“SEGMENTATION OF ANTI NEUTROPHIL CYTOPLASMIC ANTIBODIES (ANCA) IMAGES BASED ON WATERSHED AND WAVELET”

Autoimmune disease is a type of disease where immune system unable to tell between the good side and bad side which lead to the misguided attack on the healthy cells and tissues. Autoimmune disease can be classified to more than 80 types depending on the affected area. The test also varies according to the suspected type of disease. Some examples of the test are Enzyme-Linked Immunosorbent Assay (ELISA) test, Indirect Immunofluorescence (IIF) test of Antinuclear Antibody (ANA) by using HeP-2 Cells and IIF test for Anti Neutrophil Cytoplasmic Antibodies (ANCA). However in this project, author only focus on the ANCA images with two major staining patterns which are P-ANCA and C-ANCA. Currently the positivity of the images depends solely on the experience of the physician which led to variety of result and lack of reliability. Besides the time to get the result is time consuming. Thus an automatic classification system has been developed to overcome the manual process. The vital process inside the automatic system is the segmentation part. Many researchers suggest different techniques of segmentation to segment the ANCA images before being further processed. In this research, author focus on Watershed technique to segment the ANCA images by implementing the algorithm in Matlab. Author use Wavelet transform to suppress noise to avoid from over segmentation of the ANCA images. Using Rand Index method, the result of segmentations is verified. Combination of Watershed and Wavelet transform gives a very promising result. Recommendation for future work is to explore on automatic determination of noise variance inside images.