An artificial intelligence (AI) model that calculates key indicators that critically impact the prognosis of knee surgeries more than 10 times faster than measurements by specialists has been developed by domestic researchers.
Professor Noh Doo-hyun and Research Professor Kim Sung-eun from the Department of Orthopedics at Seoul National University Hospital, along with a joint research team from the University of Minnesota in the United States and the University of Bergen in Norway, announced on the 26th that they developed a deep learning model capable of quickly and accurately measuring the tibial posterior slope angle based on over 10,000 lateral X-ray images of knee joints taken from 2009 to 2019.
The tibial posterior slope angle is the angle that indicates how much the joint surface tilts backward when viewed from the side of the knee, directly affecting the lifespan of artificial joints and surgery outcomes. A larger angle increases the risk of ligament damage and can shorten the lifespan of artificial joints.
However, due to varying imaging conditions at medical institutions, there was a problem of differing results for the same patient across different hospitals, as there was no standardized measurement method.
The AI model developed by the research team automatically recognizes six anatomical reference points of the knee bones, calculates the joint line and central axis, and derives the slope of the tibia.
The time taken for this process averages 2.6 seconds, which is more than 10 times faster than when a specialist measures it directly (an average of 26.1 seconds).
The accuracy was also found to be significantly consistent with specialist measurements. The inter-observer correlation coefficient was at least 91%, and the intra-observer correlation coefficient, which indicates measurement consistency, showed perfect agreement at 100%.
The research team analyzed knee images of 289 Norwegian patients for follow-up validation. Even then, the correlation coefficient between the AI model and specialist measurements reached 80%, confirming applicability even in patient groups with different races and environments.
Research Professor Kim Sung-eun (corresponding author) noted, "This result is a case of successfully validating medical AI technology developed in Korea across various races," and stated, "We will seek ways to expand the versatility of this model to establish it as the standard for measuring tibial posterior slope angles through subsequent research."
The results of this study were recently published in the international academic journal "Orthopaedic Journal of Sports Medicine."
Reference materials
Orthopaedic Journal of Sports Medicine (2025), DOI: https://doi.org/10.1177/23259671251333607