How AI aids in tackling tuberculosis

Today, Tuberculosis affects close to 10 million people per year and is one of the top causes of death too. Also, the trouble in diagnosing TB is due to the early symptoms which are similar to some of the common respiratory diseases, thereby missing the window for early treatment. While modern medicine and cost effective chest screening is one way for early detection and treatment, artificial intelligence can be a great support in the treatment of TB.

Google researchers have developed an AI tool that helps identify potential TB patients. A system of deep learning could be used for the same. Google in its blog stated that the AI-powered screening can save 80% of the cost for each positive TB case found. The AI system gives a number between 0 and 1 that would indicate a severity or risk of TB. The AI tool could detect accurately active pulmonary TB cases with false- negative and positive detection rates similar to 14 radiologists.

The model worked for a wide range and from de-identified data from nine countries which was used to train the model and tested on cases from five countries. Google’s research suggests that any clinic could start using the model from the default threshold and be assured that the model can match performance with that of a radiologist. From there on the threshold can be adjusted basis the local requirements and resources.