Clinical summary
A computer-aided detection (CADe) system significantly improves polyp detection in routine practice
| Title | A Computer-Aided Detection (CADe) System Significantly Improves Polyp Detection in Routine Practice |
| Authors | Keswani RN, Thakkar U, Sais A, et al. |
| Journal | Clinical Gastroenterology and Hepatology |
| Pub date | |
| DOI |
Abstract
Takeaway message
- Computer-aided detection (CADe) is found to significantly improve screening and surveillance polypectomy rates as well as screening adenoma detection rate (ADR) and serrated detection rate (SDR) in majority users.
- Further work is needed to determine best practices for reducing individual and system barriers to widespread implementation of CADe in practice.
Background and aims
The ADR, an important metric in preventing colorectal cancer (CRC), measures how often neoplastic polyps are found. Various methods have been implemented to improve ADR, but their impact has been limited. Artificial intelligence, specifically CADe, has shown promise in enhancing polyp detection in clinical trials. However, there is conflicting data on its effectiveness in routine practice. The study aimed to evaluate the impact of CADe on polyp detection in a large group of experienced colonoscopists after implementation in their practice.
Study design
A retrospective analysis was conducted at a single urban academic medical center to assess the implementation of the GI Genius™ CADe system during screening or surveillance colonoscopies. The study included patients who underwent these procedures over a nine-month period from March 2022 to December 2022. The CADe system was introduced in a staggered manner, with colonoscopists using it on some days ("CADe room") and not using it on others ("non-CADe room"). Colonoscopists who performed at least 100 colonoscopies over the 9 month period, both with and without CADe, were included in the study. Colonoscopists were encouraged but not mandated to utilize CADe.
The primary outcome of the study was the rate of polyp removal during screening and surveillance colonoscopies. Secondary outcomes included the ADR for screening colonoscopies, the SDR, as well as the total procedure time and withdrawal time.
Key data
In this study, 27 colonoscopists performed a total of 7,456 screening or surveillance colonoscopies, with 1,972 using a CADe system and 5,454 without. Among the colonoscopists, 22 completed a survey about their CADe usage, with 10 self-identifying as CADe majority users (use in > 50% of cases) and 12 as CADe minority users (use in < 50% of cases). The primary reasons for not using CADe were concerns about distracting "false positives" and a perception that CADe didn't significantly improve polyp detection. More than half of colonoscopists surveyed felt that CADe helped detect polyps they would have otherwise missed and 9% believed CADe helped detect clinically significant polyps.
The study found that in rooms where CADe was used, there was a significant increase in screening and surveillance polypectomy rates, screening colonoscopy ADR, and SDR, as seen in the table below. These improvements were more pronounced among CADe majority users compared to CADe minority users. Total procedure time did not differ significantly between CADe and non-CADe rooms.
| CADe not available | CADe available | P value | |
|---|---|---|---|
| All Colonoscopists (n=27) | |||
| All Procedures Polypectomy Rate | 60.02% | 66.93% | <0.0001 |
| Screening/Surveillance Polypectomy Rate | 65.27% | 70.99% | <0.0001 |
| Screening ADR | 39.70% | 44.42% | 0.003 |
| Screening + Surveillance Adenoma Detected | 45.20% | 49.49% | 0.001 |
| Screening SDR | 14.94% | 17.56% | 0.03 |
| Screening + Surveillance Serrated Detected | 15.71% | 18.56% | 0.004 |
| Colonoscopists who self-reported using CADe for majority of cases (n=10)* | |||
| All Procedures Polypectomy Rate | 60.33% | 69.21% | <0.0001 |
| Screening/Surveillance Polypectomy Rate | 62.42% | 71.97% | <0.0001 |
| Screening ADR | 37.59% | 45.75% | 0.0003 |
| Screening + Surveillance Adenoma Detected | 43.69% | 50.63% | 0.0001 |
| Screening SDR | 14.20% | 17.73% | 0.03 |
| Screening + Surveillance Serrated Detected | 15.19% | 19.40% | 0.002 |
| Colonoscopists who self-reported using CADe for minority of cases (n=12)* | |||
| All Procedures Polypectomy Rate | 58.48% | 65.13% | <0.0001 |
| Screening/Surveillance Polypectomy Rate | 67.76% | 68.92% | 0.5 |
| Screening ADR | 42.58% | 43.58% | 0.7 |
| Screening + Surveillance Adenoma Detected | 46.70% | 48.04% | 0.5 |
| Screening SDR | 16.29% | 17.36% | 0.6 |
| Screening + Surveillance Serrated Detected | 16.29% | 17.67% | 0.4 |
* Five users did not fill out survey to self-report CADe usage |
|||
Conclusion
CADe significantly improves colonoscopy quality in routine practice, specifically among colonoscopists who reported using it for the majority of cases. The study suggests that addressing barriers to CADe adoption may further improve its implementation. However, the study acknowledges some limitations, such as potential biases introduced by the study design and the lack of a mandate for CADe usage.
