Khalid Maimoona, Sajid Muhammad Zaheer, Youssef Ayman, Khan Nauman Ali, Hamid Muhammad Fareed, Abbas Fakhar
Department of Computer Software Engineering, Military College of Signals, National University of Science and Technology, Islamabad 44000, Pakistan.
Department of Computers and Systems, Electronics Research Institute, Cairo 11843, Egypt.
Diagnostics (Basel). 2024 Nov 27;14(23):2679. doi: 10.3390/diagnostics14232679.
Diabetic retinopathy, hypertensive retinopathy, glaucoma, and contrast-related eye diseases are well-recognized conditions resulting from high blood pressure, rising blood glucose, and elevated eye pressure. Later-stage symptoms usually include patches of cotton wool, restricted veins in the optic nerve, and buildup of blood in the optic nerve. Severe consequences include damage of the visual nerve, and retinal artery obstruction, and possible blindness may result from these conditions. An early illness diagnosis is made easier by the use of deep learning models and artificial intelligence (AI). This study introduces a novel methodology called CAD-EYE for classifying diabetic retinopathy, hypertensive retinopathy, glaucoma, and contrast-related eye issues. The proposed system combines the features extracted using two deep learning (DL) models (MobileNet and EfficientNet) using feature fusion to increase the diagnostic system efficiency. The system uses fluorescence imaging for increasing accuracy as an image processing algorithm. The algorithm is added to increase the interpretability and explainability of the CAD-EYE system. This algorithm was not used in such an application in the previous literature to the best of the authors' knowledge. The study utilizes datasets sourced from reputable internet platforms to train the proposed system. The system was trained on 65,871 fundus images from the collected datasets, achieving a 98% classification accuracy. A comparative analysis demonstrates that CAD-EYE surpasses cutting-edge models such as ResNet, GoogLeNet, VGGNet, InceptionV3, and Xception in terms of classification accuracy. A state-of-the-art comparison shows the superior performance of the model against previous work in the literature. These findings support the usefulness of CAD-EYE as a diagnosis tool that can help medical professionals diagnose an eye disease. However, this tool will not be replacing optometrists.
糖尿病视网膜病变、高血压视网膜病变、青光眼以及与造影剂相关的眼部疾病是由高血压、血糖升高和眼压升高导致的公认病症。后期症状通常包括棉絮斑、视神经中静脉变细以及视神经内血液淤积。严重后果包括视神经损伤、视网膜动脉阻塞，这些情况可能导致失明。利用深度学习模型和人工智能（AI）可使疾病早期诊断更加容易。本研究引入了一种名为CAD-EYE的新方法，用于对糖尿病视网膜病变、高血压视网膜病变、青光眼以及与造影剂相关的眼部问题进行分类。所提出的系统结合了使用两种深度学习（DL）模型（MobileNet和EfficientNet）提取的特征，并通过特征融合来提高诊断系统的效率。该系统使用荧光成像作为一种图像处理算法来提高准确性。添加该算法是为了增强CAD-EYE系统的可解释性。据作者所知，该算法在以往文献中未用于此类应用。本研究利用从知名互联网平台获取的数据集来训练所提出的系统。该系统在收集的数据集中的65,871张眼底图像上进行训练，分类准确率达到98%。对比分析表明，在分类准确率方面，CAD-EYE超过了ResNet、GoogLeNet、VGGNet、InceptionV3和Xception等前沿模型。与现有技术的比较显示了该模型相对于文献中先前工作的卓越性能。这些发现支持了CAD-EYE作为一种诊断工具的实用性，它可以帮助医学专业人员诊断眼部疾病。然而，这个工具不会取代验光师。
