FEATURES EXTRACTION OF HEP-2 IMMUNOFLUORESCENCE PATTERNS BASED ON TEXTURE AND REGION OF INTEREST TECHNIQUES

Autoimmune disease is a disease that happens when improper immune response in the body fighting against substance, cells and tissues that naturally exists and needed in human’s body. This will later on cause autoantibody disease such as SLE where internal organ failed to perform their basic functions. Antinuclear antibody (ANA) test is a way to test the presence of autoantibodies in individual blood serum. This study focuses on ANA test that is done using indirect immunofluorescence HEp-2 cell coating slides that are used to place the blood serum. However, there are several problems encountered with current technique, such as inaccuracy of the result as the result is viewed by naked eyes of operator. There is no objective definition for positive, negative or border line of infection. This project involves developing features extraction technique of HEp-2 cell of 2 main patterns namely Nucleolar and Centromere using texture and region of interest technique. Next, to design an algorithm that can automatically identify the 2 patterns of the HEp-2 cell tested using ANA. To execute features extraction, image pre-processing is performed to enhance image in terms of its intensity, brightness and contrast. Only clear and good input image will produce good results. Image segmentation will be done after pre-processing completed to further enhance the image according to its edge or region to be used for the input image. Different methods of features extraction will be used and compared for better outcome. To differentiate between one pattern from another, image classification is done by evaluating the properties of internal image from features extraction and a boundary is drawn between Centromere and Nucleolar pattern. The result shows four different types of properties of internal cells which are homogeneity, contrast, energy and correlation. After analysis has been done, energy between Centromere and Nucleolar are different from each other and used to classify the pattern in SVM classifier. Tools used in this study are MATLAB software and image processing tools in MATLAB.