An Intelligent Detection System for Rheumatoid Arthritis (RA) Disease using Image Processing

Hajyyev, Abdyrahym (2014) An Intelligent Detection System for Rheumatoid Arthritis (RA) Disease using Image Processing. [Final Year Project] (Unpublished)

[thumbnail of Abdyrahym Hajyyev_12989_EE_Final Project  Report.pdf]
Abdyrahym Hajyyev_12989_EE_Final Project Report.pdf

Download (2MB) | Preview


Rheumatoid Arthritis (RA) is an autoimmune disease that causes chronic pain, stiffness, redness or loss of function in the joints. Other than early diagnosis, there is yet a cure available for RA. Diseases with similar symptoms such as lupus, osteoarthritis, gout cause difficulty in diagnosing RA. Currently, indirect immunofluorescence (IIF) test performed to identify ANA in Hep-2 cells. Thus, image processing techniques vital to make diagnosis more efficient, accurate and less time-consuming.
For this project standardized staining pattern classifier to be designed by using image processing techniques. Current manual techniques has limited accuracy and time consuming. In IFF procedures, unsuitable microscope to read Hep-2 cell slides, or photo bleaching effect where cells bleached extremely in short period of time are disadvantages. Another downside is test results being subject to change with experts knowledge and years of experience. These factors lead to low accuracy and it becomes a lengthy process due to large number of images. Out of five types of staining patterns nucleolar and centromere share similar visual appearance and the same is true to homogeneous, fine-speckled, coarse-speckled patterns. This is one of the major factors affecting classification accuracy due to results being subjective.
In this research, First and Second Order Statistics Feature Extraction, Mamdani Fuzzy Logic Classification methods utilized to develop automatic detection system for RA with the help of Matlab R2012b, Fuzzy Logic Toolbox, and Image Processing Toolbox. The algorithm tested on the publicly available Mivia Hep-2 Cell image dataset.
Fuzzy logic classified 85 out of 250 images wrongly. It has 66% accuracy. The images obtained from MIVIA dataset has been manually segmented to cell level from the image level. Developing an automated segmentation algorithm might give better results.

Item Type: Final Year Project
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Departments / MOR / COE: Engineering > Electrical and Electronic
Depositing User: Users 2053 not found.
Date Deposited: 09 Oct 2014 16:13
Last Modified: 25 Jan 2017 09:37

Actions (login required)

View Item
View Item