APPLICATION OF ARTIFICIAL NEURAL NETWORK FOR DIAGNOSIS OF TUBERCULOUS MYCOBACTERIUM IN INFECTED LUNGS USING IMAGE PROCESSING TECHNIQUES
Publication Date : 01/08/2017
The high level of manual work flow in the diagnostic procedure for Mycobacterium Tuberculous (TB) introduces varying degrees of subjectivity at different stages in the process. This affects the final accuracy and increases diagnosis time. This paper presents an automatic method of segmenting TB bacilli using cascade threshold filters. A multi layer artificial neural network (ANN) with scaled conjugate gradient descent algorithm was used to classify the presence or absence of TB bacilli in the processed images. Results of the ANN classifier gave a MSE of 0.025 and accuracy of 94.7%. These results suggest that the proposed procedure can help detect the presence or absence of TB bacilli in Ziehl-Neelsen (ZN)-stained sputum smear samples with high accuracy.
No. of Downloads :