Web Application for Early Cataract Detection Using a Deep Learning Cloud Service

• Autores: Galindo-Vilca, Fatima Dayana, Astorayme-Garcia, Fredy Daniel, Aliaga-Cerna, Esther
• Formato: artículo
• Fecha de Publicación: 2024
• Institución: Universidad Peruana de Ciencias Aplicadas
• Repositorio: UPC-Institucional
• Lenguaje: inglés
• OAI Identifier: oai:repositorioacademico.upc.edu.pe:10757/676094
• Enlace del recurso: http://hdl.handle.net/10757/676094
• Nivel de acceso: acceso embargado
• Materia: Azure Custom Vision; Cataract; Deep Learning; Fundus image; Web Application

Cataracts are a degenerative disease that causes opacity in the crystalline lens. They represent one of the leading causes of blindness worldwide, making early detection crucial to prevent severe damage to patients. Current studies on cataract detection face limitations, particularly due to the high cost of imaging devices and their limited accessibility for users. In this study, we propose a web application that utilizes a Deep Learning service to analyze fundus images and provide a cataract diagnosis. This application aims to assist healthcare personnel in medical centers lacking specialist ophthalmologists or facing limited resources for cataract diagnosis. We designed the physical architecture of the application using Azure services, enabling its deployment and operation in the cloud. Azure Custom Vision facilitated the training of our model with a dataset of 1446 fundus images, encompassing both cataract and non-cataract cases. Subsequently, we implemented the web application using React.js and Express.js technologies, integrating the Deep Learning model to perform diagnoses through the web interface. The results demonstrated that the model achieved sensitivity, specificity, precision, and accuracy levels exceeding 90%, showcasing that our proposed tool allows for reliable initial cataract diagnoses in patients without the need for high-cost equipment.