Utilizing Deep Learning to Identify an Ultrasound-guided Nerve Block Target Zone

Area of Science:

• Anesthesiology
• Medical Imaging
• Artificial Intelligence

Background:

• Ultrasound-guided nerve blocks are crucial for perioperative analgesia.
• Nerve identification in ultrasound images is challenging for less experienced anesthesiologists.
• Accurate nerve localization is vital for effective and safe anesthesia delivery.

Purpose of the Study:

• To develop a deep learning algorithm for automatic identification of the transversus abdominis plane in ultrasound images.
• To assist anesthesiologists in performing ultrasound-guided nerve blocks.
• To improve the safety and efficiency of perioperative pain management.

Main Methods:

• Utilized a U-Net architecture for training a deep learning model.
• Employed artificial data augmentation to enhance the training dataset.
• Evaluated model performance using the Dice score coefficient against expert anesthesiologist labels.

Main Results:

• The deep learning model achieved a global Dice score of 73.31% on the test set.
• The model demonstrated effectiveness in identifying the transversus abdominis plane region.
• Results indicate the model's potential as an ultrasound decision support system.

Conclusions:

• Deep learning algorithms show promise in automating ultrasound image analysis for anesthesia.
• The developed model can potentially serve as a decision support tool for anesthesiologists.
• Further development could enhance ultrasound-guided regional anesthesia practices.