An efficient approach to sputum image segmentation using improved fuzzy local information C means clustering algorithm for tuberculosis diagnosis

Tuberculosis is one of an infectious threatening disease affected one third of world's population but its fatality rate can controlled by diagnosing and treating at an early stage itself. Ziehl-Neelsen stained sputum smear microscopy is the most popular diagnosis method in developing countries. The stained images do not always have the sufficient contrast and hence the clinicians feel hard to inspect bacteria on it. Our Proposed algorithm automatically detects and segments the tuberculosis bacteria from background using improved fuzzy local information C means (IFLICM) clustering algorithm that overcomes the drawbacks of existing fast-generalized fuzzy C means clustering algorithm and improves the performance of clustering. Experimental result shows that IFLICM algorithm is efficient, robust to noise and a feasible alternative while comparing with traditional fuzzy based algorithms by giving an average segmentation accuracy of 96.05 on sputum image dataset.