TOP 100 and detection of colorectal lesions in colon capsule endoscopy: more than meets the eye

Aims Colon capsule endoscopy(CCE) can identify colorectal lesions(CRL). However, there are no studies reporting the accuracy of TOP100, a CCE software tool, for automatic detection of CRL in CCE. We aimed to evaluate the performance of TOP100 in detecting CRL in patients submitted to CCE for incomplete colonoscopy(IC) compared to classic reading(CR).

Methods Retrospective cohort-study including all adult patients submitted to PillCam COLON2 Medtronic for IC between 2017-2022. One experienced reader performed CR and other, blinded to CR results, used TOP100 to identify CRL. Detection of CRL, namely polyps, angiectasia, blood, diverticula, erosions/ulcers, neoplasia and subepithelial lesions(SEL) was assessed and TOP100 performance was evaluated.

Results 188 CCE were included. Prevalence of polyps, angiectasia, blood, diverticula, erosions/ulcers, neoplasia and SEL using TOP100 and CR, was 45.7%vs45.7%, 9.0%vs8.0%, 1.6%vs1.6%, 48.4%vs48.4%, 0.5%vs0.5%, 0.5%vs0.5%, 0.5%vs1.1%, respectively. TOP100 and CR had a strong agreement(Kappa 0.897,P<0.001) regarding the presence of CRL and in detecting polyps(Kappa 0.893,P<0.001), angiectasia(Kappa 0.932, P<0.001), blood(Kappa 1.000,P<0.001), diverticula(Kappa 0.957,P<0.001), erosions/ulcers(Kappa 1.000,P<0.001), neoplasia(Kappa 1.000,P<0.001) and SEL(Kappa 1.000,P<0.001). There was 88.3%(n=166) of complete agreement in the detection of CRL. However, in eleven CCE, TOP100 identified lesions missed using CR and in eleven CCE CR identified lesions missed using TOP100.

Conclusions TOP100 is a useful tool in assisting the reader in decreasing missed CRL and should be systematically assessed after CR of the entire video of CCE.

 

Scroll to top

Conflicts of interest

Authors do not have any conflict of interest to disclose.

Publication History

Article published online:
14 April 2023