National Taiwan University Hospital fully deploys pancreatic cancer imaging AI

National Taiwan University Hospital has recently launched a new AI-powered diagnostic imaging service for pancreatic cancer.

Offered as a self-pay service at the NTUH Department of Medical Imaging since late 2024, it features an AI system touted as the first of its kind in the world. 

WHAT IT'S ABOUT 

The PANCREASaver, which the hospital developed with the Institute of Applied Mathematical Sciences at National Taiwan University, leverages deep learning algorithms and a multi-image training dataset to automatically analyse CT scans, delineate the pancreas, and flag suspicious lesions. The system directly integrates with NTUH's PACS and is accessible across departments, including gastroenterology, surgery, and oncology.

In national clinical validation trials, the AI system demonstrated 80% sensitivity in detecting early-stage pancreatic tumours of less than 2 centimetres and an overall diagnostic accuracy of over 90%. 

The technology has received regulatory approval from the Taiwan Food and Drug Administration, as well as a breakthrough device designation from the United States Food and Drug Administration. 

WHY IT MATTERS

According to NTUH, pancreatic cancer is one of the most challenging cancers to treat and diagnose, with death cases gradually increasing in Taiwan in recent years and the five-year survival rate remaining at around 10%. It notes how early-stage pancreatic tumours – especially those below 2 cm that are imperceptible to the human eye – are difficult to spot, "with approximately 40% of lesions missed." 

Hence, the research team behind PANCREASaver developed their AI system to address this challenge in pancreatic cancer diagnosis. 

The team is now working to expand the system's application to other pancreatic-related conditions, such as pancreatitis and pancreatic cystic lesions. They are also incorporating more multi-modal imaging and clinical data to further expand the diagnostic system's coverage and accuracy. 

THE LARGER TREND

Fujitsu and the Southern Tohoku General Hospital in Japan also developed a similar AI-driven solution for diagnosing pancreatic cancer early. They started creating the diagnostic software in 2022, leveraging Fujitsu's AI that was trained with the anonymised pancreas CT images.

Meanwhile, NTUH is also working on large language model development. Late last year, it acquired two new supercomputers, which enable it to incorporate more data types into LLMs. The hospital aims to develop multimodal LLMs to optimise operations and enhance patient services.