Automatic brachial plexus segmentation to facilitate targeted echo block using convolutional neural networks

Kurzfassung

The brachial plexus is a network of nerves located in the shoulder. It carries movement and sensory signals from the spinal cord to the arms and hands. The brachial plexus block is the most widely used anesthetic method in surgical procedures of the upper limbs. By using ultrasound imaging, trained professionals proceed to carry out echo-directed brachial plexus block procedure. However, the identification and location of the brachial plexus is difficult, which in turn may lead to health complications to the patient. Convolutional neural networks could assist this procedure by providing an automatic identification and segmentation of the region of interest. In this paper, by using the dataset provided by the competition Ultrasound Nerve Segmentation, an U-Net model is trained. Our implementation automatically locates and segments the brachial plexus in the aforementioned dataset. In our experiment, the model achieves a dice coefficient (DSC) score of DSC = 0.87. It is concluded that the proposed method satisfactory locates and segments the brachial plexus. Our findings are in line with other deep learning works, sugg