▲ (from left) Professor Kang Mo Kim of the Division of Gastroenterology and Professor Namkug Kim of the Department of Convergence Medicine

Hepatocellular carcinoma (HCC) presents a wide range of factors to consider, including the location and size of the tumor, metastasis, and residual liver function, with various treatment methods available. Therefore, it is challenging to determine the direction of treatment. To address this, a research team led by Professor Kang Mo Kim of the Division of Gastroenterology and Professor Namkug Kim of the Department of Convergence Medicine at Asan Medical Center developed an AI-based clinical decision support system (CDSS) that suggests treatment methods for individual patients and predicts survival rates.

The research team collected data from 2,685 patients diagnosed with HCC from nine domestic medical institutions, including AMC, and treated with various therapies from January 2010 to December 2012. Artificial intelligence was trained for their basic clinical information, type of initial treatment after cancer diagnosis, and post-therapy survival data divided by hospital. The research findings showed treatment prediction accuracies of 87.27% and 86.06% for internal and external datasets of AMC, respectively, with survival prediction accuracies of 91.89% and 86.48%, demonstrating high diagnostic performance.

Notably, in this study, different treatment methods were recommended to the same patient based on the characteristics of each institution, and predicted survival rates differed for each treatment. From this, there is potential for use as a ‘digital twin’ that converts real situations into virtual models to simulate various situations.

Professor Namkug Kim of the Department of Convergence Medicine stated, “This study proves the expansion of artificial intelligence beyond the diagnostic field into the treatment area. It signifies the possibility of a data-driven clinical decision support system tailored to each hospital’s characteristics for HCC patients for whom determining treatment options is challenging.

The research findings were recently published online in the ‘npj Digital Medicine.’