A Randomized Controlled Multicenter Study of Artificial Intelligence Assisted Digestive Endoscopy

NCT04071678 | Unknown

• 1 other identifier: 研2019-262
• Study Type: observational
• Target: 3,600
• Locations: 1 country
• Sites: 1

Brief Summary

Digestive endoscopy center of the second affiliated hospital of medical college of zhejiang university and engineers of naki medical co., ltd. in Hong Kong independently developed an ai-assisted diagnostic model of digestive endoscopy in the early stage, namely the deep learning model.The deep learning model through the early stage of the study, is able to identify lesions of digest tract.The sensitivity for the diagnosis of some diseases, such as colon polyps, is 99%. On the one hand, this auxiliary diagnostic model can guide endoscopic examination for beginners; on the other hand, it can improve the detection rate of lesions and reduce the rate of missed diagnosis; on the other hand, the overall operating efficiency of the endoscopic center is improved, which is conducive to the quality control of endoscopic examination. Now the AI-assisted diagnostic model has been further improved, and it is planned to carry out further clinical verification in the digestive endoscopy center of our hospital. It is connected to the endoscopic system of our hospital and used simultaneously with the existing image-text system of endoscopy to compare the practicability, sensitivity and specificity of AI-assisted diagnosis model in the diagnosis of digestive tract diseases, and focus on the quality control of endoscopic examination.

Conditions

Artificial Intelligence

Outcome Measures

Primary Outcomes (2)

• Changes of detection rate of digestive tract lesions assisted by artificial intelligence gastroenteroscopy

  Endoscopic examination has a high dependence on the clinical experience and status of endoscopists, and the quality of endoscopic examination of endoscopists can be reduced by high-load work, and problems such as incomplete examination site coverage, incomplete detection of lesions, and incomplete image collection are easy to occur. Artificial intelligence does not have this weakness. It does not reduce its ability to work over a long period of time, and its assistance is expected to improve the detection rate of lesions

  2 years

• The accuracy of AI-assisted diagnostic model evaluating the intestinal readiness score

  The quality of intestinal preparation determines the quality of colonoscopy, which is evaluated by endoscopists through the Boston score. The ai-assisted diagnostic model can also be automatically graded.The Boston bowel score is used to determine whether the bowel is adequately prepared. The Boston bowel score is divided into 4 grades (0\~3 points) from worst to cleanest. The higher the score is, the better the bowel is prepared and more conducive to colonoscopy.

  2 years

Study Arms (3)

A: Model A

Mode A was silent mode, back-to-back with endoscopic physicians to simultaneously display endoscopic images and record video, but did not interfere with the operation of endoscopic physicians.After the operation, the AI model automatically generates an endoscopy report, which is compared with the official report given by the endoscopy doctor in the endoscopy system. If the difference is large, video verification shall be played back immediately or endoscopic examination shall be performed again before the patient wakes up

Behavioral: Careful examination during endoscopic procedures to identify lesions

B: Model B

Mode B is a delayed reminder mode. If the lesion is found during the operation, it is required to be moved to the middle of the visual field within 5 seconds. If the lesion has been detected by the AI model (the lesion has been circled in the picture), but the doctor does not move the lesion to the middle of the visual field within 5 seconds, the AI system will give an alarm prompt

Behavioral: Careful examination during endoscopic procedures to identify lesions

C: Model C

Mode C is a real-time reminder mode, which is an alarm prompt when the focus is captured in the visual field.

Behavioral: Careful examination during endoscopic procedures to identify lesions

Interventions

Careful examination during endoscopic procedures to identify lesions BEHAVIORAL

When the AI model alarms, check carefully to confirm the lesion

A: Model A; B: Model B; C: Model C

Eligibility Criteria

• Age: 18 Years+
• Sex: all
• Healthy Volunteers: No
• Age Groups: Adult (18-64), Older Adult (65+)
• Sampling Method: Probability Sample

Study Population

Patients who underwent painless gastroenteroscopy at the endoscopy center from September 2019 to August 2021

You may qualify if:

• Voluntarily sign the informed consent for this study
• Stable vital signs
• Over 18 years old
• Patients requiring painless gastroenteroscopy for various reasons

You may not qualify if:

• Unable or unwilling to sign a consent form, or unable to follow research procedures
• have contraindications to painless gastroenteroscopy
• Vital signs are unstable
• The lesions have been identified by gastroenteroscopy in other hospitals, which is to further confirm the patients who come to our hospital for endoscopic examination
• Endoscopic treatment, such as polypectomy, pylorus narrow dilatation and so on

Sponsors & Collaborators

• Second Affiliated Hospital, School of Medicine, Zhejiang University: lead

Study Sites (1)

Cai J Ting

Hangzhou, Zhejiang, 310000, China

RECRUITING

Study Officials

• Cai J Ting, Dr

  Second affiliated hospital of school of medicine, zhejiang university

  STUDY DIRECTOR

Central Study Contacts

Wang J An, Dr

CONTACT

057187783759; HREC2013@126.com

Cai J Ting, Dr

CONTACT

15267019902; 1173920428@qq.com

Study Design

• Study Type: observational
• Observational Model: COHORT
• Time Perspective: PROSPECTIVE
• Sponsor Type: OTHER
• Responsible Party: SPONSOR

Study Record Dates

• First Submitted: August 26, 2019
• First Posted: August 28, 2019
• Study Start: August 1, 2019
• Primary Completion: August 1, 2021
• Study Completion: December 30, 2021
• Last Updated: October 22, 2019

Record last verified: 2019-08

Data Sharing

• IPD Sharing: Will not share

Locations

CN (1)