Automated Detection of Tuberculosis on Chest X-Rays Using Artificial Intelligence.
Authors
Affiliations (2)
Affiliations (2)
- Mathematics and Informatics laboratory (LAMI), Ouagadougou.
- Virtual University (UV) and Public Health Laboratory (LASAP), Ouagadougou.
Abstract
Detecting tuberculosis on X-rays remains complex and challenging for clinicians. In Burkina Faso, the scarcity of radiologists and their heavy workload increase the risk of misinterpretation, particularly in areas with high demand. To address this constraint, we have designed a comprehensive solution based on artificial intelligence, ranging from data analysis to the training of advanced deep learning models, enabling the automatic identification of tuberculosis abnormalities from chest X-rays. In this article, we present our three-step approach for tuberculosis detection, directly inspired by the clinical reasoning of the radiologist. First, we classify the images into those that are normal and those that are not. Next, we segment the images to keep only the pulmonary area. Finally, in this area, we perform a detection of tuberculosis-related anomalies.