Comparison Of An Artificial Intelligence Capsule Endoscopy System (Navicam) with the conventional Capsule Endoscopy (Pillcam Sb3) In The Study Of Small Bowel Pathology: Navipill Pilot Study

Aims Artificial intelligence can improve reading times and lesion detection rates in Small Bowel Capsule Endoscopy (SBCE).Compare the reading with artificial intelligence (AIR) of the Navicam Capsule with the conventional reading (CR) of Pillcam^R SB3 (gold standard).

Methods Prospective clinical trial comparing AIR and CR (non-inferiority study). The 2 SBCE were administered to the same patient in randomized order with 60-minute intervals. AIR and CR were performed by two independent expert investigators. Diagnostic performance (percentage of lesions compatible with the indication for the test), diagnostic accuracy and reading times of the AIR compared to CR were evaluated.

Results Twenty patients (median age 60 ± 10 years) were included. SBCE studies were complete with Navicam and Pillcam in 100% and 95% (p=1). The diagnostic yield by AIR and CR was 80% and 90% (p=0.5). The two readings agreed in 90% of the cases (Cohen's k 0.62). In the analysis per patient, the AIR presented sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and diagnostic accuracy of 88%, 100%, 100%, 60%, and 90%, respectively. In the analysis per lesion, the AIR showed sensitivity, specificity, PPV, NPV, and diagnostic accuracy of 97%, 50%, 95%, 60%, and 93%, respectively. Mean reading time was 3.2 ±3.9 minutes with AIR and 27.5 ±9. 4 minutes with CR (p < 0.001).

Conclusions AIR with Navicam^R shows high agreement with CR with Pillcam^R. AI demonstrates high diagnostic yield and accuracy, with a significant reduction of reading times.

Publication History

Article published online:
14 April 2023