Kakao Brain, the artificial intelligence research and development arm of South Korea’s Kakao Corp., has made significant strides in medical AI by unveiling the Labeler Project on GitHub. This generative AI technology marks a breakthrough in medical diagnostics by extracting disease names from chest X-ray reports without directly analyzing the X-ray images. Instead, it interprets notes from medical personnel, identifying the presence of 13 thoracic diseases with an impressive 90.4% accuracy rate—substantially higher than the current competitive technologies’ accuracy of about 76%.
This development is poised to offer substantial support in interpreting chest X-ray images, an area Kakao Brain is targeting for service introduction within the year. By publishing the AI’s learning methodologies on arXiv and planning to release a test model for AI technology research benchmarking, Kakao Brain is demonstrating its commitment to advancing medical AI research and application.
In addition to the Labeler Project, Kakao Brain has previously contributed to the AI community with Honeybee, a technology that interprets images and answers text-based questions, also available on GitHub. With a focus on utilizing generative AI for educational projects and integrating it into various services, including KakaoTalk, Kakao Brain is at the forefront of leveraging AI to enhance both medical and consumer applications, underscoring its innovative approach to technology development.
