Deep Learning-Driven Diabetic Retinopathy Classification: From Optimized CNN to Flask-Based Deployment

Emeraldo, Reyhan (2026) Deep Learning-Driven Diabetic Retinopathy Classification: From Optimized CNN to Flask-Based Deployment. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026201095-Undergraduate_Thesis.pdf Download (7MB)

Abstract

Retinopati Diabetik (DR) merupakan salah satu penyebab utama kebutaan akibat diabetes baik di negara maju maupun negara berkembang. Identifikasi dini sangatlah penting, namun metode skrining tradisional sering sulit diterapkan karena memerlukan waktu yang lama dan membutuhkan keahlian khusus. Tujuan dari proyek ini adalah merancang dan mengembangkan sistem otomatis yang dapat mendeteksi retinopati diabetik (DR) menggunakan Convolutional Neural Network (CNN) tingkat lanjut dan aplikasi web berbasis Flask yang mampu melakukan analisis gambar secara real-time. Model dilatih menggunakan dataset EyePACS, yang memungkinkan proses klasifikasi antara retinopati diabetik (DR) dan non-retinopati diabetik (No DR). Beberapa teknik preprocessing diterapkan untuk mengatasi ketidakseimbangan kelas serta meningkatkan performa model. Teknik-teknik tersebut mencakup normalisasi gambar, mengubah ukuran menjadi 150 × 150 piksel, zooming, flipping, dan rotating. Model CNN berhasil mencapai akurasi pengujian sebesar 93% setelah 20 epochs pelatihan. Model juga menunjukkan nilai precision, recall, dan F1-score yang baik. Selain itu, model dirancang agar ringan dan cepat. Flask mempermudah pengguna dalam mengakses model yang telah dilatih. Pengguna dapat mengunggah foto fundus retina melalui antarmuka grafis sederhana dan langsung memperoleh prediksi secara instan. Teknologi ini mampu mendeteksi permasalahan dengan cepat dan akurat. Karya ini memberikan solusi yang sederhana dan dapat diskalakan untuk mendeteksi DR secara dini. Selain itu, proyek ini membuka peluang untuk pengembangan di masa depan, seperti klasifikasi multi-kelas, integrasi explainable AI, serta validasi klinis.
======================================================================================================================================
Diabetic Retinopathy (DR) is a leading cause of diabetes-induced blindness both in developed and developing countries. Identifying problems early is crucial, but traditional screening methods can be difficult to apply because they are time- consuming and require expertise. The objective of this project is to design and develop an automated system that can detect diabetic retinopathy (DR) within the use of an advanced CNN and a Flask-based web app, which performs real-time image analysis. The model was trained on the EyePACS dataset, which facilitate the differentiation between diabetic retinopathy (DR) and non-diabetic retinopathy (No DR). Some preprocessing techniques were applied to mitigate the class imbalance and enhance the model's performance. These techniques included normalising images, resizing to 150 x 150 pixels, zooming, flipping, and rotating. The CNN model achieved a test accuracy of 93% after 20 training epochs. It also had confident precision, good recall, and the F1-score. It was going to be light and fast. Flask made it simple for people to use the trained model. Customers can use a simple graphical interface to upload retinal fundus photos and get predictions right away. The technology was able to quickly and correctly find problems. This work gives us a simple and scalable way to find DR early on. It also sets the stage for future progress, such as multi-class classification, explainable AI integration, and clinical validation.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Retinopati Diabetik (DR), Convolutional Neural Network (CNN), Sistem Deteksi Otomatis, Analisis Gambar Real-Time, Deteksi Dini Penyakit, Aplikasi Web Flask, Diabetic Retinopathy, Convolutional Neural Network, Automated Diagnosis System, Real-Time Image Analysis, Early Disease Detection, Flask Web Application.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Reyhan Emeraldo
Date Deposited:	30 Jan 2026 07:18
Last Modified:	30 Jan 2026 07:18
URI:	http://repository.its.ac.id/id/eprint/131271

Actions (login required)

View Item