臨床研究等提出・公開システム

Clinical performance of computer-aided detection of colorectal neoplasia using artificial intelligence

Velidation research of colonoscopy using artificial intelligence

415

We enrolled 415 subjects who underwent colonoscopy at our institution from January 2021 to March 2022. The eligibility criteria were as follows (1) aged 21 to 81 years old on the date of colonoscopy, (2) patients undergoing endoscopy as a primary endoscopic screening for CRC, (3) patients who tested positive for the fecal immunochemical test (FIT) for occult blood, (4) patients with colorectal neoplasia undergoing endoscopic resection, and (5) willingness to participate in the RCT. In contrast, the exclusion criteria included patients who (1) underwent colorectal surgery and (2) had an inflammatory bowel disease. Participants who provided written informed consent and for participation in this study were consecutively enrolled and randomly allocated either to the CADe (n = 207) or control groups (n = 208) at the start of each colonoscopy. The randomization method used in this study was the numbered container method according to the CONSORT guideline. (Moher D, Hopewell S, Schulz KF, Montori V, et al. CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group randomized trials. Int J Surg. 2012 10(1) 28 to 55.) The allocation message was generated by the computer program of the Python random module and then sealed in sequentially numbered identical containers according to the allocation sequence. After an eligibility check, the endoscopist opened the allocation message, and subsequently, examinees were randomly assigned to the CADe or control groups. The background information (Baseline Characteristics) of the target subjects is as follows. The number of people in each age group, in the CADe group, there was 1 individual aged 20-29, 12 individuals aged 30-39, 56 individuals aged 40-49, 63 individuals aged 50-59, 54 individuals aged 60-69, and 21 individuals aged 70-81. On the other hand, the Control group had 1 individual aged 20-29, 15 individuals aged 30-39, 37 individuals aged 40-49, 79 individuals aged 50-59, 52 individuals aged 60-69, and 24 individuals aged 70-81. In addition, the CADe and control groups, mean age was 54.9 and 55.9 years (not significant: n.s.), male ratio was 73.9% and 69.7% (n.s.), the primary screening ratio was 7.2% and 8.2% (n.s.), FIT-positive ratio was 56.5% and 51.9% (n.s.), and surveillance ratio after polypectomy was 36.2% and 39.9% (n.s.), respectively.

IRB registration: January 26, 2021. Participant registration closed: March 31, 2022. Statistical analysis completed and submission to Digestion journal: June 7, 2022. Publication of statistical analysis results in Digestion journal: January 4, 2023. The study concluded with these events.

No adverse events.

The ADR, a primary outcome, was 59.4% in the CADe group and 47.6% in the control group (p = 0.018), and the AMRrs, a secondary outcome, was 11.9% in the CADe group and 26.0% in the control group (p = 0.037).

The colonoscopy with the CADe system yielded an 11.8% higher ADR than that performed by experienced endoscopists alone. We believethat the novel CADe system can lead to considerable advances in colorectal cancer diagnosis.

Jan. 04, 2023

Jan. 04, 2023

https://karger.com/dig/article-abstract/104/3/193/836213/Clinical-Evaluation-of-Computer-Aided-Colorectal?redirectedFrom=fulltext

No

https://jrct.mhlw.go.jp/latest-detail/jRCT1030220081

Nakashima Hirotaka

Foundation for Detection of Early Gastric Carcinoma

2-6-12, Nihombashikayabacho, Chuo Ward, Tokyo. Zip; 103-0025

nakashima@soiken.or.jp

Nakashima Hirotaka

Foundation for Detection of Early Gastric Carcinoma

2-6-12, Nihombashikayabacho, Chuo Ward, Tokyo. Zip; 103-0025

+81-3-3668-6803

nakashima@soiken.or.jp

Complete

Jan. 04, 2021

Interventional

randomized controlled trial

single blind

no treatment control/standard of care control

parallel assignment

diagnostic purpose

Colorectal cancer screening endoscopy, colonoscopy after fecal occult blood positive test, surveillance after colorectal polyp resection

After colorectal surgery or inflammatory bowel disease

Both

Neoplastic polyp of the colon and rectum

We enrolled 415 subjects who under went colonoscopy at our institution from January 2021 to March 2022.The randomization method used in this study was the numbered container method according to the CONSORT guideline. The allocation message was generated by the computer program of the Python random module and then sealed in sequentially numbered identical containers according to the allocation sequence. After an eligibility check, the endoscopist opened theallocation message, and subsequently, examinees were randomly assigned to the CADe or control groups.

colorectal neoplasia

Numbered container method

D018256

D016449

Adenoma detection rate in total colonoscopy

Adenoma miss rate in recto-sigmoid scopy

+81-3-3668-6803

Nov. 19, 2020

none