GFNN: Gaussian-Fuzzy-Neural network for diagnosis of tuberculosis using sputum smear microscopic images

Abstract

Tuberculosis (TB) is one of the infectious diseases spread by the infectious agent, Mycobacterium tuberculosis. Sputum smear microscopy is a primary tool used for the diagnosis of pulmonary TB, but it has limitations, like low sensitivity, large observation time, and so on. Hence, an automated technique is preferred for the diagnosis of TB. This work proposes the Gaussian-Fuzzy-Neural network (GFNN) by combining the Gaussian mixture model (GMM) along with the fuzzy and the neural network for the TB detection. Initially, the input sputum smear microscopic image is subjected to a color space (CS) transformation, for which the thresholding is applied to obtain the segmented result. Then, the texture feature and other features are extracted for GFNN-based classification which classifies the segments into few-bacilli, non-bacilli and overlapping bacilli. Again, the overlapping bacilli are applied to the classifier to find the number of bacilli in the overlapping bacilli. The experimentation of the proposed GFNN classifier utilizes the Zeihl-Neelson (ZN) database containing the sputum smear images. From the experimental results, it is clear that the proposed GFNN model achieved better performance with the values of 0.91379, 0.04596, and 1.55714 for the performance metrics segmentation accuracy (SA), mean squared error (MSE), and missing count (MC), respectively.