Is software assisted polyp detection reliable in the presentation of colorectal polyps in colon capsule endoscopy?
Aims: Compared to conventional colonoscopy, colon capsule endoscopy (CCE) proved to have a high sensitivity with respect to the detection of clinically relevant colorectal polyps (CP). However, a major limitation is the time consuming video reading.
Objective: To assess the feasibility and accuracy of the software assisted analysis “Quickview” (QV) in the detection of CP, the numbers of frames showing a colon pathology in normal-mode and QV-mode were compared.
Design and settings: Prospective, double center trial.
Patients: This analysis involved CCE-2 videos of 65 patients (mean age 56 years).
Intervention: CPs detected during normal mode reading of CCE-2 videos were distributed into episodes and frames. At increasing QV-levels (10, 20 – 80%) all frames showing the CP were recorded.
Main Outcome Measurements: Efficiency of QV in presenting polyps < 6 mm and ≥6 mm, and identifying patients with the need for subsequent therapeutic colonoscopy, capsule egestion rate and cleansing level of bowel preparation were assessed.
Results: At a QV-setting of 30%, the QV video presented at least one frame in 89% of the polyps ≥6 mm and in 86% of polyps of any size (per polyp analysis). With a sensitivity of 95% QV presented ≥1 image in patients with significant polyps at a QV setting of 10% (per patient analysis). The capsule excretion rate was 74%. Overall colon cleanliness was adequate in 85% of the patients.
Limitations: No blinding of the QV reading to the normal mode results, heterogenous patient population, nonconsecutive patients.
Conclusion: The informatics algorithm Quickview presents ≥1 frame in nearly 90% of significant polyps at a 30% setting. With respect to the per patient analysis QV is even more effective at a setting of 10% in determinating patients with significant polyps in 95% of the cases.