An Electronic Equipment for Automatic Detection of Mycobacteria in MGIT Cultures Based on Digital Image Processing

In this work we propose an automated electronic equipment oriented to the detection of mycobacteria by means of the analysis of fluorescence in MGIT tube cultures using digital image processing. The equipment aims to improve detection efficiency, since this procedure is mostly performed by visual inspection and has low reliability due to subjectivity. Part of the proposed system acts as an incubator for the growth of mycobacteria, which will maintain a temperature of 37°C±1°C. It also has a mobile camera for the acquisition of images of each tube in the enclosure, as well as an algorithm implemented in a single-board computer. The algorithm detects if a tube emits fluorescence, which is an indicator of the presence of mycobacteria in the broth. The steps of the algorithm involve transforming the RGB images to the HSV color model, filtering and classification by K-Nearest Neighbor. Equipment validation was made from a set of samples processed by a medically certified and approved professional electronic equipment. The results were compared with those obtained by the BD Bactec 960 MGIT equipment achieving a Cohen Kappa coefficient of 0.9355 in the evaluation of 40 samples selected by the specialist.