Enhancing Peripheral Nerve Blocks: Real-Time Ultrasound Guidance for Nerve Detection and Needle Placement

Area of Science:

• Medical imaging
• Artificial intelligence in medicine
• Regional anesthesia

Background:

• Ultrasound-guided nerve blocks are crucial for localized pain relief but face challenges in complexity and interpretation.
• Current limitations hinder widespread adoption and require significant practitioner expertise.

Purpose of the Study:

• To develop and validate an AI-assisted platform for real-time nerve and needle detection in ultrasound-guided transversus abdominis plane blocks.
• To address the procedural complexity and interpretation challenges associated with ultrasound-guided nerve blocks.

Main Methods:

• Utilized a dataset of annotated ultrasound images to train R2U-Net based segmentation models for nerve and needle detection.
• Integrated trained models into an Android application for real-time visualization with low latency (<200 ms) and high frame rates (24 FPS).

Main Results:

• Achieved a Dice score of 0.8384 for nerve detection and 0.8309 for needle detection.
• The AI platform demonstrated real-time performance suitable for clinical application.

Conclusions:

• The AI-assisted system has the potential to improve precision, reduce practitioner variability, and increase access to ultrasound-guided nerve blocks.
• This tool can support clinicians, especially those with less experience, by enhancing procedural guidance and confidence.