Agreement Between Artificial Intelligence and Anesthesiologists in Ultrasound-Guided Axillary Brachial Plexus Block

Overview

This prospective observational study aims to evaluate the agreement between artificial intelligence (AI)-assisted target point identification and experienced anesthesiologists during ultrasound-guided axillary brachial plexus block. Ultrasound guidance is widely used in regional anesthesia to improve block success and safety. However, accurate identification of anatomical structures and optimal injection points remains operator-dependent. Artificial intelligence-based systems have the potential to assist clinicians by identifying anatomical landmarks in real time. In this study, AI-generated target points will be compared with those determined by experienced anesthesiologists. The level of agreement between the two methods will be analyzed. Secondary outcomes will include block performance parameters and image quality. The findings of this study may contribute to understanding the clinical utility of AI in ultrasound-guided regional anesthesia.

Ultrasound-guided axillary brachial plexus block is a widely used regional anesthesia technique for upper extremity surgeries. The success of the procedure largely depends on accurate identification of neural structures and optimal injection points, which are operator-dependent. Artificial intelligence (AI) has recently emerged as a promising tool for assisting ultrasound interpretation by automatically identifying anatomical structures. However, the level of agreement between AI-based target point identification and expert anesthesiologists has not been sufficiently investigated, particularly in axillary brachial plexus block. In this prospective observational study, patients undergoing upper extremity surgery under axillary brachial plexus block will be included. No additional intervention will be performed on patients within the scope of the study. All evaluations will be based on real-time ultrasound imaging obtained as part of routine clinical practice. During routine ultrasound examination prior to block performance, images will be observed in real time. Experienced anesthesiologists will determine anatomical structures and optimal target injection points during the procedure. Simultaneously, the AI-based system will analyze the same real-time ultrasound images and identify target points. For each identified nerve (median, ulnar, radial, and musculocutaneous), both AI and anesthesiologists will determine target injection points. The spatial difference between AI-generated and expert-defined target points will be calculated in millimeters. The primary objective is to evaluate the agreement between AI and anesthesiologists in target point identification using the intraclass correlation coefficient (ICC). Additionally, a difference of ≤5 mm between measurements will be considered clinically acceptable agreement. Secondary outcomes will include: Proportion of measurements within ≤5 mm agreement Agreement in nerve identification Procedure-related parameters All expert evaluations will be performed independently and blinded to AI outputs. This study aims to determine whether AI can reliably assist clinicians in identifying anatomical targets during ultrasound-guided regional anesthesia without introducing any additional risk to patients.