<![CDATA[The incidence of keratitis globally ranges from 0.4 to 5.2 per 10,000 people annually. Keratitis can only be identified by an ophthalmologist using a slitlamp as a fundamental instrument for specific eye examination in secondary care facilities. In primary care facilities, eye specialists and slitlamps are not available. This causes delay in the diagnosis and treatment of keratitis patients in public health centers or areas with limited facilities and access to doctors/ophthalmologists. This research aims to develop a keratitis identification model using the convolutional neural network (CNN) method and training data consisting of images produced by smartphones and combined with slitlamp images. The training accuracy of the developed model is 92% with a dropout layer set at 0.3, and the average validation accuracy is 83%, indicating that the model training did not experience overfitting. The testing results with new data achieved an accuracy of 90%. Next, the parameters of the best model will be integrated into an application running on the Android operating system. However, the application’s functionality and UX/UI performance need to be improved to facilitate seamless use of the model.]]>
Copyrights © 2023
