"The system has the potential to assist endoscopists in obtaining targeted biopsies and reducing gastric neoplasm miss rates in clinical practice," Dr. Honggang Yu of the Renmin Hospital of Wuhan University, in China, told Reuters Health by email.
The first-line tool for detecting gastric neoplasms is white-light endoscopy; however, the subtle nature of these neoplasms and the varying skills among endoscopists can result in missed lesions.
While previous research suggests trained AI models can improve the detection rate of early gastric cancer during white-light endoscopy, many of these studies were retrospective and rarely confirmed in randomized trials, Dr. Yu and his colleagues note in The Lancet Gastroenterology & Hepatology.
In their new study, they tested an AI system designed to identify focal lesions and detect gastric neoplasms during endoscopy. They enrolled patients from the Renmin Hospital who were undergoing routine upper GI endoscopy for either screening, surveillance, or assessment of symptoms.
The patients were randomly assigned to undergo AI-assisted upper GI endoscopy first followed by routine upper GI endoscopy (n=907) or routine upper GI endoscopy first followed by AI-assisted upper GI endoscopy (n= 905). At the end of the second examination, researchers performed targeted biopsies for all detected lesions.
The AI system used by the researchers, ENDOANGEL-LD, is a deep-learning-based system designed to detect focal lesions in the gastric mucosa. Dr. Yu told Reuters Health in an email that the AI system has been applied in approximately 350 hospitals in China.
"The system can achieve real-time diagnosis during endoscopy and does not affect the routine endoscopic operation," he explained, adding that using the tool did not increase procedure time.
In the study, patients who underwent AI-assisted upper GI endoscopy first had a significantly lower gastric neoplasm miss rate compared with those who first underwent a routine GI endoscopy without AI (6.1% vs. 27.3; P=0.015).
The miss rate per patient was 6.4% in the AI-first group versus 25.6% in the routine-first group (P=0.024).
Bleeding from a post-biopsy targeted lesion was the only adverse event reported. This event, which was managed with hemostasis, was reported in the routine-first group.
An advantage of AI that the researchers didn't expect to find, according to Dr. Yu, was that the system could build a communication bridge between endoscopists and assistants.
"According to feedback from endoscopists and nurses who used the system in clinical practice, it could expedite communication and increase the efficiency of biopsy," Dr. Yu said. "In future AI clinical trials, time cost and satisfaction evaluations of endoscopists and assistants could be added to evaluate the effects of AI on improving working efficiency."
The study did not have commercial funding.
SOURCE: https://bit.ly/3C1BGF5 Lancet Gastroenterology & Hepatology, online July 20, 2021.
By Brandon May
Posted on
Previous Article
« Icotinib tops chemo as adjuvant therapy for EGFR-mutant NSCLC Next Article
CT imaging may help distinguish COVID-19 from interstitial lung disease »
« Icotinib tops chemo as adjuvant therapy for EGFR-mutant NSCLC Next Article
CT imaging may help distinguish COVID-19 from interstitial lung disease »
Related Articles
July 28, 2021
Lenvatinib plus gefitinib promising in liver cancer
November 6, 2020
Surgeon technical skill has big impact on colon-cancer outcomes
© 2024 Medicom Medical Publishers. All rights reserved. Terms and Conditions | Privacy Policy