Kakao Brain releases medical GenAI tech for free

Kakao Brain, an artificial intelligence (AI) R&D subsidiary of South Korea's Kakao Corp., announced on Tuesday that it has uploaded the Labeler Project, a generative AI technology that extracts disease names from chest X-ray reports to the open-source platform GitHub.

The AI used in this technology does not analyze X-ray images directly.

Instead, it collects and analyzes notes written by medical staff while observing various images.

Through this, it determines the presence or absence of 13 thoracic diseases, including fractures, pleural lesions, and pneumothorax.

The interpretation accuracy is 90.4%, which is higher than the accuracy (about 76%) of existing competitive technologies, according to Kakao Brain.

Kakao Brain aims to introduce a service that assists in interpreting chest X-ray images this year.

The AI's learning process used for report analysis has been published on the academic website arXiv.

Kakao Brain also plans to release the test model of this research result to be used as a benchmark for AI technology research, according to Kim Il Doo, the CEO of Kakao Brain.

Last month, Kakao Brain also released its own AI technology Honeybee on GitHub.

Honeybee is a technology that analyzes images uploaded by users and answers text questions.

Kakao Brain is considering using this technology for educational projects.

Kakao plans to sequentially launch various services by incorporating generative AI technology into KakaoTalk.

Write to Ju-Hyun Lee at 2juhyun@hankyung.com