TY - JOUR
T1 - Performance of artificial intelligence in colonoscopy for adenoma and polyp detection
T2 - a systematic review and meta-analysis
AU - Hassan, Cesare
AU - Spadaccini, Marco
AU - Iannone, Andrea
AU - Maselli, Roberta
AU - Jovani, Manol
AU - Chandrasekar, Viveksandeep Thoguluva
AU - Antonelli, Giulio
AU - Yu, Honggang
AU - Areia, Miguel
AU - Dinis-Ribeiro, Mario
AU - Bhandari, Pradeep
AU - Sharma, Prateek
AU - Rex, Douglas K.
AU - Rösch, Thomas
AU - Wallace, Michael
AU - Repici, Alessandro
N1 - Publisher Copyright:
© 2021 American Society for Gastrointestinal Endoscopy
PY - 2021/1
Y1 - 2021/1
N2 - Background and Aims: One-fourth of colorectal neoplasia are missed at screening colonoscopy, representing the main cause of interval colorectal cancer. Deep learning systems with real-time computer-aided polyp detection (CADe) showed high accuracy in artificial settings, and preliminary randomized controlled trials (RCTs) reported favorable outcomes in the clinical setting. The aim of this meta-analysis was to summarize available RCTs on the performance of CADe systems in colorectal neoplasia detection. Methods: We searched MEDLINE, EMBASE, and Cochrane Central databases until March 2020 for RCTs reporting diagnostic accuracy of CADe systems in the detection of colorectal neoplasia. The primary outcome was pooled adenoma detection rate (ADR), and secondary outcomes were adenoma per colonoscopy (APC) according to size, morphology, and location; advanced APC; polyp detection rate; polyps per colonoscopy; and sessile serrated lesions per colonoscopy. We calculated risk ratios (RRs), performed subgroup and sensitivity analyses, and assessed heterogeneity and publication bias. Results: Overall, 5 randomized controlled trials (4354 patients) were included in the final analysis. Pooled ADR was significantly higher in the CADe group than in the control group (791/2163 [36.6%] vs 558/2191 [25.2%]; RR, 1.44; 95% confidence interval [CI], 1.27-1.62; P <.01; I2 = 42%). APC was also higher in the CADe group compared with control (1249/2163 [.58] vs 779/2191 [.36]; RR, 1.70; 95% CI, 1.53-1.89; P <.01; I2 = 33%). APC was higher for ≤5-mm (RR, 1.69; 95% CI, 1.48-1.84), 6- to 9-mm (RR, 1.44; 95% CI, 1.19-1.75), and ≥10-mm adenomas (RR, 1.46; 95% CI, 1.04-2.06) and for proximal (RR, 1.59; 95% CI, 1.34-1.88), distal (RR, 1.68; 95% CI, 1.50-1.88), flat (RR, 1.78; 95% CI, 1.47-2.15), and polypoid morphology (RR, 1.54; 95% CI, 1.40-1.68). Regarding histology, CADe resulted in a higher sessile serrated lesion per colonoscopy (RR, 1.52; 95% CI, 1.14-2.02), whereas a nonsignificant trend for advanced ADR was found (RR, 1.35; 95% CI,.74-2.47; P =.33; I2 = 69%). Level of evidence for RCTs was graded as moderate. Conclusions: According to available evidence, the incorporation of artificial intelligence as aid for detection of colorectal neoplasia results in a significant increase in the detection of colorectal neoplasia, and such effect is independent from main adenoma characteristics.
AB - Background and Aims: One-fourth of colorectal neoplasia are missed at screening colonoscopy, representing the main cause of interval colorectal cancer. Deep learning systems with real-time computer-aided polyp detection (CADe) showed high accuracy in artificial settings, and preliminary randomized controlled trials (RCTs) reported favorable outcomes in the clinical setting. The aim of this meta-analysis was to summarize available RCTs on the performance of CADe systems in colorectal neoplasia detection. Methods: We searched MEDLINE, EMBASE, and Cochrane Central databases until March 2020 for RCTs reporting diagnostic accuracy of CADe systems in the detection of colorectal neoplasia. The primary outcome was pooled adenoma detection rate (ADR), and secondary outcomes were adenoma per colonoscopy (APC) according to size, morphology, and location; advanced APC; polyp detection rate; polyps per colonoscopy; and sessile serrated lesions per colonoscopy. We calculated risk ratios (RRs), performed subgroup and sensitivity analyses, and assessed heterogeneity and publication bias. Results: Overall, 5 randomized controlled trials (4354 patients) were included in the final analysis. Pooled ADR was significantly higher in the CADe group than in the control group (791/2163 [36.6%] vs 558/2191 [25.2%]; RR, 1.44; 95% confidence interval [CI], 1.27-1.62; P <.01; I2 = 42%). APC was also higher in the CADe group compared with control (1249/2163 [.58] vs 779/2191 [.36]; RR, 1.70; 95% CI, 1.53-1.89; P <.01; I2 = 33%). APC was higher for ≤5-mm (RR, 1.69; 95% CI, 1.48-1.84), 6- to 9-mm (RR, 1.44; 95% CI, 1.19-1.75), and ≥10-mm adenomas (RR, 1.46; 95% CI, 1.04-2.06) and for proximal (RR, 1.59; 95% CI, 1.34-1.88), distal (RR, 1.68; 95% CI, 1.50-1.88), flat (RR, 1.78; 95% CI, 1.47-2.15), and polypoid morphology (RR, 1.54; 95% CI, 1.40-1.68). Regarding histology, CADe resulted in a higher sessile serrated lesion per colonoscopy (RR, 1.52; 95% CI, 1.14-2.02), whereas a nonsignificant trend for advanced ADR was found (RR, 1.35; 95% CI,.74-2.47; P =.33; I2 = 69%). Level of evidence for RCTs was graded as moderate. Conclusions: According to available evidence, the incorporation of artificial intelligence as aid for detection of colorectal neoplasia results in a significant increase in the detection of colorectal neoplasia, and such effect is independent from main adenoma characteristics.
UR - http://www.scopus.com/inward/record.url?scp=85092180388&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85092180388&partnerID=8YFLogxK
U2 - 10.1016/j.gie.2020.06.059
DO - 10.1016/j.gie.2020.06.059
M3 - Review article
C2 - 32598963
AN - SCOPUS:85092180388
SN - 0016-5107
VL - 93
SP - 77-85.e6
JO - Gastrointestinal endoscopy
JF - Gastrointestinal endoscopy
IS - 1
ER -