Dr. Eugene Koay, Associate Professor in the Department of Gastrointestinal Radiation Oncology at The University of Texas MD Anderson Cancer Center, is the lead researcher on an innovative study funded by Pancreatic Cancer Canada that is harnessing artificial intelligence for the early detection of pancreatic cancer.

Early detection seems to be a buzz word these days when it comes to cancer. Why is early detection so important for pancreatic cancer and what makes this research/approach so unique?

The vast majority of patients who are diagnosed with pancreatic cancer are diagnosed at late stages that are incurable because the disease has spread (or metastasized) beyond the pancreas. It is known that most patients have vague symptoms or signs of the cancer months to years prior to diagnosis when the disease is still confined to the pancreas and is curable. Identifying these patients is the focus of biomarker research. Our unique research project aims to validate promising biomarkers and combine these biomarkers with the subsequent step of using advanced imaging and artificial intelligence algorithms to quantify the risk of the patient to develop the disease and, when appropriate, identify small abnormalities in the pancreas that may become cancer.

Why is there no general population screening right now for pancreatic cancer? Could this study help provide insights towards that? 

The screening for pancreatic cancer is currently focused only on patients with a higher risk of developing pancreatic cancer than the general population because the disease is not common. Although screening with highly sensitive and highly specific biomarkers could identify many patients who may have the disease, it would also give false positives in millions of people if deployed for the general population. Because the workup for pancreatic cancer is invasive and carries significant risk with sticking a needle into bowel and the pancreas, there is significant risk of harm if a person has a false positive test.

Our approach of using biomarkers to quantify risk of developing the disease to enrich the patients and further refining that risk or detecting the disease with advanced imaging and artificial intelligence aims to improve this risk/benefit ratio of screening. We want to make it more favorable to patients to be more likely to benefit from the workup of the disease over the risk of harm from the invasive procedures needed to arrive at a diagnosis.

What are the key goals for this study?

We have immediate and long-term goals. One immediate goal is to validate promising blood biomarkers in a large study of several thousand patients who had blood collected due to the development of new onset diabetes, which has been shown to be a risk factor for developing pancreatic cancer in the near future for some patients.

While we validate these blood-based biomarkers, we will also validate promising image processing and artificial intelligence algorithms for early detection in nearly 2,000 patients who had diagnostic imaging 6 months to 5 years prior to the diagnosis of pancreatic cancer. This is a unique data source for validation of our computer algorithms because it provides the opportunity to look at the pancreas and other tissues in the time period when the cancer probably was present but not detected visually by radiologists using conventional techniques. While we pursue our immediate goals with blood and imaging approaches to early detection, we will be working on the next generation of biomarkers and deep learning techniques that we believe will be more powerful than what we have now. We will demonstrate proof of concept with these next generation techniques in a separate group of patients that we recruit during the study period.

What are the possible implications/outcomes once you have reached those goals?
Specifically, if we can fund this work faster, what will that achieve?

The implications of our immediate goals are that we will have well validated blood and imaging methods to quantify risk and detect pancreatic cancer. Achieving those milestones for blood and imaging in the near term would justify a prospective trial to fully demonstrate the value of our current technologies. If the prospective trial supports the use of our approach, it would lead to integration of the technologies into the clinic for patients. Typically, this process takes many years to finish, but additional funding would enable us to work with many other high-volume centers that would accelerate the adoption of our approaches and get into the clinic faster. Similarly, with our next generation blood and imaging techniques, we would have the ability to demonstrate proof of principle and bring these to patients sooner.

What implications could this study have for other harder to detect cancers, or cancer in general?

While we are currently focused on pancreatic cancer for this project, we think that our blood and imaging techniques can be deployed more broadly for multi-cancer detection, following a similar approach of enrichment with the blood-based biomarkers to identify those at highest risk of developing or harboring a cancer somewhere, and then detecting the cancer using advanced imaging and artificial intelligence in the body. This study is enabling us to develop the foundations relevant to a multi-cancer detection approach but to fully validate these ideas, we will need to collect the proper specimens and imaging to pursue this ambition.

What is the connection between new-onset diabetes (NOD) and pancreatic cancer and how will this project address the two?

Researchers have shown that pancreatic cancer cells alter the ability of our body to process glucose (or sugar), making it more advantageous of the cancer cells to grow and spread to other parts of the body (metastasize). New onset diabetes is essentially a sign that there is a higher risk of developing pancreatic cancer in the near future for a given patient. By focusing on this higher risk population and enriching it further with our blood-based biomarker approach, we aim to improve the ability to detect or rule out pancreatic cancer in the many patients who develop new onset diabetes in the population.