Mycobacterium Tuberculosis Bacilli Cells Recognition Using SVM

The proper and rapid diagnosis of pulmonary tuberculosis is essential for treatment as well as control. A technique commonly used consist of analyzing sputum images for detecting bacilli. Manual method is time consuming, tedious & it may require skilled persons for proper, accurate & specific results, sometime it may confuse with some stain residue & non tuberculosis bacilli. Due to this reason the need of atomization is required for exact identification of tuberculosis, reducing time and increasing specificity of test. We present an object (M.TB) recognition method using Support Vector Machine. The object i.e. M.TB bacilli cells are extracted using color segmentation thresholding & boundary detection method from Ziehl-Neelsen stained sputum smears images. The various extracted shapes contains some of M. TB bacilli cells, non TB cells, stain residue etc, for proper recognition the moment invariant and eccentricity feature of these shapes re extracted on this basis SVM classifier is applied for recognition of M.TB bacilli cells. The accuracy obtains using this technique is 85% for M.TB cells and specificity obtains 98.81%.