Colorectal Adenoma, Colorectal Polyp
Conditions
Keywords
Artificial Intelligence, Polyp Size, Deep Learning, Colonoscopy, Depth Map
Brief summary
Accurate polyp size measurements are essential for risk stratification, selection of polypectomy techniques, and surveillance interval assignments. Evidence indicated that the clinical implementation of artificial intelligence is an optimal tool to improve the measurement of polyps during colonoscopy. This study aimed to evaluate the performance of a computer-aided measuring (CAM) system (EndoDASS) and compare its accuracy with routine sizing methods during real-time colonoscopy.
Detailed description
This study will be conducted in 2 phases: the first phase will evaluate the value of the application of the computer-aided measuring (CAM) system for polyp size measurement accuracy in a small sample of clinical videos, recording the corresponding clinical videos for CAM measurements after measuring polyp size using real-time visual assessment and non-scaled device (open biopsy forceps or snare) assessment, and comparing the different methods of polyp size measurement through a video-based analysis \[Autonomous artificial intelligence (AI) measurement, AI-assisted human measurement, non-scaled device assessment, and visual assessment\] with different groups of endoscopists ( experts, gastroenterologists, surgeons, fellows) evaluating the corresponding clinical videos during real-time measurements, to obtain pilot data on the relative accuracy of polyp size measurement when using the CAM system, to test the feasibility of size measurement of freshly resected polyp specimens and to determine the sample size and reference for evaluating the CAM system in the subsequent phases of a randomized controlled trial of the study. The second phase will assess the value of using the AI-assisted human measurement versus the non-scaled device assessment for polyp size measurement accuracy in a clinical randomized controlled trial using a prospective, multicenter, patient-single-blind, randomized controlled trial design in which subjects are randomly assigned to the CAM group and the non-scaled device measurement group in a 1:1 ratio. Each patient will have a maximum of 3 polyps included in the study.
Interventions
DIAGNOSTIC_TESTPolyp size measurement using autonomous AI measurement or AI-assisted human measurement with the CAM system
The study of real-time polyp size measurement using the CAM system will be conducted in two phases. Phase I (pilot phase, n=24 polyps, about 27 patients) will be used to assess the feasibility of applying the CAM system in real-time in a clinical video in order to obtain pilot data on the relative accuracy of assessing polyp sizes using autonomous AI measurement and AI-assisted human measurement and to determine the relative accuracy of assessing polyp size in Phase II of the study ( Randomized Controlled Trial) to assess the sample size and reference standards for the CAM system. The sample size for the second phase of the randomized controlled trial (comparing AI-assisted human measurement to non-scaled device) will be based on pilot data in order to compare the relative accuracy of polyp size measurement with AI-assisted human measurement versus non-scaled device in clinical practice.
Sponsors
Changhai Hospital
Study design
Allocation
RANDOMIZED
Intervention model
FACTORIAL
Primary purpose
DIAGNOSTIC
Masking
SINGLE (Subject)
Masking description
Randomization will be concealed. The patient will be blind to the assigned groups. The endoscopist will be blind to the allocated groups until the polyp is identified.
Intervention model description
prospective, multi-endoscopist, single-center or multi- centers
Eligibility
Sex/Gender
ALL
Age
18 Years to 85 Years
Healthy volunteers
No
Inclusion criteria
1. Adults aged 18-75, any gender; 76-85 years eligible case-by-case based on health status. 2. Colonoscopy screening, surveillance, or diagnostic participants. 3. Informed consent obtained.
Exclusion criteria
1. Anticoagulant use (e.g., aspirin, warfarin) within 7 days prior to colonoscopy or coagulopathy. 2. Inflammatory bowel disease. 3. Aronchick score \>3 at entry. 4. Incomplete Case Report Form (CRF) data. 5. Emergency colonoscopy. 6. Pregnancy or lactation. 7. Gastrointestinal obstruction. 8. Refusal to participate.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Evaluation of the computer-aided measuring (CAM) system | 7 days | To evaluate the relative accuracies of the computer-aided measuring (CAM) system measurement (Autonomous AI measurement and AI-assisted human measurement), non-scaled device (open biopsy forceps or snare) assessment or visual assessment for measuring polyp size compared to with measurement of freshly resected polyp specimens by a digital vernier caliper as reference standard. Relative accuracy is defined as \[1-(CAM system measurement - ground truth measurement)/ground truth measurement\] x100%. Scale: 0-100. Higher score is a better outcome. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| video-based analysis | 7 days | Compare in video-based analysis with different endoscopists (experts, gastroenterologists, surgeons, fellows) the relative size measurements accuracy of CAM system measurement with visual and non-scaled device assessment. Relative accuracy is defined as \[1-(CAM system measurement ground truth measurement)/ground truth measurement\] x100%. Scale: 0-100. Higher score is a better outcome. |
| Reliability between CAM system measurement and ground truth measurement | 7 days | Reliability between CAM system measurement, non-scaled device assessment, visual assessment and ground truth measurement is evaluate by concordance correlation coefficient (CCC). CCC is a statistical measure used to evaluate the agreement between two continuous variables.The CCC considers both the correlation and the bias between the two variables. It ranges from -1 to 1；1: Perfect agreement. 0: No agreement. Negative values: Indicate that the variables diverge. |
| Time taken for polyp size measurement | 7 days | Measurement time of the AI-assisted human measurement, non-scaled device assessment, and visual estimation refers to the time taken by endoscopists using various measurement methods, starting from when a polyp appears in the field of view for measurement until the report of the polyp size is completed. |
| Percentage differences between the AI-assisted human measurement and non-scaled device assessment | 7 days | Calculate percentage differences between the AI-assisted human measurement and non-scaled device assessment. |
| Instances of overestimation or underestimation by the AI-assisted human measurement and non-scaled device assessment | 7 days | Evaluate the instances of overestimation or underestimation by the AI-assisted human measurement and non-scaled device assessment. |
Countries
China
Contacts
Primary ContactCheng-Long Wang, MD
Backup ContactXiang-Yu Sui, MD
Outcome results
None listed