DEVELOPMENT OF DIABETIC RETINOPATHY EARLY DETECTION AND ITS IMPLEMENTATION IN ANDROID APPLICATION

• Main Author: Amanda, Isca
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/40004
• Institution: Institut Teknologi Bandung

Diabetic retinopathy (DR) is a diabetes complication causing blindness which symptoms are not perceived in earlier stage or nonproliferative diabetic retinopathy (NPDR). Early identification of DR is critical for good prognosis, meanwhile with the growing number of DR patients it is hard to answer the needs of this with manual diagnosis methods. In this final project, an algorithm to detect NPDR is developed and implemented in Android application. Deep neural network and transfer learning method is being used on fundus images to train classifier model. Model development is done with Messidor (4 class) and Messidor-2 (2 class) dataset, image pre-processing, InceptionV3 and MobileNetV1 network, configuration of test set-train set split, optimizer, and learning rate. A test accuracy of 86% is acquired with InceptionV3 and Messidor-2 which then implemented in Android application. It yields an accuracy, sensitivity, and specificity of 88%, 80%, and 76% respectively.