AI tool boosts pancreatic cancer detectionBy Lin Hui-chin and Dennis Xie / Staff reporter, with staff writerNational Taiwan University (NTU) Hospital has developed an artificial intelligence (AI) algorithm to improve radiologists’ ability to detect pancreatic cancer during computed tomography (CT) scans. Photo: Lin Hui-chin, Taipei TimesWhile CT scans are the main tool used to detect pancreatic cancer, about 40 percent of tumors smaller than 2cm evade detection, Liao said. During the early stage, pancreatic cancer shows almost no signs or symptoms, and the tumor often has indistinct borders, making it difficult to identify even by professional radiologists, he said. Liao cited a case in which a CT scan showed no signs of pancreatic cancer, but an endoscopic ultrasonography identified a 1.5cm tumor in the patient. The algorithm, trained by 3,000 pieces of image data, can yield a 98.3 percent sensitivity rate in distinguishing image patches of pancreatic cancer from non-cancerous ones, higher than the 92.9 percent of radiologists, he said.

Source: Taipei Times November 04, 2020 15:56 UTC