Study on the Use of Artificial Intelligence (Fujifilm) for Polyp Detection in Colonoscopy (Fuji AI)

Universitätsklinikum Hamburg-Eppendorf

Status: Enrolling

Conditions: Colonoscopic Control After Polypectomy; Screening Colonoscopy; Suspected Colon Polyps

Treatments: Procedure: colonoscopy

Study type: Interventional

Funder types: Other

Identifiers

NCT04894708

PV7284

Details and patient eligibility

About

Colonoscopy is currently the best method of detection of intestinal tumors and polyps, particularly because polyps can also be biopsied and removed. There is a clear correlation between the adenoma detection rate and prevented carcinomas, so adenoma detection rate is the main parameter for the outcome quality of diagnostic colonoscopy. The efficiency of preventive colonoscopy needs optimisation by increase in adenoma detection rate, as it is known from many studies that approximately 15-30% of all adenomas can be overlooked. This mainly applies to smaller and flat adenomas. However, since even smaller polyps may be relevant for colorectal cancer development, the aim of colonoscopy should be to preferably be able to recognize all polyps and other changes.The latest and by far the most interesting development in this field is the use of artificial intelligence systems. They consist of a switched-on software with a small computer connected to the endoscope processor; the patient's introduced endoscope is completely unchanged.

The present study therefore compares the adenoma detection rate (ADR) of the latest generation of devices with high-resolution imaging from Fujifilm with and without the connection of artificial intelligence.

Full description

Methods of Computer Vision (CV) and Artificial Intelligence (AI) provide completely new opportunities, e.g. in the automatic polyp detection and differentiation of a lesion based on its endoscopic image. Computer vision using artificial intelligence methods means the application of "trained" so-called deep neural net (DNN) with a set of defined images (e.g. everyday scenes) and well-known solutions ( e.g. name of the pictured item; c.f. e.g. the "ImageNet Challenge"). The technical feasibility of using AI algorithms in endoscopy has already been proven in many cases. In the present study, it is an AI system from Fujifilm, which is already clinically usable. By using Fujifilm high-resolution imaging devices in colonoscopies, AI will be added randomly.

Enrollment: 1,572 estimated patients

Sex: All

Ages: 35+ years old

Volunteers: Accepts Healthy Volunteers

Inclusion criteria

• Persons> 35 years of age who are capable of giving informed consent
• Planned diagnostic colonoscopy (clarification of symptoms, polyp follow-up)
• Screening colonoscopy for men >50 or women > 55 years of age

Exclusion criteria

• Colon bleeding
• Colon carcinoma
• Known polyps for removal
• Inflammatory bowel disease
• Colonic stenosis
• Other suspected colon disease for further clarification
• Follow-up care after colon cancer surgery (partial colon resection)
• Anticoagulant drugs that make a biopsy or polypectomy impossible
• Poor general condition (ASA IV)
• Incomplete colonoscopy planned

Trial design

Primary purpose: Diagnostic

Allocation: Randomized

Interventional model: Parallel Assignment

Masking: None (Open label)

---

1,572 participants in 2 patient groups

AI colonoscopy

Other group

Description:

colonoscopy with artificial intelligence added

Treatment:

Procedure: colonoscopy

conventional colonoscopy

Sham Comparator group

Description:

conventional colonoscopy

Treatment:

Procedure: colonoscopy