Skripsi

SEGMENTASI PEMBULUH DARAH PADA CITRA RETINA MENGGUNAKAN ARSITEKTUR RESVNET

---

The use of retinal images on the eye fundus is becoming an important tool in the medical world, especially for the diagnosis of eye diseases. Segmentation of blood vessels in retinal images is one of the important steps in medical analysis to detect and diagnose various eye diseases, such as diabetic retinopathy, glaucoma, and macular degeneration. Using the ResVNet method, segmentation is performed with a deep learning approach that combines the advantages of the residual network (ResNet) and V-Net, which is a U-Net-based architecture combining the strengths of ResNet and U-Net. ResNet with its residual approach allows for more effective training of very deep networks, thus capturing more complex and abstract features. The initial stage in image processing is to improve image quality by noise removal, aiming to increase accuracy in the segmentation and image extraction process. On the other hand, V-Net is designed to work effectively with medical data that has three dimensions, although in the case of retinal image segmentation, it is adapted to work on two-dimensional data. By combining ResNet and V-Net, ResVNet is able to identify important features of blood vessels in retinal images. The image quality improvement steps in preprocessing involve grayscale, gaussian blur and clahe. The methods used for blood vessel segmentation are the residual network ResVNet (ResNet) and V-Net methods. Evaluation of the results of applying image quality enhancement and segmentation techniques using the U-Net method was performed on the DRIVE, STARE, training and testing datasets. The measurement results of blood vessel segmentation using the U-Net method on the DRIVE dataset (accuracy 96.57%, sensitivity 96.27%, precission 97.97%, and jaccard score 96.31%), STARE dataset (accuracy 96.71%, sensitivity 96.28%, precission 98.14%, and jaccard score 96,48%), Of the two datasets used, the STARE dataset obtained better results than the DRIVE dataset, where STARE showed better performance in terms of accuracy , sensitivity , precision and Jaccard score. The focus of the next research is to classify diseases on retinal images.

---

Availability

Detail Information

Series Title

-

Call Number

T1598012024

Publisher

Indralaya : Program Magister Ilmu Komputer, Fakultas Ilmu Komputer Universitas Sriwijaya., 2024

Collation

xviii, 151 hlm.; ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

006.370 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Visi Komputer
Program Magister Ilmu Komputer

Specific Detail Info

-

Statement of Responsibility

TUTI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment