86% accuracy in predicting complications in AI development by screening high-risk groups with severe complications after the cure of chronic hepatitis C

Nov 05, 2024

A research team led by Kim Seung-up and Lee Hye-won of Severance Hospital's gastroenterology department has developed an AI that can select high-risk patients who may show liver-related complications after the cure of chronic hepatitis C.

The findings are published in the journal Journal of the European Liver Society 'Journal of Hepaticology (IF 26.8)'.

Chronic hepatitis C is not transmitted only by routine contact, but is infected by contaminated blood such as needle stabbing, acupuncture, and tattoos. It can be fixed with 'Direct-acting antiviral agents (DAA)' that most people eat.



Hepatitis C tends to ignore the risk of developing hepatocellular carcinoma if there is no cirrhosis after cure. However, even without cirrhosis, some cured patients may show serious complications such as hepatocellular carcinoma, so caution is needed. It is very important to predict the risk because it is difficult to recognize the risk in the absence of cirrhosis diagnosed by abdominal ultrasound.

The research team developed an AI that can select high-risk groups with poor prognosis, such as hepatocellular carcinoma, even though they do not have cirrhosis among those cured of hepatitis C, and verified its performance.



First, AI was created using data from 913 hepatitis C patients without cirrhosis who were treated with antiviral drugs at Severance Hospital. AI predicts severe complications by using the results of six blood tests that can check liver elasticity values, age, gender, platelet count, and bilirubin levels.

We then tested the accuracy of predicting liver-related complications such as hepatocellular carcinoma, decompensated cirrhosis, liver transplantation, and death using data from 1264 patients in Hong Kong and France. Closer to 1, the higher the predictive performance, the 'C-index' value recorded 0.86, boasting an 86% predictive accuracy of AI. When only hepatocellular carcinoma was separately identified, the level rose to 0.87.



In fact, the incidence of liver-related complications within five years of high-risk patients with AI calculated values of 0.7 or higher was 3.2 to 6.2%, up to 30 times higher than the incidence (0.2 to 0.5%) of low-risk groups below 0.7.

Professor Kim Seung-up said "Chronic hepatitis C can be easily treated with antiviral drugs, but serious complications such as hepatocellular carcinoma can be experienced even after a cure."Using the AI developed in this study, we can select high-risk groups that need regular follow-up even after a complete cure" he said.

86% accuracy in predicting complications in AI development by screening high-risk groups with severe complications after the cure of chronic hepatitis C
Professors Kim Seung-up (left) and Lee Hye-won





bellho@sportschosun.com