Meeting News

Artificial intelligence system automatically detects polyps during colonoscopy

Artificial intelligence-assisted endoscopy was feasible for automatic polyp detection during colonoscopy, according to new research presented at UEG Week.

“The most remarkable breakthrough with this system is that artificial intelligence enables real-time optical biopsy of colorectal polyps during colonoscopy, regardless of the endoscopists’ skill,” study investigator Yuichi Mori, MD, PhD, from Showa University in Yokohama, Japan, said during the opening plenary at UEG Week, according to a press release. “This allows the complete resection of adenomatous polyps and prevents unnecessary polypectomy of non-neoplastic polyps.”

The system adds computer-aided diagnosis to endocytoscopy, a modality delivered by a prototype, ultra-magnifying endocytoscope provided by Olympus. This scope provides a 500-fold magnified image, which can analyze about 300 polyp features when using narrow-band imaging (NBI) mode or methylene blue dye. The system then predicts a lesion’s pathology in less than a second by comparing it with more than 30,000 other endocytoscopic images that were used for machine learning. Preliminary studies have shown this is feasible, but the current study is the first prospective one to do so, according to the press release.

In this study, Mori and colleagues compared the artificial intelligence-assisted system’s prediction vs. the findings in the pathology reports from 250 patients in whom endocytoscopy found colorectal polyps. The system evaluated 306 polyps in real-time, and detected neoplasia with 94% sensitivity, 79% specificity, 86% accuracy, a 79% positive predictive value and a 93% negative predictive value, per the press release.

“We believe these results are acceptable for clinical application and our immediate goal is to obtain regulatory approval for the diagnostic system,” Mori said in the press release. He and his colleagues’ next steps include a multicenter study and development of an automatic polyp detection system.

“Precise on-site identification of adenomas during colonoscopy contributes to the complete resection of neoplastic lesions,” he added. “This is thought to decrease the risk of colorectal cancer and, ultimately, cancer-related death.” – by Adam Leitenberger

References:

Mori Y, et al. Abstract OP001. Presented at: UEG Week; Oct. 28 to Nov. 1, 2017; Barcelona.

Disclosures: Some of the researchers report financial relationships with Cybernet, which developed custom software for analyzing endocytoscopy images, and Olympus, which manufactures the endocytoscope.

 

Artificial intelligence-assisted endoscopy was feasible for automatic polyp detection during colonoscopy, according to new research presented at UEG Week.

“The most remarkable breakthrough with this system is that artificial intelligence enables real-time optical biopsy of colorectal polyps during colonoscopy, regardless of the endoscopists’ skill,” study investigator Yuichi Mori, MD, PhD, from Showa University in Yokohama, Japan, said during the opening plenary at UEG Week, according to a press release. “This allows the complete resection of adenomatous polyps and prevents unnecessary polypectomy of non-neoplastic polyps.”

The system adds computer-aided diagnosis to endocytoscopy, a modality delivered by a prototype, ultra-magnifying endocytoscope provided by Olympus. This scope provides a 500-fold magnified image, which can analyze about 300 polyp features when using narrow-band imaging (NBI) mode or methylene blue dye. The system then predicts a lesion’s pathology in less than a second by comparing it with more than 30,000 other endocytoscopic images that were used for machine learning. Preliminary studies have shown this is feasible, but the current study is the first prospective one to do so, according to the press release.

In this study, Mori and colleagues compared the artificial intelligence-assisted system’s prediction vs. the findings in the pathology reports from 250 patients in whom endocytoscopy found colorectal polyps. The system evaluated 306 polyps in real-time, and detected neoplasia with 94% sensitivity, 79% specificity, 86% accuracy, a 79% positive predictive value and a 93% negative predictive value, per the press release.

“We believe these results are acceptable for clinical application and our immediate goal is to obtain regulatory approval for the diagnostic system,” Mori said in the press release. He and his colleagues’ next steps include a multicenter study and development of an automatic polyp detection system.

“Precise on-site identification of adenomas during colonoscopy contributes to the complete resection of neoplastic lesions,” he added. “This is thought to decrease the risk of colorectal cancer and, ultimately, cancer-related death.” – by Adam Leitenberger

References:

Mori Y, et al. Abstract OP001. Presented at: UEG Week; Oct. 28 to Nov. 1, 2017; Barcelona.

Disclosures: Some of the researchers report financial relationships with Cybernet, which developed custom software for analyzing endocytoscopy images, and Olympus, which manufactures the endocytoscope.