Evaluation of Cerebral Autoregulation Function Based on TCD Signal

Transcranial Doppler (TCD) ultrasound is a commonly used clinical method to evaluate cerebral vascular function. The cerebral autoregulation (CA) function is a mechanism to keep the cerebral blood flow constant when the blood pressure fluctuates. Certain cardio-cerebrovascular diseases may undermine the function to some extent. However, there is no gold standard to evaluate whether the function is normal or not. The most widely used TFA method and ARI index only give a qualitative rule. Therefore, it is necessary to find a reliable evaluation method to identify people with impaired CA function. Based on the TCD signals of normal group and diabetes patient group in the physionet dataset, this paper first uses the features and indicators extracted by the above two methods to classify the subjects. Secondly, a one-dimensional convolution network structure is designed, and its classification accuracy is better than that of machine learning method, reaching 85.33%. Finally, the structure of siamese neural network is used and a new loss function is designed to further improve the accuracy to 92.00%. The method proposed in this paper has preliminarily verified the feasibility of assessing whether the CA function of different groups is damaged, but more clinical patient data is necessary to improve the accuracy and reliability of the method.