Application of fuzzy c-mean clustering technique for mycobacterium tuberculosis detection in Ziehl-Neelsen stained tissue images

Abstract

Automatic detection of Mycobacterium tuberculosis improves accuracy, sensitivity and efficiency of diagnosis compared to manual method. However, the process is difficult, especially in Zeihl-Neelsen stained tissue images due to intensity inhomogeneity and tissue background complexity. In this paper, an automated approach to segment Mycobacterium tuberculosis from tissue slide images using fuzzy c-mean clustering procedure is proposed. The procedure provides a basic step for detecting the presence of tuberculosis bacilli. First, initial filter is used to assist the clustering process by removing the tissues images which remain blue after counterstaining process. Then, fuzzy c-mean clustering is applied to segment the bacilli. Three colour models, RGB, HSI and C-Y are analysed to identify the colour model that perform significant segmentation performance. Finally, a 5×5 median filter and region growing was used to eliminate small regions and noises. The proposed methods have been analysed for several TB slide images under various conditions. The results indicated that fuzzy c-mean clustering using saturation component of C-Y colour model has achieved the best segmentation result with an accuracy of 99.54%.