CAD4TB is a computer-aided detection system whose main goal is to detect tuberculosis-related abnormalities in posterior anterior chest X-rays. CAD4TB takes a single chest X-ray as its input, in the form of a DICOM image, and produces several outputs: a quality assessment of the input image, a heat map highlighting possible abnormal areas, and a score between 0 and 100 indicating the likelihood of the X-ray being abnormal and the subject on the X-ray being affected by tuberculosis. Users can take these outputs into account in their clinical workflow. They can decide, for example, that a new image should be acquired, in case the quality assessment indicates suboptimal image quality, or they can decide that the subject should undergo further testing, in case the heat map displays suspicious regions that are verified by a human operator or the score is above a certain threshold.
CAD4TB has been developed following the principles of supervised machine learning: in the process of computing its score, it compares regions in the input image with regions extracted from normal and abnormal images previously processed by the system, which constitute the so-called training set. One of the conditions for proper supervised learning is that this training set should be representative of the test data; otherwise, results may not be reliable. To fulfill this condition, and thus make CAD4TB applicable to diverse scenarios, the system has been trained with data from several populations and several X-ray devices.