Blockchain enabled IoMT and transfer learning for ocular disease classification

Abstract For detecting and diagnosing a wide range of ophthalmological diseases, fundus images are used as a primary and basic tool. Early and accurate diagnosis of these ocular diseases can substantially improve the quality of treatment as well as important for preventing permanent vision loss. Changes in the anatomical structures like the optic disc, macula, blood vessels, and fovea show the presence of diseases like age-related macular degeneration, glaucoma, diabetic retinopathy, cataracts, myopia, and hypertension. In the proposed work, six different automated convolutional neural network architectures based on the Internet of Medical Things (IoMT) using transfer learning techniques were implemented for the classification of fundus images that can detect ocular diseases. These pre-trained neural networks were tuned by modifying the four layers before training them on the dataset. The proposed models incorporate blockchain technology-based private clouds for the security of the patient’s data. Transfer learning provides a promising framework with the combination of IoMT technologies and blockchain technology layers to enhance the diagnosing capabilities of Ocular disease. An ocular disease dataset was used, which was classified into four classes Cataract, Glaucoma, Myopia, and AMD. Class-wise Accuracy, Precision, Sensitivity, F1 Score, Specificity, and Misclassification Rate were computed with up to 96.88% training and 95.40% testing accuracy. The second part of this research is a comparative analysis of implemented models. The performance of AlexNet, GoogleNet, MobileNetV2, DarkNet-19, VGG-19, and DenseNet-201 compared in terms of accuracy. GoogleNet yielded unquestionably impressive results when compared to AlexNet and MobileNetV2.