A Gaussian Mixture Model-Inspired Convolutional Neural Network with Probabilistic Attention for Diabetic Retinopathy Severity Grading
DOI:
https://doi.org/10.65204/djes.v3i1.728Keywords:
Diabetic retinopathy; Gaussian mixture models; Attention mechanisms; Medical image classification; Convolutional neural networksAbstract
Diabetic retinopathy (DR) is the leading cause of preventable vision loss in working-age adults, affecting about 103 million people, with prevalence expected to rise. Current deep learning methods for DR screening face challenges such as lesion localization uncertainty and limited generalization across diverse datasets. This study presents GMM-CNN, a novel framework combining Gaussian Mixture Model (GMM) mechanisms with convolutional neural networks to overcome these issues. Three synergistic architectural innovations are introduced: (1) a mixture-inspired spatial attention module that generates probabilistic attention weights over K=8 component distributions for multi-focal lesion localization; (2) mask-aware rescaled pooling that prevents feature dilution from sparse pathological lesions; and (3) a GMM-style dense branch that models multi-modal feature distributions across DR severity grades. Evaluated on APTOS 2019 and Messidor-2 benchmark datasets, GMM-CNN achieves 89.27% accuracy (κ=0.8421) and 88.94% accuracy (κ=0.8389) respectively, surpassing state-of-the-art methods by 1.09–3.44 percentage points while using 2.2× fewer parameters than ViT-CapsNet. Cross-dataset evaluation confirms robust generalization across diverse imaging protocols, demonstrating clinical deployment readiness for automated DR screening.
References
References
International Diabetes Federation, "IDF Diabetes Atlas," 10th ed. Brussels, Belgium: IDF, 2021.
Y. Zheng, M. He, and N. Congdon, "The worldwide epidemic of diabetic retinopathy," Indian J. Ophthalmol., vol. 60, no. 5, pp. 428–431, 2022.
S. D. Solomon et al., "Diabetic retinopathy: A position statement by the American Diabetes Association," Diabetes Care, vol. 47, no. Suppl. 1, pp. S142–S153, 2024.
N. Shaukat et al., "Classification and segmentation of diabetic retinopathy: A systemic review," Appl. Sci., vol. 13, no. 5, p. 3108, 2023.
B. Tymchenko, P. Marchenko, and D. Spodarets, "Deep learning approach to diabetic retinopathy detection," in Proc. ICPRAM, 2023, pp. 501–509.
Early Treatment Diabetic Retinopathy Study Research Group, "Photocoagulation for diabetic macular edema: ETDRS report 1," Arch. Ophthalmol., vol. 103, no. 12, pp. 1796–1806, 1985.
R. Taylor and S. Batey, Eds., Handbook of Retinal Screening in Diabetes, 3rd ed. Chichester, UK: Wiley, 2023.
G. Litjens et al., "A survey on deep learning in medical image analysis," Med. Image Anal., vol. 42, pp. 60–88, 2022.
V. Gulshan et al., "Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs," JAMA, vol. 316, no. 22, pp. 2402–2410, 2016.
K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proc. IEEE CVPR, 2016, pp. 770–778.
C. Szegedy et al., "Rethinking the Inception architecture for computer vision," in Proc. IEEE CVPR, 2016, pp. 2818–2826.
M. Tan and Q. V. Le, "EfficientNet: Rethinking model scaling for convolutional neural networks," in Proc. ICML, 2019, pp. 6105–6114.
M. Aatila et al., "Diabetic retinopathy classification using ResNet50 and VGG-16 pretrained networks," Int. J. Comput. Eng. Data Sci., vol. 3, no. 2, pp. 1–7, 2023.
A. M. Abdulghani et al., "Vision Health Monitoring System for Pilots," in Proc. CSCI, Las Vegas, NV, USA, 2023, pp. 1406–1411.
N. S. Shaik and T. K. Cherukuri, "Hinge attention network: A joint model for diabetic retinopathy severity grading," Appl. Intell., vol. 52, no. 13, pp. 15105–15121, 2022.
J. Fan et al., "A hybrid model merging convolutional neural network and differential vision transformer for diabetic retinopathy identification," Biomed. Signal Process. Control, vol. 115, p. 109435, 2026.
M. T. Al-Antary and Y. Arafa, "Multi-scale attention network for diabetic retinopathy classification," IEEE Access, vol. 10, pp. 54190–54200, 2022.
R. Fan, Y. Liu, and R. Zhang, "Multi-scale feature fusion with adaptive weighting for diabetic retinopathy severity classification," Electronics, vol. 11, no. 12, p. 1369, 2022.
X. Li et al., "CANet: Cross-disease attention network for joint DR and DME grading," IEEE Trans. Med. Imaging, vol. 39, no. 5, pp. 1483–1493, 2020.
M. R. Islam et al., "Applying supervised contrastive learning for the detection of diabetic retinopathy," Comput. Biol. Med., vol. 146, p. 105602, 2022.
C. M. Bishop, Pattern Recognition and Machine Learning. New York, NY, USA: Springer, 2006.
K. Van Leemput et al., "Automated model-based tissue classification of MR images of the brain," IEEE Trans. Med. Imaging, vol. 18, no. 10, pp. 897–908, 1999.
A. M. Abdulghani et al., "Network-aware Gaussian mixture models for multi-objective SD-WAN controller placement," Electronics, vol. 14, no. 15, p. 3044, 2025.
A. M. Abdulghani et al., "Dynamic multi-objective controller placement in SD-WAN: A GMM-MARL hybrid framework," Network, vol. 5, no. 4, p. 52, 2025.
A. M. Abdulghani et al., "Enhancing Healthcare Network Effectiveness Through SD-WAN Innovations," in Studies in Systems, Decision and Control, vol. 233. Springer, Cham, 2025.
S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, "CBAM: Convolutional block attention module," in Proc. ECCV, 2018, pp. 3–19.
J. D. Bodapati et al., "Composite deep neural network with gated-attention mechanism for diabetic retinopathy severity classification," J. Ambient Intell. Humaniz. Comput., vol. 14, no. 10, pp. 9825–9839, 2023.
M. Oulhadj et al., "Diabetic retinopathy prediction based on vision transformer and modified capsule network," Comput. Biol. Med., vol. 175, p. 108523, 2024.
S. S. Mondal et al., "EDLDR: An ensemble deep learning technique for detection and classification of diabetic retinopathy," Diagnostics, vol. 13, no. 1, p. 124, 2023.
