Application of Convolutional Neural Network for the Task of Recurrent Laryngeal Nerve Identification

The paper represents the results of research efficiency of the application of machine learning for the task of recurrent laryngeal nerve (RLN) identification. The main approaches to the identification of the recurrent laryngeal nerve in the area of surgical intervention using software and technical tools were analyzed. In particular, the main problem is the classification of information signals that are obtained as a result of electric current stimulation of the tissues of a surgical wound. The use of machine learning is proposed for the classification of these signals. The main concept of data analysis in machine learning, namely the process of audio recognition by using a neural network, is also described. The article considers two types of neural networks Long short-term memory networks (LSTMs) and Convolutional neural networks (CNNs) and justifies the choice of one of them. The process of training the CNNs model on test data of patients during thyroid surgery is described. The results of the assessment of the reliability of recognition of various fragments of patients' information signals are presented and the effectiveness of the proposed approach is confirmed.