Korean researchers have developed an artificial intelligence (AI) model that can predict recurrence and survival rates in patients with gallbladder cancer, one of the most intractable cancers along with pancreatic cancer. As predicting different prognoses for each patient is the first step toward implementing personalized precision treatment, the development is expected to serve as a stepping stone to improving biliary tract cancer survival rates.
Samsung Medical Center announced on the 22nd that a research team comprising professors Park Joo-kyung, Lee Kyu-taek, and Choi Young-hoon of the Department of Gastroenterology, Professor Kim Hong-beom of the Department of Hepatobiliary-Pancreatic Surgery, and Dr. Kim Hye-min of the Intractable Cancer Early Diagnosis Team at the Future Medicine Research Institute, developed and validated technology that analyzes the tumor microenvironment (TME) of gallbladder cancer patients and predicts prognosis using AI-based spatial analysis technology.
The gallbladder, commonly referred to as ssulgae in Korean, serves as a storage reservoir for bile produced by the liver. Even when cancer develops in the gallbladder, there are no clear symptoms until it has progressed considerably. Nonspecific symptoms such as weight loss, fatigue, loss of appetite, nausea, vomiting, and upper abdominal pain rarely appear, so the disease is often discovered only after it has advanced. According to the Korea Central Cancer Registry at the National Cancer Center, the five-year relative survival rate for gallbladder cancer patients is 29%, the second lowest among major solid cancers after pancreatic cancer (17%). Statistically, this means more than seven out of every 10 gallbladder cancer patients die within five years of diagnosis.
The research team analyzed an external validation cohort of 41 patients based on data from 225 gallbladder cancer surgery patients, then quantified key indicators of the tumor microenvironment—including the density of tumor-infiltrating lymphocytes (TIL) around cancer cells, the number of tertiary lymphoid structures (TLS), and fibroblast density—and incorporated them into the AI model. The team identified three key factors for predicting prognosis in gallbladder cancer patients: low TIL density within the tumor microenvironment, a small number of TLS, and high fibroblast density. This is because reports show that as these risk factors increase, both overall survival (OS) and disease-free survival (DFS) periods of gallbladder cancer patients shorten sharply. OS refers to the period from the start of treatment until the cancer patient's death, while DFS refers to the period a cancer patient survives without cancer recurrence after treatment.
As a result, the patient group with none of the three risk factors had 87% and 80% lower risks of recurrence and death, respectively, compared with the group that had all three. The team's analysis indicated that patients with more risk factors tended to have worse outcomes, including recurrence and death.
"Gallbladder cancer has a particularly poor prognosis among biliary tract cancers and is difficult to predict survival rates for," Professor Kim Hong-beom said. "A path has now opened to precisely analyze the prognosis of gallbladder cancer using AI technology."
"This study confirmed the possibility that AI can deeply analyze the biological characteristics of cancer to predict patient prognosis," Professor Park Joo-kyung said. "It will be of great help in providing customized treatment optimized for each individual patient after gallbladder cancer surgery."
The findings were published in a recent issue of the International Journal of Surgery.
