AI for the early detection of pancreatic cancer

Scientists at the DAMO Academy, a research initiative of the Alibaba Group, have developed an early detection method based on AI technologies that recognises pancreatic cancer with high accuracy. The method is an important medical breakthrough on the way to comprehensive screening for this type of cancer. Until now, pancreatic cancer has often only been detected at a late stage, which makes treatment more difficult. As a result, pancreatic cancer is the seventh most common cause of cancer-related death worldwide, with an average five to ten per cent annual survival rate.

The algorithm used in the screening was trained with more than 3,200 image sets, according to the scientists. It achieved a specificity of 99.9% and a sensitivity of 92.9%. Out of 1,000 tests, there was only one false positive result. The algorithm thus surpasses the performance of human radiologists. In collaboration with more than ten of the world's leading medical institutions, the DAMO Academy researchers have applied the AI-based screening method to more than 20,000 patients. In the process, 31 cases of pathological changes were detected that had not previously been recognised by doctors. The model has so far been used over 500,000 times in hospitals and medical examinations in China.

„Early detection of pancreatic cancer is difficult with conventional examinations, leading to late diagnosis and therefore a poor prognosis. AI and non-contrast CT technology promise to become an effective and cost-effective tool for early detection of pancreatic cancer and enable comprehensive pancreatic cancer screening,“said Le Lu, head of the medical AI team at Alibaba's Damo Academy and a fellow of the IEEE (Institute of Electrical and Electronics Engineers), a global professional association of engineers, technicians, scientists and related professions, mainly in the fields of electrical engineering and information technology.

Further tests required

The accuracy of the algorithm is better than that of some recognised early detection methods such as the Pap test for cervical cancer or mammography for breast cancer. It therefore makes sense to integrate this special method into large-scale screening programmes. AI-based screening is a promising approach that could have clinical implications in the near future," comment Jörg Kleeff and Ulrich Ronellenfitsch from Martin Luther University Halle-Wittenberg and the University Hospital Halle (Saale). However, the professors also point out that further tests are required before the procedure can be introduced into widespread practice.

The results were published in Nature Medicine.

Message of the "transcript" from 09/02/2024