South Korea has long had one of the highest cancer incidence rates in the developed world, driven partly by diet, partly by an ageing population, and partly by a national culture of intensive health screening. What it lacked, until recently, was the radiological workforce to read all those scans with consistent accuracy — particularly outside Seoul and Busan.
The National AI Diagnostics Integration Act, passed in 2023 and implemented across all 312 public hospitals by March 2024, mandated that every CT, MRI, and mammography scan be processed by a certified AI diagnostic layer before human radiologist review. The system — developed by a consortium including Kakao Healthcare, Seoul National University Hospital, and Canadian firm Nines — flags anomalies, prioritises worklists by clinical urgency, and provides probability scores for 14 cancer types.
Ministry of Health data for the first full operating year shows late-stage diagnosis rates fell 47 percent nationally compared to the 2019 to 2022 baseline. In rural North Gyeongsang and Gangwon provinces, where radiologist shortages had been most acute, the reduction reached 61 percent. Five-year survival projections for newly diagnosed patients in those regions have improved by an estimated 18 percentage points as a result of earlier detection.
Radiologists initially resisted mandatory AI integration, citing concerns about deskilling and liability. A revised protocol — in which AI flags are treated as a second opinion rather than a primary read, with the human radiologist retaining diagnostic authority — has largely resolved professional disputes, and satisfaction surveys among radiologists show 74 percent now consider the tool an asset.
"The algorithm does not replace the doctor," said Dr. Park Ji-hoon of Seoul National University Hospital. "It ensures the doctor sees what matters first."