Classification and Counting of Mycobacterium Tuberculosis from Sputum Microscopic Image using Fuzzy Logic

Abstract

The diagnosis of tuberculosis (TB) is done by detecting and counting the number of mycobacterium tuberculosis in a sputum examination done manually using a microscope. It is considered ineffective because it requires a long time and different diagnostic results. To overcome this problem, this paper implements digital image processing. There are 5 processes used on the system. Preprocessing with the RGB to HSV method is used to clarify the color of the image. Segmentation to separate objects from background images using thresholding. Feature extraction to get the value of area, perimeter, and level of roundness of the object. Classification uses fuzzy logic to classify mycobacterium tuberculosis based on features. The next is the process of counting mycobacterium tuberculosis. And the last is the process of classify into IUATLD scale based on the number of mycobacterium tuberculosis. From the results of tests conducted on 15 data, the system show that the level of accuracy, precision, sensitivity and specificity of system in calculate mycobacterium tuberculosis is 89%, 90%, 91.66% and 78.88% respectively. And also level of sensitivity, specificity and accuracy of system in classifying the level of infection is 100%, 80 % and 93% respectively. This system was tested on a microscopic sputum image database of RSUD Dr. Soetomo from a different patient.