The improvement of automatic scanning microscope based on intelligent systems to identify Mycobacterium tuberculosis

This research yielded the conventional light microscope which could do screening and identification of mycobacterium tuberculosis automatically in sputum smear slide with Ziehl-Neelsen staining. The tool consists of electromechanical side which was assembled to move the X-Y direction of microscope desk automatically. The microscope was provided with the computer aided diagnose software to identify mycobacterium tuberculosis which it consists of image processing, segmentation, feature extraction, and classification methods. The most important in software development in this research is the segmentation process. It could influence the accuracy of mycobacterium tuberculosis observation. We tried some methods on segmentation in which k-Nearest Neighbors gave the better accuracy than other methods. But k-Nearest Neighbors gave the long computational times. After segmentation process, we did classification to the reddish object using neural network with feature extraction based on geometrical shape to become neural network input. The neural network gave very good accuracy 100% on classification of mycobacterium tuberculosis and not mycobacterium tuberculosis.