Deep Learning-Based Tracking of Neurovascular Features Toward Semi-Automated Ultrasound-Guided Peripheral Nerve...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Regional Anesthesia

Background:

• Peripheral nerve blocks reduce general anesthesia and opioid use for critical pain management.
• Underutilization of nerve blocks stems from the high skill required for accurate needle insertion.
• Ultrasound image guidance with AI offers a semi-automated solution for regional anesthesia.

Purpose of the Study:

• Develop and characterize deep learning algorithms for real-time ultrasound image interpretation.
• Identify anatomical landmarks and aimpoints for safe needle placement in nerve blocks.
• Establish a foundation for an integrated AI-guided regional anesthesia platform.

Main Methods:

• Trained deep learning algorithms on over 55,000 ultrasound images from 20 porcine models.
• Evaluated AI system performance for in vivo landmark detection in the femoral nerve region.
• Conducted prospective live animal testing for aimpoint identification accuracy and speed.

Main Results:

• Achieved an average area under the precision-recall curve of 0.92 (SD = 0.03) for landmark detection.
• Demonstrated a 98.3% success rate in identifying aimpoints during live animal testing.
• Recorded an average aimpoint identification time of 40.5 seconds (SD = 33.5).

Conclusions:

• AI-powered ultrasound interpretation can accurately identify anatomical landmarks for nerve blocks.
• The developed AI system shows significant potential for improving regional anesthesia delivery.
• Future integration with robotics aims to create more accessible anesthesia methods.