Objective Quality Control for Colonoscopy

Abstract

Purpose: Objective data to explain why a significant number of large polyps and cancers are not detected during colonoscopy do not exist. Here we present a novel, objective approach to automatically extract endoscopic metrics reflecting quality from digital video files created during colonoscopy. Methods: Digital Video Signal Capture. A workstation was developed that digitally captures and stores the complete video file generated during colonoscopy. Quality Control Algorithms. We developed software that automatically extracts 5 metrics from digitized video files created during colonoscopy. Metric 1 measures the overall duration of the insertion phase termed the insertion time. Metric 2 measures the duration of the withdrawal phase termed the withdrawal time. Metric 3 measures the clear withdrawal time defined as the duration of the withdrawal phase without out-of-focus frames. Metric 4 reflects the number of back and forth movements. Metric 5 includes fractions of Metric 3 that are spent on close inspections of the colon wall (off-axial or wall view) or global inspections (axial or lumen view). Results: We created approximately 250 digital video files during colonoscopy. Several critical algorithms were developed that allowed discrimination between clear and out-of-focus frames, forward or backward movement, and presence or abscence of the appendix. The image recognition algorithms that detect out-of-focus frames and movement direction have high sensitivity and specificity (>95%). Appendix recognition on a test dataset had a sensitivity of 90% and a specificity of 86%. Conclusions: We have created novel software that automatically extracts 5 objective quality control metrics from colonoscopy video files. Our method has the potential to provide large scale, continuous quality control for colonoscopy in the day-to-day medical practice setting. Lastly, our method may be useful to assess progress during colonoscopy training, or as part of endoscopic skills assessment evaluations. [figure 1]Figure 1