Withdrawal time (WT) is the interval during which the gastroenterologist inspects the colon mucosa for abnormalities, from the moment of cecal intubation (reaching the endpoint of the large intestine) to the complete withdrawal of the colonoscope. It is one of the most important quality indicators of a colonoscopy: studies show that longer withdrawal times are associated with higher adenoma detection rates (ADR – Adenoma Detection Rate), and clinical guidelines recommend a minimum of 6–9 minutes.

This service provides gastroenterologists with an automated and standardized way to analyze the quality of their colonoscopies from a withdrawal time perspective, based on video recordings of the procedures. Unlike a single global WT measurement, the service calculates withdrawal time separately for each anatomical segment of the colon, enabling identification of regions where inspection time was insufficient.

A key aspect of the calculation is that non-informative intervals are automatically excluded from the quality assessment: cleaning periods, moments when the colonoscope is in close contact with the colon wall, and other situations with poor visibility. The report therefore reflects only the effective useful inspection time.

What the service analyzes

The AI system automatically processes the colonoscopy video recording and provides:

• Automatic frame-by-frame identification of the anatomical colon segments (cecum, ascending colon, transverse colon, descending colon, sigmoid colon, rectum)
• Calculation of withdrawal time per segment — how much time was spent in each region of the colon
• Automatic separation of useful inspection time from non-informative intervals (cleaning, close wall contact, etc.)
• A structured report with quality indicators, comparable to thresholds recommended by clinical guidelines

Who this is for

This service is intended for gastroenterologists who want to evaluate and improve their own colonoscopy performance from a withdrawal time quality perspective. By providing objective and reproducible measurements, the service supports:

• Periodic self-assessment of withdrawal technique
• Identification of colonic segments where inspection time is systematically insufficient
• Progress monitoring over time and comparison against recommended clinical standards

Input data

The service accepts colonoscopy video recordings in .mp4 format.

The video must cover the colonoscope withdrawal phase, starting from the cecum.

Results

The AI model analyzes each frame of the video and produces a classification label for the corresponding colonic segment, as well as a frame-type classification (useful inspection or non-informative). Based on these predictions, a downloadable report is generated with a detailed breakdown of withdrawal time per anatomical segment — total time, time per colonic segment, and the percentage of useful versus non-informative time.

The service also provides an interactive visualizer of the analyzed video, featuring a color-coded timeline indicating the identified colonic segments throughout the procedure, marking of non-informative intervals, and the WT report displayed directly in the interface.

How to access the service

The service runs on a dedicated external platform, accessible through your institutional account. Authentication uses the same credentials as the RNCC platform (Keycloak).

Usage workflow:

• Access the platform using the button below
• Log in with your institutional credentials
• Upload the colonoscopy video in .mp4 format
• Wait for automatic processing — duration depends on video length
• Download the report and explore the annotated visualizer
• The platform retains a full history of all previously processed videos