Retina Hastalıklarının Tespiti için Yerel ve Küresel Özellik Öğreniminin Entegrasyonu: EfficientNetB3-ViT Hibrit Modeli

Öz

Başlangıç aşamasında teşhis edilemeyen göz rahatsızlıkları, görmeyi giderek bozarak kritik görme bozukluklarına yol açabilir. Fundus görüntüleme, özellikle diyabetik retinopati, glokom ve yaşa bağlı makula dejenerasyonu gibi çeşitli retina hastalıklarının tanımlanmasını sağlamıştır. Bununla birlikte, bu görüntülerin manuel olarak değerlendirilmesi hem zaman alıcıdır hem de uzman görüşüne bağlıdır. Bu bağlamda, yapay zekâ tabanlı otomatik tanı sistemleri, tıbbi görüntü analizi alanında önemli bir ihtiyaç haline gelmiştir. Bu çalışmada, fundus görüntülerinden çok sınıflı göz hastalığı tespiti için derin öğrenme tabanlı modeller karşılaştırılmış ve yeni bir hibrit model önerilmiştir. Bu çalışmada, EfficientNetB3, DenseNet-121, ResNet-50, MobileNetV2 ve ViT dahil olmak üzere mevcut derin öğrenme mimarileri kullanılmıştır. Tüm modeller bağımsız olarak eğitilmiş ve test edilmiş ve etkinlikleri yaygın olarak kullanılan sınıflandırma metrikleri kullanılarak değerlendirilmiştir. Ek olarak, Transformer ve CNN mimarilerinin güçlü yönlerini birleştiren EfficientNetB3 ve ViT tabanlı bir hibrit model tasarlanmıştır. Tüm modeller, adil bir karşılaştırma sağlamak amacıyla GridSearch yöntemi kullanılarak hiperparametre optimizasyonuna tabi tutulmuştur. Elde edilen sonuçlar, hibrit modelin fundus görüntülerindeki hem yerel hem de küresel özellikleri başarıyla öğrendiğini ve diğer modellere kıyasla daha yüksek performans sergilediğini göstermiştir. Bu bağlamda, geliştirilen sistemin gelecekteki klinik karar destek sistemlerine entegre edilme potansiyeli bulunmaktadır.

Anahtar Kelimeler

• Göz hastalıkları
• derin öğrenme
• efficientnet
• vision transformer

Integrating Local and Global Feature Learning for Retinal Disease Detection: An EfficientNetB3-ViT Hybrid Model

Öz

Undiagnosed ocular conditions in initial phases can progressively impair vision, potentially leading to critical visual dysfunctions. Fundus imaging has enabled the identification of several retinal disorders, notably diabetic retinopathy, glaucoma, and age-related macular degeneration. However, manual evaluation of these images is both time-consuming and dependent on expert judgment. In this context, artificial intelligence-based automatic diagnosis systems have become an important need in the field of medical image analysis. In this study, deep learning-based models were compared for multi-class eye disease detection from fundus images, and a unique hybrid model was proposed. In this study, current deep learning architectures, including EfficientNetB3, DenseNet-121, ResNet-50, MobileNetV2, and ViT, were employed. All models were independently trained and tested, and their effectiveness was evaluated using commonly employed classification metrics. Additionally, a hybrid model based on EfficientNetB3 and ViT was designed, combining the strengths of Transformer and CNN architectures. All models were subjected to hyperparameter optimization using the GridSearch method, allowing for a fair comparison. The results obtained showed that the hybrid model successfully learned both local and global features in fundus images and exhibited higher performance compared to other models. In this context, the developed system has the potential to be integrated into future clinical decision-support systems.

Anahtar Kelimeler

• Ocular disease
• deep learning
• efficientnet
• vision transformer

Kaynakça

1. Steinmetz J D, Bourne R R, Briant P S, Flaxman S R, Taylor H R, Jonas J B, Morse A R F. Causes of blindness and vision impairment in 2020 and trends over 30 years, and prevalence of avoidable blindness in relation to VISION 2020: the Right to Sight: an analysis for the Global Burden of Disease Study. The Lancet Global Health 2021; 9(2): e144-e160.
2. Vergroesen J E, Thee E F, Ahmadizar F, van Duijn C M, Stricker B H, Kavousi M, Ramdas W D. Association of diabetes medication with open-angle glaucoma, age-related macular degeneration, and cataract in the Rotterdam study. JAMA ophthalmology 2022; 140(7): 674-681.
3. Deme T G, Mengistu M, Getahun F. Prevalence and associated factors of visual impairment among adults aged 40 and above in Southern Ethiopia, 2022. Scientific Reports 2024; 14(1): 2542.
4. Setyawati R, Astuti A, Utami T P, Adiwijaya S, Hasyim D M. The importance of early detection in disease management. Journal of World Future Medicine, Health and Nursing 2024; 2(1): 51-63.
5. Marmamula S, Yelagondula V K, Varada R, Khanna R C. Improving access to eye care in low and middle-income countries–challenges, opportunities, and the way forward. Expert Review of Ophthalmology 2023; 18(6): 365-377.
6. Haleem A, Javaid M, Singh R P, Suman R. Medical 4.0 technologies for healthcare: Features, capabilities, and applications. Internet of Things and Cyber-Physical Systems 2022; 2: 12-30.
7. Puttagunta M, Ravi S. Medical image analysis based on deep learning approach. Multimedia tools and applications 2021; 80(16): 24365-24398.
8. Sarvamangala D R, Kulkarni R V. Convolutional neural networks in medical image understanding: a survey. Evolutionary intelligence 2022; 15(1): 1-22.

