A woman on a medical table gets ready for an MRI examShare on Pinterest
Pancreatic cancer can be difficult to detect early with current technologies. FS Productions/Getty Images
  • Pancreatic cancer is deadly if not detected in its early stages.
  • Researchers say a new tool that uses artificial intelligence could help diagnose pancreatic cancer earlier.
  • They plan to conduct further studies that combine this tool with the emerging technology of liquid biopsies.

Pancreatic cancer has been in the news a lot in recent years.

A number of well-known figures have died from this deadly cancer, including Jeopardy host Alex Trebek, singer Aretha Franklin, and Supreme Court Justice Ruth Bader Ginsburg.

The 5-year survival rate for people with pancreatic cancer in the United States is 11%, according to the American Society of Clinical Oncology.

If caught early, pancreatic cancer is treatable. Experts say that early detection is the best way to improve the chances of survival as the prognosis worsens significantly once the tumor grows beyond 2 centimeters.

Presently, there are no validated screening tests that can reliably identify early-stage pancreatic cancer in people who have no symptoms.

Regular CT scans don’t always detect the tumor.

However, a study from Taiwan published today about an artificial intelligence (AI) program to more accurately detect pancreatic cancer is offering new hope.

The new technique deploys a deep-learning program that may be able to accurately detect pancreatic cancer at early stages.

The study’s senior author, Weichung Wang, Ph.D., a professor at National Taiwan University and director of the university’s MeDA Lab, told Healthline that this technology could be a breakthrough.

During the research, Dr. Wang said, something called the clinically applicable computer-aided detection tool identified the pancreas automatically.

“This is an important advance because the pancreas borders multiple organs and structures and varies widely in shape and size,” said Wang, who with his colleagues developed the tool with 546 people with pancreatic cancer and 733 control participants.

Researchers said the tool achieved 90% sensitivity and 96% specificity in the internal test set. The sensitivity for detecting pancreatic cancers less than 2 centimeters was 75%.

Validation followed with 1,473 individual CT exams from institutions throughout Taiwan.

The scientists, whose work was supported by the Ministry of Science and Technology, Ministry of Science and Technology All Vista Healthcare Subcenter, and National Center for Theoretical Sciences Mathematics Division, are planning further studies on more diverse populations.

Hong-Wen Deng, Ph.D., a professor and the director of the Tulane Center of Biomedical Informatics and Genomics at Tulane University School of Medicine in New Orleans, said the potential for AI to advance cancer care is vast.

“AI has great potential for the diagnosis and early detection and prognosis of cancer,” he told Healthline. “In our AI work in colorectal cancer, we compared our model with pathologists, six different models, and ours was better in terms of performance.”

Wang said that in subsequent studies he hopes to deploy liquid biopsies, which are blood tests that identify biomarkers in the blood that can identify the presence of cancer.

Wang said liquid biopsy and AI complement each other.

“For early pancreatic cancers, most of which have no or vague symptoms, a sensitive and specific blood-based exam such as liquid biopsy is needed for screening the populations at risk,” he said.

“That is how liquid biopsy complements AI. Subjecting all those individuals to CT scan is cost-prohibitive and results in unnecessary radiation for many,” Wang added.

For individuals who are suspected of having pancreatic cancer, the next step, Wang said, is to undergo a CT scan, preferably with AI analysis to enhance the ability of detection on the scan.

“It is not possible to determine a subject as having pancreatic cancer based on only liquid biopsy without being able to confirm the presence of tumor on imaging,” he said.

Meanwhile, another screening technology using AI has flagged more than 95 percent of stage 1 pancreatic cancers, according to a pilot study published earlier this year in the journal Nature Communications Medicine.

Dr. Scott Lippman, director of the UC San Diego Moores Cancer Center and co-senior author of that study, said in a press statement that something called high-conductance di-electrophoresis detected 95 percent of early pancreatic cancers.

The technology detects extracellular vesicles, which contain tumor proteins that are released into circulation by cancer cells as part of a poorly understood intercellular communication network.

AI-enabled protein marker analysis is then used to predict the likelihood of malignancy.

This method detected 95% of stage 1 pancreatic cancers, with more than 99% specificity.

If the study results are validated, Lippman said, “we can greatly reduce the mortality from this disease which will soon become the second-leading cause of cancer mortality in the U.S.”