Image preprocessing to improve Acid-Fast Bacilli (AFB) detection in smear microscopy to diagnose pulmonary tuberculosis

Pulmonary tuberculosis (TB) is a highly infectious disease. TB is curable if it is diagnosed opportunely. Worldwide, the most used diagnostic method is the analysis of smear microscopy, which consists in, using a microscope, detecting and counting the bacilli in the smear. The automatic detection of pulmonary tuberculosis usually involves processing and analyzing digital images related to smear microscopy. The main problem in this analysis is the color variation and low contrast in the images. This paper presents a quick and easy method to minimize these variations by using image preprocessing, changing the RGB color space to the HSV space, analyzing and modifying the original images characteristics to standardize them. The results are validated by using a further segmentation step of the images using Artificial Neural Networks (ANNs) and comparing the results obtained with and without the image preprocessing method.