Analisis Retinopati Diabetik Dengan Implementasi Menggunakan Residual Neural Network

Utomo, Satrio Heru (2024) Analisis Retinopati Diabetik Dengan Implementasi Menggunakan Residual Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07211940000053-Undedrgraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (9MB) | Request a copy

Abstract

Retinopati diabetik adalah komplikasi mikrovaskular diabetes dan merupakan penyebab utama kebutaan di antara orang dewasa usia kerja di seluruh dunia. Deteksi dan intervensi dini sangat penting untuk mencegah kehilangan penglihatan dan meningkatkan hasil pengobatan pasien. Namun, metode skrining tradisional sering kali memiliki keterbatasan dalam hal akurasi dan aksesibilitas. Penelitian ini mengusulkan penerapan Residual Neural Network (ResNet) untuk deteksi dan klasifikasi DR secara otomatis dari gambar fundus. Penelitian ini bertujuan untuk berkontribusi pada kemajuan diagnosis DR otomatis dan pada akhirnya meningkatkan perawatan pasien melalui intervensi dini dan strategi perawatan yang dipersonalisasi. Model paling akurat yang dicapai adalah ResNet-18 yang memiliki akurasi validasi terbaik tanpa penyesuaian apa pun pada beban class, dengan nilai 0,8211. Selain itu, model dengan skor Kappa tertinggi adalah model ResNet-18 yang memiliki akurasi Training terbaik tanpa modifikasi apa pun pada beban class, yang menghasilkan nilai Kappa sebesar 0,7584.
======================================================================================================================
Diabetic retinopathy (DR) is a microvascular complication of diabetes and is the leading cause of blindness among working-age adults worldwide. Early detection and intervention are crucial to prevent vision loss and improve patient outcomes. However, traditional screening methods often face limitations in accuracy and accessibility. This study proposes the implementation of a Residual Neural Network (ResNet) for automated DR detection and classification from fundus images. By achieving these objectives, this study aims to contribute to the advancement of automated DR diagnosis and ultimately improve patient care through early intervention and personalized treatment strategies. The most accurate model, ResNet-18, achieved the best validation accuracy without any adjustment to the class weight, with a value of 0.8211. Additionally, the model with the highest Kappa score was ResNet-18, which had the best training accuracy without any modification to the class weight, resulting in a Kappa score of 0.7584.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Retinopati Diabetik, ResNet, Analisis Angiografi OCT, Diabetic retinopathy, ResNet, Deep Learning, Optical Coherence Tomography Angiography Analysis
Subjects:	R Medicine > R Medicine (General) > R858 Deep Learning R Medicine > RE Ophthalmology > RE48 Eye--Diseases. Ophthalmoscopy.
Divisions:	Faculty of Electrical Technology > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Satrio Heru Utomo
Date Deposited:	26 Jul 2024 07:49
Last Modified:	26 Jul 2024 07:49
URI:	http://repository.its.ac.id/id/eprint/109117

Actions (login required)

View Item