An Image Segmentation Approach for Synthetic Nervous Tisular Data at The Human Forearm Region with Convolutional Neural Networks
Electrical impedance tomography allows biomedical images to be obtained safely, non-invasively and at low cost. In the region of the human forearm, aspects such as the low dynamism and the conductive nature of the tissues impede the reliable recognition of the nerve terminals. The segmentation of images in electrical impedance tomography is a challenging field and has not been addressed for the area of the human forearm up to date. This study presents a segmentation tool for synthetic tomograms that emulates the conductive behavior of the human forearm, through a U-Net architecture CNN. Segmentation has close to 96 % accuracy and is able to detect the nerve conductivity regions in different positions. Keywords - Electrical Impedance Tomography, Human Forearm, Image Segmentation, Convolutional Neuronal Networks, U-Net.