Advances in artificial intelligence have enabled Korean medical researchers to develop a technology that more accurately predicts the prognosis of patients undergoing pancreatic cancer surgery.
Using AI to analyze tumor-infiltrating lymphocyte (TIL) density, the new approach offers improved accuracy in forecasting postoperative outcomes for pancreatic cancer patients.
Samsung Medical Center said Monday that a team, led by Professor Park Joo-Kyung of the Department of Gastroenterology at Pancreaticobiliary Cancer Center, Professor Han In-woong of the Department of Hepatobiliary Surgery, and Professor Jang Kee-taek of the Department of Pathology, conducted a study showing that AI-based spatial TIL density analysis is closely related to the survival rate of pancreatic cancer patients.
The study, published in the world-renowned journal JAMA Surgery, was accompanied by a special commentary from JAMA Surgery titled “The Future of Oncology is Now.” The research is analyzed as the best explanation for why people with the same cancer respond differently to treatment.
TILs are considered a prognostic marker for prognosis after cancer treatment based on characteristics that reflect the immune response to the tumor. However, it is time-consuming for medical professionals to manually measure TIL density, and inter-observer variability has limited its use in actual medical practice.
The research team addressed this problem by utilizing Lunit SCOPE IO, an AI-based immunophenotyping platform, to quantify TIL density in the tumor microenvironment, compartmentalize tumor tissue and surrounding stroma, and classify immune phenotypes.
In a retrospective cohort analysis of 304 patients undergoing pancreatic cancer surgery, the researchers found that assessing tumor-infiltrating lymphocyte density and distribution in the surgical tissues of pancreatic cancer patients can more accurately predict cancer recurrence and survival outcomes.
“AI-based tumor-infiltrating lymphocyte analysis is attracting attention as a technology that can be practically applied in clinical practice due to its high reproducibility and scalability,” the researchers said. “Immune microenvironment analysis can complement existing pathological staging systems and serve as a promising biomarker to accurately predict patient prognosis.”
When patients were stratified by immunophenotype using tumor microenvironment analysis, the study found that the median survival time was 35.11 months for patients with an “immune-inflamed phenotype,” which is characterized by an abundance of tumor-infiltrating lymphocytes.
This was nearly three times longer than the median survival of 11.6 months for patients with the immune-desert phenotype. The median progression-free survival for pancreatic cancer was also twice as long for the immune-active phenotype, at 14.63 months, compared to 6.57 months for the immune-desert phenotype.
“Higher tumor-infiltrating lymphocyte density was associated with significantly better survival outcomes,” the researchers explained.
In this paper, the researchers report that in some cases, survival was reversed based on immunophenotype. While stage 1 pancreatic cancer is generally considered to have a better prognosis than stage 2, among stage 2 pancreatic cancer patients, immune-active patients had a better prognosis than non-immune-active patients with stage 1 pancreatic cancer.
“The era has begun where artificial intelligence can provide important information to guide the direction of cancer treatment,” Professor Park said. “This study shows that AI-based immunophenotyping will play a key role in precise prognosis prediction and personalized treatment strategies for pancreatic cancer patients in the future.”