▲ (from the left) Professor Namkug Kim and Specialist Seung Min Ryu

A team led by Professor Namkug Kim of the Department of Convergence Medicine and Specialist Seung Min Ryu of the Department of Orthopedic Surgery has published research findings regarding the development of an artificial intelligence(AI) model that accurately diagnoses flatfeet based on deep learning technology utilizing X-ray images.

The team developed an AI model using 1,200 X-ray images of a weight-bearing lateral foot, collected from 600 adult males. The results displayed a diagnostic baseline for flatfeet with an average error of 1.37°, which is narrower than the 2.35° margin of error observed by orthopedic specialists.

Diagnosing flat feet involves measuring the angle between the bones of the foot through a weight-bearing lateral foot X-ray. However, there were limitations due to discrepancies in the angle measurement process among medical professionals, known as ‘inter-observer reliability.’ To overcome these limitations, the research team trained the AI model using 1,050 X-ray images captured between June and December 2020.

Additionally, the X-ray images were used to train the AI to identify 22 landmarks utilized for measuring flatfeet. Subsequent validation of the developed AI model using 150 X-ray images resulted in a diagnostic baseline with an average error of 1.37°.

Following the guidance of AI, when interpreting the 150 X-ray images, orthopedic specialists reduced their average error from 2.35° to 1.55°, accompanied by a reduction in measurement time from 195 to 135 minutes. General practitioners also experienced a decrease in error from 1.99° to 1.56°, along with a reduction in time from 205 to 155 minutes.

"The significance of this study is that AI can be utilized to solve the problem of inter-observer reliability in the flatfeet diagnostic process," said Professor Namkug Kim.

These research findings have recently been published in the 'European Radiology(IF=7.034).'