Text

SEGMENTASI EXUDATE PADA CITRA RETINA PENYAKIT DIABETIC RETINOPATHY MENGGUNAKAN MODIFIKASI ARSITEKTUR U-NET, DENSE BLOCK, DAN CONVOLUTIONAL LONG SHORT-TERM MEMORY

---

Penilaian

---

Diabetic retinopathy (DR) is an eye disease caused by high glucose and blood pressure levels that can lead to blindness. DR can be detected through the presence of exudates, which are white or yellow spots that form due to damage to the blood vessels of the retina. The process of image segmentation is performed to obtain exudate features in the retinal image. Automatic segmentation development can use the Convolutional Neural Network (CNN) method. U-Net is a CNN architecture that is capable of segmentation because it has encoder and decoder parts. Encoder functions to extract features but has the risk of losing fine image features and vanishing gradient due to the down sampling process. The decoder functions to reconstruct the image but has the risk of losing important information due to skip connection.This research modifies the U-Net architecture by adding Dense Block and Convolutional Long Short-Term Memory (ConvLSTM) for exudate segmentation in retinal images. Dense Block is applied to the encoder part to retain the fine features of the image and overcome the vanishing gradient problem. ConvLSTM is inserted in the decoder to retain important information in the image. This research performs image segmentation with two labels, namely exudate and background. The exudate segmentation results with the application of the proposed architecture obtained an accuracy of 99%, indicating the model is very good at predicting all labels correctly as a whole. Sensitivity of 82%, indicating the model predicts the exudate area well. Specificity of 99%, indicating the model is very good at predicting areas that are not part of the exudate. F1-Score 75%, indicating the model has a good balance between sensitivity and specificity. IoU 61%, indicating a poor level of overlap between the prediction results and the ground truth. The results of this study show that the model predicts and performs well in segmenting exudates in retinal images. However, improvements in sensitivity, F1-Score and IoU are required to optimize the performance of the segmentation model.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2507003123	T174031	T1740312025	Central Library (Reference)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T1740312025

Publisher

: Prodi Ilmu Matematika, Fakultas Matematika Dan Ilmu Pengetahuan Alam Universitas Sriwijaya., 2025

Collation

xvii, 59 hlm.: Ilus., tab.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

510.07

Content Type

-

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Matematika

Specific Detail Info

-

Statement of Responsibility

EM

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment