Please use this identifier to cite or link to this item: http://oaps.umac.mo/handle/10692.1/83
Title: Retinal Image Analysis System
Authors: CHAN, CHON KIT (陳俊傑)
Department: Department of Computer and Information Science
Faculty: Faculty of Science and Technology
Issue Date: 2015
Citation: CHAN, C. K. (2015). Retinal Image Analysis System (Outstanding Academic Papers by Students (OAPS)). Retrieved from University of Macau, Outstanding Academic Papers by Students Repository.
Abstract: In this project, we separate to three main parts, the Retinal blood vessel extraction, Optic disc detection and the Microaneurysm detection. For retinal blood vessel extraction, accurate extraction of the blood vessel is important task for the computer-aided medical technology. There is a simple filter to match the vessel from the retinal image – Matched Filter (MF), but MF will generate also the non-vessel edges not only the vessels. In our project, we propose an extension of MF and used to detect only the blood vessels exclude the non-vessel edge – Matched Filter compose the Zero-Mean Gaussian function. For optic disc (OD) detection, the result of it is important for developing the automated screening programs. In our project, we propose to use two main steps for the whole algorithm – a. OD vessel candidate detection and OD vessel candidate matching. In OD vessel candidate detection, we can use the method based on our retinal blood vessel extraction to generate the blood vessel result for the OD. In order to complete the OD vessel candidate matching well, we make the original blood vessel result much thinner, it’s used to fit the candidate matching by applied the Vessels’ Directional Matched Filter (VDMF) of various dimensions. The OD location result will then generated by applied VDMF from the OD vessel candidate. For microaneurysm (MA) detection, the microaneurysm can provide the important useful information about the retina, for example the diabetic retinopathy (DR) is a serious diabetes and related to the formation of the lesions in the retina and the serious result will lead to the blindness. Therefore our purpose is to provide the microeurysms’ location and mark at the original retinal image. Mainly our proposed method is separated into two parts, Microaneurysm candidate detection and Microaneurysm classification. The detection part will basically by applying the filter to detect the possible microaneurysm candidates, and the candidates will pass to the classification part in order to check the microaneurysm candidate is true or not. The classification part will calculate by extract 31 different features of each candidate and compare the value with our true microaneurysm statistics to check, after go through all the candidates, the remaining candidates will be our detected result.
Instructor: Prof. ZHANG, YIBO
Programme: Bachelor of Science in Computer Science
URI: http://hdl.handle.net/10692.1/83
Appears in Collections:FST OAPS 2015

Files in This Item:
File Description SizeFormat 
OAPS_2015_FST_006.pdf15.72 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.