Detecting pancreatic disease using artificial intelligence and human ingenuity

Talk like a pancreatologist

(AI) — computer systems able to perform complex tasks that traditionally were only possible using human intelligence

Computed tomography (CT) scan — a medical imaging technique used to obtain detailed images inside the body

Gastroenterology — the branch of medicine that involves studying, diagnosing and treating disorders of the digestive system and related organs, including the oesophagus, stomach, intestines, pancreas, gallbladder and liver

Glucagon — a hormone produced by the pancreas which signals the liver to release glucose into the bloodstream

Glucose — a sugar molecule that is mostly found in carbohydrates and is the body’s primary source of energy

Insulin — a hormone produced by the pancreas which signals the liver to take in glucose from the bloodstream

Jaundice — a condition that causes the skin to appear yellow or green

Metabolism — the chemical reactions happening within the body to convert food into energy

Metastasis — the spread of cancer cells from where they first formed to other sites in the body

Pancreatic ductal adenocarcinoma (PDAC) — a highly aggressive cancer that begins in duct cells in the pancreas

Deep in your abdomen, nestled behind your stomach, sits your pancreas, a leaf-shaped organ about the size of a hand. “The pancreas has important roles in both facilitating digestion and in regulating metabolism,” says Dr Stephen Pandol, a pancreatologist from Cedars-Sinai Medical Center. To aid digestion, the pancreas produces digestive juices full of enzymes that travel to the small intestine, where they break down nutrients from food so they can be absorbed by the body. The pancreas also secretes hormones that enter the bloodstream, which help regulate what the body does with the nutrients it gains. Most significantly, it synthesises insulin and glucagon, the hormones that tell the liver whether to store or release glucose respectively, depending on the body’s needs.

Diseases of the pancreas that disrupt these processes can be serious – even fatal – if not diagnosed and treated properly. The most well-known pancreatic disease is Type 1 diabetes, where the pancreas is unable to produce insulin, preventing the body from regulating its blood sugar levels. Pancreatitis is another disease in which the digestive enzymes produced by the pancreas attack the organ itself. Pancreatic cancers, on the other hand, usually affect the quantity of digestive enzymes produced, which inhibits the body’s ability to absorb nutrients, often leading to acute weight loss. If these cancers metastasise, they can prove fatal.

Studying the pancreas

Stephen’s research group studies the pancreas and its diseases in detail. “In particular, we research pancreatic cancer and pancreatitis,” says Stephen. “We investigate genetic and lifestyle factors that affect the probability of these diseases emerging, as well as their effects on the body, such as pain, weight-loss and jaundice.” Stephen’s group focuses on these diseases because current practices are often unable to treat them effectively. “Treatments can include supportive care, surgery and partially-effective medications,” explains Stephen. “However, a lot of work needs to be done to make these treatments as comprehensive and effective as possible.”

One focus of the group is an especially aggressive type of cancer called pancreatic ductal adenocarcinoma (PDAC), which accounts for over 90% of pancreatic cancer cases worldwide. In its early stages, PDAC can be asymptomatic, which makes it difficult to detect. Once symptoms emerge, the cancer is normally at an advanced stage and has often already metastasised, making treatment particularly challenging. Alarmingly, only around 10% of people who are diagnosed with PDAC will survive longer than five years. However, if the cancer is detected and diagnosed earlier, the survival rate increases dramatically.

Early symptoms of PDAC may include abdominal pain and weight loss, which are common symptoms for many disorders. While patients with these symptoms might undergo computed tomography (CT) scans, the cancer is often missed as it is not visible on the resultant images – at least, not to humans.

Detecting cancer using artificial intelligence

Artificial intelligence (AI) is a hot topic in a diverse array of fields, and the scope of its potential applications within medicine is staggering. “We are using AI to detect early signs of pancreatic cancer before they can be detected by humans,” says Stephen. “Early detection massively improves the likelihood of successful treatment of the disease.”

The team on this project is led in collaboration with Dr Debiao Li, a leading imaging scientist at Cedars-Sinai Medical Center. Debiao and his team trained an AI algorithm using three types of images: CT scans with visible PDAC tumours; CT scans with no visible tumours from patients who later went on to be diagnosed with PDAC; and CT scans of healthy pancreases with no tumours, as a control. Once trained, the AI system was able to identify features invisible to the human eye that could be used to predict and diagnose PDAC. By rolling out this system, it is possible that PDAC could be detected at an earlier stage, which could make treatment much more successful.

What is next?

Stephen’s team is not only focusing on diagnosis. “We are also developing therapeutic drugs that target the underlying mechanisms of pancreatic diseases,” says Stephen. “If we are able to show the effectiveness of these drugs in clinical trials, they could be rolled out to help improve patient outcomes across the world.” Early diagnoses, hand in hand with effective treatments, have the potential to dramatically improve the lives of patients suffering from pancreatic diseases.

AI could further accelerate this trend thanks to its ability to spot patterns in large datasets. People are complex, and each of us has a different array of interacting factors that can affect how we might respond to treatments. AI can search for patterns in data that describe how different patients responded to treatments and use its findings to recommend specific treatments for future patients depending on specific traits, such as other health issues, lifestyle factors or genetic traits. This could pave the way for truly personalised healthcare that ensures each of us receives the most effective treatment possible.