

[16] V. Lakshminarayanan, H. Kheradfallah, A. Sarkar, and J. J. Balaji, “Automated detection and diagnosis of diabetic retinopathy: A comprehensive survey,” J. Imaging, vol. 7, no. 9, 2021, doi: 10.3390/jimaging7090165.

[17] J. Saeidian et al., “Segmentation of choroidal area in optical coherence tomography images using a transfer learning-based conventional neural network: a focus on diabetic retinopathy and a literature review,” BMC Med. Imaging, vol. 24, no. 1, 2024, doi: 10.1186/s12880-024-01459-2.

[18] E. Maliagkani et al., “Artificial Intelligence Algorithms for Epiretinal Membrane Detection, Segmentation and Postoperative BCVA Prediction: A Systematic Review and Meta-Analysis,” Applied Sciences (Switzerland), vol. 15, no. 22, pp. 1–29, 2025, doi: 10.3390/app152212280.

[19] G. James, T. Hastie, R. Tibshirani, and D. Witten, “An Introduction to Statistical Learning,” Springer Texts, vol. 102, p. 618, 2023.

[20] M. Rezky, K. R. Syarif, and Y. M., “PENERAPAN KECERDASAN BUATAN (AI) DALAM PELAYANAN KEPERAWATAN DAN KESEHATAN: TREN, TANTANGAN, DAN IMPLIKASI ETIS.,” Jurnal Media Keperawatan: Politeknik Kesehatan Makassar, vol. 12, no. 1, pp. 149–160, 2021.

[21] E. Tjoa and C. Guan, “A Survey on Explainable Artificial Intelligence (XAI): Toward Medical XAI,” IEEE Trans. Neural Netw. Learn. Syst., vol. 32, no. 11, pp. 4793–4813, 2021, doi: 10.1109/TNNLS.2020.3027314.

[22] R. Hamza Ali and A. Katham Mtasher, “Personalized Medicine and Explainable AI for Diabetic Retinopathy and Tongue Segmentation: A Systematic Review of Current Approaches and Future Directions,” Global Scientific Journal of Computer Science, vol. 8, no. 2, pp. 2023–2029, 2023, doi: 10.5281/gsjcs.2023.8174203.

[23] D. P. Kingma and J. L. Ba, “Adam: A method for stochastic optimization,” 3rd International Conference on Learning Representations, ICLR 2015 - Conference Track Proceedings, pp. 1–15, 2015.

[24] T. Fawcett, “An introduction to ROC analysis,” Pattern Recognit. Lett., vol. 27, no. 8, pp. 861–874, 2006, doi: 10.1016/j.patrec.2005.10.010.