BacillusNet: An automated approach using RetinaNet for segmentation of pulmonary Tuberculosis bacillus

Abstract

Tuberculosis is an infectious disease transmitted by Mycobacterium tuberculosis, being the leading cause of death from infection. The sputum bacilloscopy method is the technique for detecting the bacillus that is currently most used, not only in the search for infectious cases but also as a thermometer to check the effectiveness of the treatment. In this context, computational techniques have been developed to help the specialist for a better diagnosis. In this work, we promote a methodology for automated detection of the bacillus using RetinaNet. A set of the 928 images was used for evaluating this method. The results were promising, achieving an accuracy of 67.1%, recall of 86. 56%, and an F-score of 75.61%. Finally, we believe that our method is capable of acting in the diagnosis of tuberculosis.