Brachial Plexus Segmentation and Detection Through Ultrasound Image Analysis using Convolutional Neural Networks

Medical image segmentation is widely being used for nerve block/region identification, which in turn, has always been a matter of instinct for anaesthesiologists. In most of the cases, they depend on their instinct, to administer anaesthesia at precise locations. But this procedure does not always yield 100 percent accuracy since the dynamics of the place of administration is subject to variation among the patients. Hence, an ultrasound image of the area containing nerves is acquired in order to identify the nerve structures, however even for experts in the field, this identification process is difficult due to speckle noise and echo disturbances that influence the ultrasound image. Hence, nerve segmentation, incorporating ResNet architecture, makes use of the ultrasound images of required nerve groups. This procedure behaves as an aid, and helps practitioners operate on the nerves in an efficient manner, thereby easing processes such as Peripheral Nerve Blocking(PNB) regional anaesthesia. So, the model is developed by incorporating ResNet50 and few other layers like dense, dropout, and flatten layers to get an output which helps the practitioners, to identify the region of brachial plexus. The proposed model generates an output with 91.00% accuracy.