Ultrasound-guided nerve blocks has been at the forefront of pain management for patients who undergo surgery. However, identifying nerve structures in ultrasound feeds poses a challenge for novice anesthesiologists. By leveraging the power of artificial image augmentation on publicly available ultrasound images, our team was capable of training several convolutional neural networks (CNNs) to automatically highlight nerve regions in real-time. With over 1000 hours of continuous model training on state-of-the-art computing hardware, our team was able to accurately locate nerves in 10 regions:
- Transversus Abdominis Plane (TAP)
- Radial Nerve
- Median and Ulnar Nerves
- Pectoralis I
- Femoral Nerve
- Popliteal Sciatic Nerve
- Rectus Sheath
- Interscalene Axillary Nerve
- Infraclavicular Axillary Nerve
- Supraclavicular Axillary Nerve
As a result of our findings, three publications have been submitted and published in top academic journals.