SEGMENTASI PEMBULUH DARAH PADA CITRA RETINA MENGGUNAKAN ARSITEKTUR RESVNET

SYAFIRA, DIAN RAMADHANI and Erwin, Erwin and Anita, Desiani (2024) SEGMENTASI PEMBULUH DARAH PADA CITRA RETINA MENGGUNAKAN ARSITEKTUR RESVNET. Masters thesis, Sriwijaya University.

	Text RAMA_55101_09012682226004.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (4MB) | Request a copy
	Text RAMA_55101_09012682226004_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (12MB) | Request a copy
	Text RAMA_55101_09012682226004_029017101_0011127702_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (909kB)
	Text RAMA_55101_09012682226004_0029017101_0011127702_02.pdf - Accepted Version Restricted to Repository staff only Download (836kB) | Request a copy
	Text RAMA_55101_09012682226004_0029017101_0011127702_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (836kB) | Request a copy
	Text RAMA_55101_09012682226004_0029017101_0011127702_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (653kB) | Request a copy
	Text RAMA_55101_09012682226004_0029017101_0011127702_04.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (958kB)
	Text RAMA_55101_09012682226004_0029017101_0011127702_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (9kB) | Request a copy
	Text RAMA_55101_09012682226004_0029017101_0011127702_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (179kB) | Request a copy
	Text RAMA_55101_09012682226004_0029017101_0011127702_07_lamp.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (3MB) | Request a copy

Abstract

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.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	CLAHE, U-Net, ResVNet, grayscale, gaussian blur, segmentasi pembuluh darah,
Subjects:	Q Science > Q Science (General) > Q334-342 Computer science. Artificial intelligence. Algorithms. Robotics. Automation.
Divisions:	09-Faculty of Computer Science > 55101-Informatics (S2)
Depositing User:	Syafira Dian Ramadhani
Date Deposited:	25 Nov 2024 02:17
Last Modified:	25 Nov 2024 02:17
URI:	http://repository.unsri.ac.id/id/eprint/159801

Actions (login required)

View Item