Identification of nerve in ultrasound images using U-net architecture

Surgery is undesirable because it is painful and often causes at least a short-term discomfort. Anesthesia is used for reducing the pain during the surgery. When the surgery involves only a local procedure, regional anesthesia is employed instead of total anesthesia. Regional Anesthesia consists of injecting the anesthetic in a nerve block as close to the nerve as possible in order to immobilize a part of the body. While human anatomy is known reasonably perfectly, it has been noticed that anatomy changes from person to person, making it impossible to know the exact location of a nerve. Ultrasound images provide some information on the anatomy of the particular patient under surgery, and it is possible to use this information for improving regional anesthesia. This paper considers the problem of detecting nerve in an ultrasound image of a part of the neck, known as Brachial Plexus. A supervised method is used in the form of artificial neural networks in order to develop a classification rule for the said purpose. A set of ultrasound images is used as the training set, while another set of similar images is used as the test set, and is used for cross-validation of the resulting classification. The images are pre-processed using Linear Gabor Binary Patterns (LGBP) before segmenting them, and the segments are then supplied to the ANN for developing the detection function.