Deep Capsule® (Deep Capsule US)

Deep Capsule® is an artificial intelligence-assisted reading tool designed to help gastroenterologists review small bowel capsule endoscopy videos more efficiently. It automatically detects and highlights potential small bowel lesions in video frames to support the clinician's final diagnosis, and is not intended to replace clinical decision-making.

Deep Capsule® uses a convolutional neural network-based algorithm to detect small bowel lesions without differentiating them, similar to the predicate device NaviCam ProScan. The main difference is that NaviCam ProScan also identifies digestive tract location (oral cavity, esophagus, stomach, small bowel), whereas Deep Capsule® focuses only on lesion detection. Both devices output AI-selected lesion frames presented in a structured gallery format for reviewer assessment.

The submission references ISO 13485 clause 7.3 (Design controls), ISO 13484 clause 8.3 (Nonconforming product), and ISO 13485 clause 8.5 (Corrective and preventative action) as part of Quality Management System Regulation compliance. FDA guidance documents on software validation and cybersecurity were followed for verification and validation testing.

Deep Capsule® is substantially equivalent to predicate device NaviCam ProScan because both are computer-aided detection tools for capsule endoscopy that use AI-based algorithms to highlight potential small bowel lesions for the same patient population (adults with suspected small bowel bleeding). Clinical validation demonstrated non-inferior diagnostic yield (96.1% vs. 76.1% for standard-of-care) and image-level sensitivity/specificity (94% and 84.9%) comparable to predicate performance, with consistent results across demographic subgroups.