Classification of Eight Different Retinal Conditions in Retinal OCT Images Using Convolutional Neural Networks
International Conference on Computer Sciences, Zagreb, Hırvatistan, 17 - 18 Eylül 2026, ss.1508-1515, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Basıldığı Şehir: Zagreb
- Basıldığı Ülke: Hırvatistan
- Sayfa Sayıları: ss.1508-1515
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Gazi Üniversitesi Adresli: Evet
Özet
Abstract In this study, a convolutional neural network-based approach was developed to automatically classify eight retinal conditions from retinal optical coherence tomography (OCT) images. The balanced dataset comprising AMD, CNV, CSR, DME, DR, DRUSEN, MH, and NORMAL classes is divided into training, validation, and test sets in an 80%/10%/10 % split. Images were resized to 224×224 pixels, and the CNN model, consisting of three convolutional blocks, was trained for 10 epochs. The model achieved 98.68% accuracy on the training set, 81.14% on the validation set, and 79.57% on the test set. The results show that the developed model can learn distinctive structural features in retinal OCT images and offers a feasible performance for multi-class classification. However, the difference between training and test performance indicates that the model's generalization capacity needs improvement.
Keywords: Retinal Diseases, Optical Coherence Tomography (OCT), CNN.