The implantation of electrodes on the skull of mice requires precise positioning. For these purposes, it is necessary to create a coordinate system that rests on the cranial sutures. The aim of the work is to train a model based on convolutional neural networks. The model should automatically draw the seam line. Investigating this issue, a method for solving the problem of seam approximation based on U-net is proposed. This study faces the difficult problem of accurately locating the cranial suture. As part of the study, a structure similar to U-net was developed and thought out. The results of the work will allow biologists to perform more accurate stereotaxic operations, by inserting electrodes into a more precise position, to receive certain signals from the brain, in order to record changes when various aromas are applied to the mouse receptors.
Keywords: convolutional neural network (CNN), seam approximation, fully convolutional network (FCN), deep learning, semantic segmentation, object segmentation, neural network, computer vision, biomedical image, machine learning, image processing