Quality components of colonoscopy include the detection and complete removal of colorectal polyps, which are precursors to CRC. However, endoscopic ablation may be incomplete, posing a risk for the development of "interval cancers". The investigators propose to develop a solution based on artificial intelligence (AI) (CADp computer-aided decision support polypectomy) to solve this problem.This research project aims to develop CADp, a computer decision support solution (CDS) for the ablation of colorectal polyps from 1 to 20 mm.
This research project aims to develop CADp, a computer-based decision support (CDS) solution for the removal of colorectal polyps ranging from 1-20 mm. The investigators will use a video and image dataset of polypectomy procedures to train the CADp model; thus, it can provide real-time overlaid video feedback for polypectomy procedures based on five specific metrics: 1) estimation of polyp size; 2) prediction of morphology and histology; 3) suggestion of an appropriate resection accessory and technical approach based on the characteristics, size, and histology of the polyp according to current guidelines; 4) image overlay, based on semantic image segmentation technology, showing the extent of the lesion and suggestion of an appropriate resection margin contour around the polyp to ensure its complete removal; 5) post-resection analysis to identify any remnant polyp tissue or insufficient resection margin that may increase this risk. The investigators will collect a set of images and video data from live polypectomy procedures to leverage recent advances in AI technology to train deep learning models. This dataset will be obtained prospectively from a cohort of adults (ages 45-80) undergoing screening, diagnostic, or surveillance colonoscopies. To train the CADp solution, the investigators will obtain the corresponding completeness of resection status using the yield of post-resection margin biopsies. The dataset will be divided into two groups, the training, and the CADp test, respectively.
Study Type
INTERVENTIONAL
Allocation
NA
Purpose
SUPPORTIVE_CARE
Masking
NONE
The AI system will capture the live video of the procedure and the AI feedbackwill be shown on a second screen installed next to the regular endoscopy screen. Screen A will show the regular endoscopy image and screen B will show the regular endoscopy image together with the areas that might harbor a polyp and the information to help the polypectomy.
Centre Hospitalier Universitaire de Montréal
Montreal, Quebec, Canada
Accuracy of the CADp system
accuracy with which the CADp system predicts completeness of polypectomy in the test set with the reference standard for completeness being determined by the histology of post-polypectomy margin biopsies; if free from any polyp tissue (adenomatous, serrated or hyperplastic), the resection will be considered complete. If remnant polyp tissue is detected in any one or more of the margin biopsies the resection is deemed incomplete
Time frame: 3 weeks
Completeness of polypectomy
We will evaluate the agreement between the different subjective and objective ways of assessing the completeness of the polypectomy : evaluation of margins (presence or not, measurement of margins) by endoscopists self-assessment, and by expert consensus.
Time frame: 1 month
Training CADp
Evaluation of the concordance of data on polyp size, extension of margins around the polyp, quality of resection between clinical data (endoscopists' self-assessment and experts' assessments) and CADp prediction.
Time frame: 1 month
Validity of the choice of primary outcome
Based on the results and comparison of the different assessment methods, we will perform sensitivity analyses to assess the validity and robustness of the choice of primary outcome.
Time frame: 1 month
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.