When it comes to early detection of silent but deadly diseases like pancreatic cancer, finding it early and predicting disease aggressiveness are critical for increasing long-term survival.

With no universal pancreatic cancer screening tool, researchers with The Ohio State University Comprehensive Cancer Center—Arthur G. James Cancer Hospital and Richard J. Solove Research Institute (OSUCCC—James) are turning to artificial intelligence (AI) to help detect pancreatic cancer earlier. or even in precancerous stages.

Most pancreatic cancers have no clear symptoms, which results in a delayed diagnosis. About one fourth, however, begin as cysts that could develop into pancreatic cancer. The challenge is to accurately determine which cysts are at high risk versus low risk of developing into cancer.

“Among all the precancerous cysts removed through surgery, about 50% likely did not need to be removed and more than likely, they wouldn’t have progressed to cancer during a patient’s lifetime,” said Somashekar Krishna, MD, a gastroenterologist and researcher with the OSUCCC—James Molecular Carcinogenesis and Chemoprevention Research Program.

“Most patients affected by pancreatic cancer are over age 60, and many have other medical conditions that make surgery risky. If we can find a way to identify and risk-stratify lesions upfront, many patients can be treated with minimally invasive organ-saving ablation. procedures to eliminate the cyst.”

Using AI to enhance human analysis speed, accuracy

Using mini probe-based laser optical cameras, Krishna and his team reach the pancreas through a procedure called endoscopic ultrasound. This approach allows them to perform a “virtual biopsy,” providing a detailed, microscopic view of the cyst wall. They then use AI to analyze the video from inside the pancreas to identify clinical signs that indicate whether a cyst may develop into cancer.

Identifying cancer risk factors, however, in hundreds of fast-moving video images can be difficult, tedious and sometimes inaccurate, based on the training and experience of the person reviewing the images.

“It’s like asking someone to count all the black, white, and blue cars in a video where the cars are moving at 100 mph. It’s simply impossible without slowing down the video and analyzing hundreds, if not thousands, of individual frames,” said Krishna.

To speed up this process and improve the accuracy of the results, Krishna is training an AI algorithm in collaboration with Ohio State computer science engineer Harry Chao, Ph.D., to recognize and flag clinical characteristics of pancreatic cancer in seconds rather than hours.

“AI represents a potentially powerful diagnostic tool for the early detection of pancreatic cancer. It can not only save lives but also reduce the need for unnecessarily invasive surgeries and their associated costs,” said Krishna. “Additionally, it lowers the risk of medical side effects from organ removal, such as the onset of diabetes and digestive complications.”

Treating pancreatic cancer

Pancreatic cancer that has not spread beyond the pancreas is typically treated with a surgical technique known as the Whipple procedure. This is a large surgery that involves removing the head of the pancreas, the first part of the small intestine, gallbladder and bile duct. Recovery can be challenging for older adults and people with other serious health conditions.

Krishna has pioneered a new minimally invasive endoscopic procedure to treat the precancerous cysts using targeted heat through radiofrequency ablation to eliminate abnormal tissue. This approach offers an organ-saving option for patients with early-stage or precancerous conditions to avoid major abdominal surgery. While it may not be the first option for everyone, Krishna says it provides a valuable alternative for patients who may face higher risks with pancreatic surgery.