AI platform doubles detection rates for early-stage esophageal cancer

Xinhua | Updated: 2024-04-18 11:20
Share - WeChat

HANGZHOU -- A group of Chinese scientists have developed an artificial intelligence (AI) platform that doubles the detection rate for early-stage esophageal cancer.

Esophageal cancer is often asymptomatic, but experienced clinicians can detect tumors and precancerous lesions with endoscopies. The five-year survival rate is over 90 percent when the tumor is treated during the early stages while the clinical outcomes decline sharply once patients start experiencing symptoms.

The study, published Thursday in the journal Science Translational Medicine, described an architecture driven by deep learning algorithms, which are trained via datasets of more than 190,000 esophageal images gathered from several clinics in China.

The clinical trial, run by researchers from Taizhou Hospital in Zhejiang Province and Wuhan University's Renmin Hospital, recruited more than 3,000 participants who underwent endoscopies. The AI system was used on half the volunteers.

This platform helps double the detection rate for high-risk esophageal lesions (1.8 percent) compared with that in the control group (0.9 percent), according to the study.

Also, the real-time AI tool has shown a high sensitivity and specificity of 89.7 percent and 98.5 percent, respectively, in the real world.

"The endoscopic assistance system, to a significant extent, enables endoscopists with less experience to enhance detection rates of high-risk esophageal lesions, thereby, reducing the frequency of missed diagnoses," said Mao Xinli, the corresponding author of the paper, from Taizhou Hospital.

Copyright 1995 - . All rights reserved. The content (including but not limited to text, photo, multimedia information, etc) published in this site belongs to China Daily Information Co (CDIC). Without written authorization from CDIC, such content shall not be republished or used in any form. Note: Browsers with 1024*768 or higher resolution are suggested for this site.
License for publishing multimedia online 0108263

Registration Number: 130349