An automatic intelligent system for diagnosis and confirmation of Johne's disease

Abstract

Johne's disease is one of the most widespread bacterial diseases of domestic animals. It causes yearly losses of billions of dollars worldwide. In this paper an automatic intelligent computer-aided system is proposed for the diagnosis of Johne's disease, the system uses image analysis and computer vision techniques to extract features from two different microscopic images, then those features are classified using neural networks and K-nearest neighbour K-NN techniques to diagnose Johne's disease. The proposed system employs histopathological examination to extract 192 different texture features. The features are then reduced into only 8 features and classified using artificial neural networks ANN. The acid fast stain test is used to confirm the positive cases. The construction and testing of both models are carried out using a total of 294 microscopic images, 194 images for the histopathological examination test which produces an overall accuracy of 98.33%. The other 100 images are used for the acid fast stain test, and it achieves an accuracy of 96.97%.