Shanghai Wision AI Co., Ltd develops computer-aided diagnostic algorithms and systems to improve the accuracy and effectiveness of diagnostic imaging. Based in Shanghai, China, the company has extensive expertise in mathematics, algorithm development, software and hardware engineering and works closely with top-tier medical institutions in China and around the world. The company integrates medical knowledge into flexible and scalable models that leverage cutting-edge convolutional neural network and general-purpose computing to achieve high sensitivity and specificity in detection, segmentation and measurement in diagnostic imaging. Future US based clinical trials for Wision AI are expected to take place soon, and the company is looking for clinical partners in Europe. Wision AI is advancing its transformative mathematical medicine approach in multiple clinical settings, including gastroenterology, ophthalmology, and pathology.
Today the company announced results of the first-ever prospective, randomized controlled clinical trial of an artificial intelligence (AI)-based system for disease diagnosis. The study, which was conducted in 1,058 patients undergoing colonoscopy, found that the Wision AI system significantly increased adenoma and polyp detection rates and the mean number of adenomas and polyps detected per patient.
Clinicians at the Center for Advanced Endoscopy at Beth Israel Deaconess Medical Center (BIDMC), Harvard Medical School and the Sichuan Provincial People’s Hospital conducted the study using the automated polyp detection system developed by Wision AI. The study results appear in the current issue of Gut : https://gut.bmj.com/content/early/2019/02/27/gutjnl-2018-317500
“The miss rate for colon polyps can be up to 27 percent. This limits the efficacy of screening colonoscopy, which is a critical tool for reducing the incidence and mortality of colorectal cancer by detecting and removing adenomatous polyps,” said Tyler Berzin, MD, Co-Director, GI Endoscopy, and Director, Advanced Endoscopy Fellowship at BIDMC and Assistant Professor of Medicine at Harvard Medical School. “The results of this study clearly demonstrate that a high-performance, real-time automatic polyp detection system based on a deep learning algorithm can significantly increase polyp and adenoma detection, especially those that are 5 mm or less. Given its high accuracy, fidelity and stability, the Wision AI system could enable improved detection of colon polyps and adenomas in clinical practice.”
“Given its high accuracy, fidelity and stability, the Wision AI system could enable improved detection of colon polyps in clinical practice,” said Tyler Berzin, MD, Co-Director, GI Endoscopy.
A total of 1,130 patients met the eligibility criteria for the study and were randomized to undergo routine colonoscopy (n=567) or computer-assisted diagnosis (CAD) colonoscopy (n=563). Subsequent to randomization, 72 patients met exclusion criteria, resulting in 536 patients undergoing routine colonoscopy and 522 undergoing CAD colonoscopy. In the CAD group, the real-time automatic polyp detection system was used to assist the endoscopist. The endoscopist focused mainly on the original monitor during the procedure and was prompted to look at the system monitor by a sound alarm. The endoscopist was required to check every polyp that the system detected.
Key findings from the study include:
- Polyp detection
- The mean number of polyps detected per colonoscopy in the control and the CAD group were 0.51 and 0.97, respectively (p<0.001).
- There was a 1.89-fold increase in the mean number of polyps detected between the two groups (p<0.001).
- The polyp detection rate of the control and CAD group was 0.29 and 0.45, respectively (p<0.001).
- Adenoma detection
- The mean number of adenomas detected per colonoscopy in the control and CAD group were 0.31 and 0.53, respectively (p<0.001).
- There was a 1.72-fold increase in the mean number of adenomas detected between the experimental and control groups (p<0.001).
- The adenoma detection rate of the control and CAD groups was 0.20 and 0.29, respectively (p<0.001).
- There was a total of 39 false alarms (non-lesion area continuously traced) in the CAD group, averaging 0.075 false alarms per colonoscopy; all polyps detected by the endoscopist in the CAD group were also detected by the automatic detection system.
- Excluding the additional biopsy time, overall procedure time between the control and CAD groups were similar.
- The improvement in the adenoma detection rate between the CAD and control group was statistically significant in patients with normal bowel but not in patients with excellent bowel prep.
- There were no complications reported.
“The aim of a CAD system is not to show how smart the AI is, but to be useful in clinical settings, and this reflects our commitment to innovating AI-based diagnostics with proven clinical benefits,” said JingJia Liu, Chief Executive Officer at Wision AI. “The study shows that use of our system did not increase procedure time, which is mainly because of the low false alarm rate. This demonstrates that the system can improve polyp and adenoma detection without negatively impacting physician and facility efficiency. The observation that our system provided an even superior detection in patients with normal bowel prep is also important given that fully complying with bowel prep procedures may be a challenge for some patients, and not all patients achieve excellent prep. We believe the results of this study are highly compelling and warrant further evaluation of our technology in screening colonoscopy or other diagnostic indications.”