Subscribe to RSS
DOI: 10.1055/s-0042-1744561
ARTIFICIAL INTELLIGENCE FOR REAL-TIME OPTICAL DIAGNOSIS OF NEOPLASTIC POLYPS DURING COLONOSCOPY
Aims Artificial intelligence using computer-aided diagnosis (CADx) may enable colonoscopists to distinguish between neoplastic polyps requiring removal and non-neoplastic polyps not requiring removal during colonoscopy. This may reduce costs and resources and prevent polyp overtreatment in colorectal cancer screening.
Methods We performed a multicenter clinical trial comparing a novel CADx system using real-time ultra-magnifying polyp visualization during colonoscopy with standard visual inspection of small (≤5mm in diameter) polyps in the sigmoid colon and the rectum for optical diagnosis of neoplastic histology. All polyps were subsequently removed. The primary endpoint was sensitivity for neoplastic polyps compared to histopathology. Secondary endpoints were specificity and colonoscopist confidence level of optical diagnosis.
Results We assessed 1,289 individuals for eligibility at colonoscopy centers in Norway, the United Kingdom and Japan. We detected 892 eligible polyps in 518 patients and included them in analyses; 359 neoplastic and 533 non-neoplastic. Sensitivity for diagnosis of neoplastic polyps with standard visual inspection was 88.4% (95% confidence interval (CI) 84.3-91.5) compared to 90.4% (95%CI 86.8-93.1) with CADx (p=0.33). Specificity was 83.1% (95%CI 79.2-86.4) with standard visual inspection and 85.9% (95%CI 82.3-88.8) with CADx (p=0.04). The proportion of polyp assessment with high confidence was 74.2% (95%CI 70.9-77.3) with standard visual inspection versus 92.6% (95%CI 90.6-94.3) with CADx (p<0.001).
Standard diagnosis |
CADx diagnosis |
|
---|---|---|
Sensitivity -% (95% CI) |
88.4 (84.3-91.5) |
90.4 (86.8-93.1) |
Specificity -% (95% CI) |
83.1 (79.2-86.4) |
85.9 (82.3-88.8) |
High confidence of optical diagnosis -% (95%CI) |
74.2% (70.9-77.3) |
92.6% (90.6-94.3) |
Conclusions Real-time polyp assessment with CADx did not significantly increase sensitivity for neoplastic polyps, but increased specificity and improved confidence of optical diagnosis.
(UMIN no. 000035213; funding Norwegian Research Council, Norwegian Cancer Society, Japan Society for the Promotion of Science)
Publication History
Article published online:
14 April 2022
© 2022. European Society of Gastrointestinal Endoscopy. All rights reserved.
Georg Thieme Verlag KG
Rüdigerstraße 14, 70469 Stuttgart, Germany