Clinical Validation of Nerveblox, an Artificial Intelligence Software to Support Ultrasound-Guided Regional Anesthesia Procedures

This study aims to validate the performance of AI technology in regional anesthesia through a clinical evaluation conducted on the U.S. population. Data collection method and Analysis: A total of 40 healthy subjects were enrolled and scanned by three different expert anesthesiologists. Twelve different peripheral nerve and plane block regions supported by the Nerveblox software were scanned bilaterally. Each participant spent approximately 30 minutes in the ultrasound scanning session, with each ultrasound clip lasting 10 seconds. Age, gender, weight, and height data were collected and reported for each subject. Subjects were selected to balance the BMI distribution between those with BMI \< 30 kg/m² and those with BMI ≥ 30 kg/m². An FDA-cleared, commercially available ultrasound device was used during the scanning process, and the Nerveblox software was not utilized in the data collection sessions. Collected ultrasound clips included both ideal and non-ideal block views. The collected ultrasound scans were processed post hoc by the Nerveblox software, and the resulting outputs, alongside the raw ultrasound images, will be assessed by a panel of expert anesthesiologists. These clips were reviewed by two separate groups of reviewers, each consisting of three members. Reviewers were asked to evaluate: * Highlighting of Safety-Critical Anatomical Structures: For each safety-critical anatomical landmark on each clip, the majority opinion (at least 2 out of 3) was used to determine the overall panel results. * Image Quality: Experts rated image quality on a scale from 1 to 5. These results were then converted to a binary measure for analysis. * Effect of color highlighting on the perceived risks of ultrasound-guided regional anesthesia Data were expressed as frequencies and shown as a percentage of the total clips analyzed. Inter-rater agreement was reported by block region.