Lunit "Ranked No.1 in Tuberculosis Detection Ability Among 12 Global AI Products"

Medical AI company Lunit announced on the 6th that its Lunit Insight CXR showed the best performance among 12 global AI solutions in detecting tuberculosis patients, based on a direct comparison study of detection performance.

Lunit's AI chest X-ray image analysis solution 'Lunit INSIGHT CXR'
[Photo by Lunit]

According to a related study by Professor Zijin Qin's research team at Heidelberg University Hospital in Germany, published recently in the medical journal The Lancet Digital Health, the tuberculosis detection ability of Lunit Insight CXR was the highest among all AI products analyzed in the study, with an area under the curve (AUC) value of 0.902, a representative performance evaluation metric for AI models. The closer the AUC is to 1, the better the product's performance is evaluated, and models with an AUC above 0.8 are classified as high-performance models.

This study was conducted by an independent research team unrelated to AI solution developers, comparing the performance of 12 AI products using the same data. Chest X-ray data from 774 individuals used in a tuberculosis prevalence survey conducted in South Africa from August 2017 to July 2019 over two years were utilized.

In particular, Lunit Insight CXR was noted to be the closest to the detection performance targets proposed by the World Health Organization (WHO) for AI solutions. WHO recommends the use of computer-aided diagnostic devices (CAD) in the classification process of tuberculosis patients aged 15 and older. The target performance is set at 90% sensitivity and 70% specificity. Sensitivity refers to the probability of correctly diagnosing actual tuberculosis patients as positive, while specificity refers to the probability of correctly diagnosing healthy individuals without tuberculosis as negative.

When the research team set the sensitivity threshold of Lunit AI at 90%, the specificity recorded was 67.7%, the closest result to the WHO target among the 12 products. Conversely, when the specificity threshold was set at 70%, the sensitivity was 89.5%, also the closest to the target index among the 12 products.

The company explained, "These research results show the potential for Lunit Insight CXR to be used as an effective tool for tuberculosis screening not only in advanced global markets but also in developing countries." They added, "Tuberculosis is a representative infectious disease prevalent in developing countries, where there is a high demand to introduce AI solutions in the tuberculosis screening process due to limited medical resources."

Seobum Seok, CEO of Lunit, said, "This study is a major milestone objectively proving the comparative superiority of Lunit AI's performance and is significant in that it was published in a world-renowned journal. Showing performance close to the WHO tuberculosis classification target demonstrates Lunit's competitiveness in the global healthcare market, and based on this, we will actively expand the market to developing countries and others."

This content was produced with the assistance of AI translation services.