9. Khan S, Naseer M, Hayat M, Zamir S W, Khan F S, Shah M. Transformers in vision: A survey. ACM computing surveys (CSUR) 2022, 54(10): 1-41.
10. Khan A, Rauf Z, Sohail A, Khan A R, Asif H, Asif A, Farooq U. A survey of the vision transformers and their CNN-transformer based variants. Artificial Intelligence Review 2023; 56(3): 2917-2970.
11. Ergün E. High precision banana variety identification using vision transformer based feature extraction and support vector machine. Scientific Reports 2025; 15(1): 10366.
12. Emir B, Colak E. Performance analysis of pretrained convolutional neural network models for ophthalmological disease classification. Arquivos brasileiros de oftalmologia 2023; 87: e2022-0124.
13. Li Z, Xu M, Yang X, Han Y. Multi-label fundus image classification using attention mechanisms and feature fusion. Micromachines 2022; 13(6).
14. Lu Z, Miao J, Dong J, Zhu S, Wu P, Wang X, Feng J. Automatic multilabel classification of multiple fundus diseases based on convolutional neural network with squeeze-and-excitation attention. Translational Vision Science & Technology 2023; 12(1): 22-22.
15. Abbas Q, Albathan M, Altameem A, Almakki R S, Hussain A. Deep-Ocular: Improved Transfer Learning Architecture Using Self-Attention and Dense Layers for Recognition of Ocular Diseases. Diagnostics 2023; 13(20).
16. Wang D, Lian J, Jiao W. Multi-label classification of retinal disease via a novel vision transformer model. Frontiers in Neuroscience 2024; 17.
17. Wahab Sait A R. Artificial intelligence-driven eye disease classification model. Applied Sciences 2023; 13(20): 11437.
18. Al-Fahdawi S, Al-Waisy A S, Zeebaree D Q, Qahwaji R, Natiq H, Mohammed M A, Deveci M. Fundus-deepnet: Multi-label deep learning classification system for enhanced detection of multiple ocular diseases through data fusion of fundus images. Information Fusion 2024; 102.
19. Adnan F, Awan M J, Mahmoud A, Nobanee H, Yasin A, Zain A M. EfficientNetB3-adaptive augmented deep learning (AADL) for multi-class plant disease classification. IEEE Access 2023; 11:85426-85440.
20. Bazi Y, Al Rahhal M M, Alhichri H, Alajlan N. Simple yet effective fine-tuning of deep CNNs using an auxiliary classification loss for remote sensing scene classification. Remote Sensing 2019; 11(24).
21. Assaduzzaman M, Bishshash P, Nirob M A S, Al Marouf A, Rokne J G, Alhajj R. XSE-TomatoNet: An explainable AI based tomato leaf disease classification method using EfficientNetB0 with squeeze-and-excitation blocks and multi-scale feature fusion. MethodsX 2025; 14.
22. Wang X, Ren H, Wang A. Smish: A novel activation function for deep learning methods. Electronics 2022; 11(4).
23. Swaminathan A, Varun C, Kalaivani S. Multiple plant leaf disease classification using densenet-121 architecture. Int. J. Electr. Eng. Technol 2021; 12(5): 38-57.
24. Hou Y, Wu Z, Cai X, Zhu T. The application of improved densenet algorithm in accurate image recognition. Scientific Reports 2024; 14(1).
25. Alom M Z, Taha T M, Yakopcic C, Westberg S, Sidike P, Nasrin M S, Asari V K. A state-of-the-art survey on deep learning theory and architectures. Electronics 2019; 8(3).
26. Bialopetravičius J, Narbutis D. Deriving star cluster parameters with convolutional neural networks-II. Extinction and cluster-background classification. Astronomy & Astrophysics 2020; 633.
27. Cong S, Zhou Y. A review of convolutional neural network architectures and their optimizations. Artificial Intelligence Review 2023; 56(3): 1905-1969.
28. Kulkarni U, Meena S M, Gurlahosur S V, Bhogar G. Quantization friendly mobilenet (qf-mobilenet) architecture for vision based applications on embedded platforms. Neural Networks 2021; 136: 28-39.
29. Moutsis S N, Tsintotas K A, Kansizoglou I, Gasteratos A. Evaluating the performance of mobile-convolutional neural networks for spatial and temporal human action recognition analysis. Robotics 2023; 12(6).
30. Lokhande H, Ganorkar S R. Object detection in video surveillance using MobileNetV2 on resource-constrained low-power edge devices. Bulletin of Electrical Engineering and Informatics 2025; 14(1): 357-365.
31. Hassanin M, Anwar S, Radwan I, Khan F S, Mian A. Visual attention methods in deep learning: An in-depth survey. Information Fusion 2024; 108.
32. Ding S, He D, Liu G. Improving Short-Term Load Forecasting with Multi-Scale Convolutional Neural Networks and Transformer-Based Multi-Head Attention Mechanisms. Electronics 2024 13(24).
33. Maurício J, Domingues I, Bernardino J. Comparing vision transformers and convolutional neural networks for image classification: A literature review. Applied Sciences 2023; 13(9).