A breast cancer patient. [AI-generated image]
A breast cancer patient. [AI-generated image]

South Korea has launched a full-scale effort to develop an AI-powered ultrasound robot capable of detecting breast cancer early — the most common cancer among women in the country.

The Daegu Gyeongbuk Institute of Science and Technology (DGIST) announced Wednesday it had been selected for a new project under the "Real-World Active Agentic AI Technology Development" initiative, backed by the Ministry of Science and ICT and the Institute for Information and Communications Technology Planning and Evaluation (IITP). DGIST will work alongside lead agency Medicalpark Co. and Yonsei University's Industry-Academic Cooperation Foundation to commercialize an "intelligent breast ultrasound scanning solution" using autonomous robotic diagnostic technology.

The project runs through 2028 — about two and a half years — with a total budget of 10.3 billion won ($7.4 million). The Ministry of Science and ICT and IITP held a project launch briefing in Seoul on July 30, formally kicking off the government's "Agentic AI Initiative." Within the project, DGIST professors Yoo Jae-seok and Park Kyung-seo of the Department of Robotics and Mechatronics Engineering will lead development of the AI agent system at the core of the platform.

Breast cancer is the leading cancer among women in South Korea, making early detection critical. Conventional breast ultrasound examinations have long been limited by their dependence on the operator's skill level, with image quality and diagnostic accuracy varying widely between examiners. Existing robotic ultrasound devices have also struggled to adapt to individual patients' body types or real-time conditions, moving only along pre-programmed paths.

The joint research team will develop a system called US-AXBOT — a three-dimensional automated breast ultrasound robot powered by "agentic AI" that perceives, judges, acts and learns autonomously. Eight AI agents will coordinate the entire examination process without human direction, and the system's closed-loop architecture — in which the AI's image assessments feed directly into the robot's movements — is designed to deliver autonomous scanning at the level of an experienced specialist.

DGIST Department of Robotics and Mechatronics Engineering professors Yoo Jae-seok (left) and Park Kyung-seo. [Provided by DGIST]
DGIST Department of Robotics and Mechatronics Engineering professors Yoo Jae-seok (left) and Park Kyung-seo. [Provided by DGIST]

The DGIST research team will take sole responsibility for developing five of the system's core AI agents.

Safety is the top design priority, given that the robot makes direct physical contact with patients. DGIST plans to combine precise control of interaction forces between the robot and the human body with real-time hazard detection and autonomous avoidance, ensuring full protection for both patients and operators.

Medicalpark will handle robot system integration, three-dimensional diagnostic AI development and regulatory approval, while Yonsei University will focus on clinical workflow development and efficacy evaluation. By the project's end, the team aims to bring the system to Technology Readiness Level 7 — one step short of full commercialization and ready for real clinical environments — while laying the groundwork for manufacturing approval from South Korea's Ministry of Food and Drug Safety and market entry into the United States through the FDA.

"This project goes beyond simple automation where a robot repeats a fixed path — it represents a significant advance in which multiple AI agents collaborate in real time to complete an examination," said Yoo Jae-seok, a DGIST professor of robotics and mechatronics engineering. "We aim to deliver standardized ultrasound examinations that are independent of the examiner's skill level, raise the early detection rate for breast cancer, and strengthen the global competitiveness of domestically developed medical devices."


nbgkoo@heraldcorp.com