Predicting enhanced diagnostic models: deep learning for multi-label retinal disease classification

Abstract

In this study, we assess three convolutional neural network (CNN) architectures—VGG16, ResNet50, and InceptionV3 for multi classification of fundus images in the retinal fundus multi-disease image dataset (RFMID2), comprising of 860 images. Focusing on diabetic retinopathy, exudation, and hemorrhagic retinopathy, we preprocessed the dataset for uniformity and balance. Using transfer learning, the models were adapted for feature extraction and fine-tuned to our multi-label classification task. Their performance was measured by subset accuracy, precision, recall, F1-score, hamming loss, and Jaccard score. VGG16 emerged as the top performer, with the highest subset accuracy (84.81%) and macro precision (95.83%), indicating its superior class distinction capabilities. ResNet50 showed commendable accuracy (79.75%) and precision (86.70%), whereas InceptionV3 lagged with lower accuracy (66.67%) and precision (81.21%). These findings suggest VGG16’s depth offers advantages in multi-label classification, highlighting InceptionV3’s limitations in complex scenarios. This analysis helps optimize CNN architecture selection for specific tasks, suggesting future exploration of dataset variability, ensemble methods, and hybrid models for improved performance.