ENVIRONMENTAL IMPACT OF OPTICAL DIAGNOSIS BY ARTIFICIAL INTELLIGENCE IN COLONOSCOPY: A PROSPECTIVE TRIAL.

Background and aims: Artificial intelligence with computer-aided diagnosis (CADx) enables real-time optical distinction between neoplastic and non-neoplastic colon polyps during colonoscopy, potentially reducing unnecessary polypectomies and histopathological assessments, however, its environmental impact is unknown. We quantified the carbon footprint of using CADx-assisted optical diagnosis in colonoscopy in a multicentre prospective trial.

Methods: We conducted a clinical trial to assess the value of CADx-assisted optical diagnosis in colonoscopy in the UK, Norway, and Japan. We used trial data and compared the two scenarios with focus on environmental effect of colonoscopy: 1) standard care scenario where all polyps are removed and evaluated with histopathology; 2) CADx-assisted optical diagnosis scenario where ≤5mm polyps predicted as non-neoplastic in the distal colon and rectum are not removed (leave-in-situ strategy). We calculated the estimated carbon dioxide emission (CO2e) in each scenario based on life-cycle assessments of the following four components: number of snares, polyp traps, histopathological processing and CADx devices. We also projected the study results onto the annual number of colonoscopy procedures in the UK (704,125 colonoscopies per year).

Findings: We included 1,134 patients (59% male, median age 67 years) in the trial. Compared to the standard care, CADx-assisted optical diagnosis reduced the number of snares and polyp traps by 18% (95% confidence interval: 16%-20%), corresponding to 93kg CO2e reduction. Histopathological processing decreased by 26% (23%-28%), corresponding to 127kg CO2e reduction. However, implementation of CADx devices in this study increased CO2e by 153kg due to both hardware production and energy-consuming machine learning processes. Overall, the CADx assisted optical diagnosis resulted in a net reduction of 67kg CO2e per 1,134 colonoscopies, corresponding to 59g CO2e per colonoscopy. Nationwide introduction of the CADx-assisted optical diagnosis in the UK could potentially lead to a significant reduction of CO2e (approximately 41 tons). This reduction is comparable to the environmental impact of a 153,450 km drive in an average passenger car (equivalent to driving from London to Vienna a hundred times).


Conclusion: A prospective trial-based analysis showed CADx-assisted optical diagnosis could potentially reduce the environmental impact of colonoscopy.