Enhanced fuzzy Gaussian networks for sputum image based Mycobacterium detection

ABSTRACT The primary step involved in the diagnosis of Mycobacterium tuberculosis is the Sputum Smear Microscopy. The limitations of manual detection can be avoided with an automated technique which is carried out in this study. This paper uses the Enhanced Fuzzy Gaussian Networks (EFGN) for the detection of TB, by integrating the Gaussian model with the fuzzy and the neural network. In EFGN, the classified output obtained from the enhanced fuzzy and neural network is combined together depending on the Gaussian mixture model. The sputum smear microscopic image acts as the input, to which the process of thresholding is imposed to get the segmented result. Local Gradient Pattern (LGP), length, density, area, and histogram features are utilized to classify and count the bacilli objects using EFGN classifier. The performance is estimated based on the factors, such as Segmentation Accuracy (SA), and Mean Squared Error (MSE).