Expected to Complement Limitations of Conventional Imaging Tests
Easily Assesses Risk Through Simple Blood Collection Without Special Equipment

The research team at Severance Hospital announced on March 23 that they have developed an artificial intelligence (AI) model capable of predicting the risk of early liver metastasis in patients with pancreatic cancer using only blood test data.


This model, developed by a joint research team led by Professor Heeseung Lee and resident Yookyung Go from the Department of Gastroenterology, was published in the latest issue of the international journal 'BMC Cancer'.

Professor Heeseung Lee (left) and resident Yeokyung Ko from the Department of Gastroenterology at Severance Hospital participated in the study. Severance Hospital

Professor Heeseung Lee (left) and resident Yeokyung Ko from the Department of Gastroenterology at Severance Hospital participated in the study. Severance Hospital

View original image

A significant number of pancreatic cancer patients are diagnosed after the cancer has already metastasized to other organs, and liver metastasis is a particularly important factor in determining surgical options and prognosis. However, existing imaging tests such as CT or MRI sometimes fail to detect small liver metastases.


Professor Heeseung Lee’s team developed an AI model called 'LiMPC' to assess the risk of liver metastasis in pancreatic cancer patients. The main feature of this model is that it utilizes routine blood test data collected at the time of pancreatic cancer diagnosis, allowing risk assessment without the need for additional tests or equipment.


The research team analyzed blood test data taken at diagnosis from 2,657 pancreatic cancer patients at Severance Hospital to develop the AI model. They then conducted external validation on 272 patients at five medical institutions in Korea, including Gangnam Severance Hospital and Yongin Severance Hospital. As a result, the AI model demonstrated a certain level of predictive performance in distinguishing the risk of early liver metastasis. In external validation, the sensitivity was 0.81, meaning that the model correctly identified about 81% of patients with actual liver metastasis as high-risk. This indicates a relatively high detection rate for patients with liver metastasis.


The negative predictive value was 0.87, indicating that among patients classified as low-risk, 87% did not actually have liver metastasis. In other words, if the model assessed the risk of liver metastasis as low, there was a high probability that metastasis was indeed absent. Through these results, the research team suggested that this tool could be used as an auxiliary method for evaluating liver metastasis risk, especially when it is difficult to confirm with conventional imaging tests.


The researchers expect that, since this AI model does not require any special equipment, it can be used not only in large hospitals but also in regions with limited medical resources. They have also completed the development of an online calculator to enable medical professionals to easily use the model and plan to release it soon.


Professor Heeseung Lee stated, "It is often difficult to accurately determine metastasis in pancreatic cancer using only initial imaging tests. This study is significant in that it demonstrates the potential to predict the likelihood of liver metastasis through standard blood tests routinely performed in hospitals, thereby supplementing standard imaging diagnostic tests."



The first author of the paper, resident Yookyung Go, said, "This AI model can be used without expensive diagnostic equipment, making it highly accessible in clinical settings. I hope to further develop the model through additional research so that it can be even more useful in clinical practice in the future."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing