基于多域脑电参数分析的麻醉深度评估

提出一种结合随机森林模型的输出和脑电参数共同评估麻醉深度的方法，以提高评估麻醉深度的可靠性。首先通过滤波方式处理脑电信号，然后把信号分割成等长的多段，从每段中提取非线性域、频域、时域的10种参数，得到脑电参数-BIS值数据集；然后建立评估麻醉深度的随机森林回归模型，并在这些脑电参数中筛选出用于辅助模型评估的参数；最后在测试集上验证模型和参数的效果。该模型在测试集上的估计值与真实值之间存在很好的一致性和相关性（Pearson相关性=0.975），筛选出的参数在测试集上也达到了82.3%的总准确率，表明该方法在评估麻醉深度方面具有较好的应用价值。

Estimating depth of anesthesia based on analysis of multi-domain EEG parameters

Abstract: A method that combines the output of random forest model and EEG parameters is proposed to estimate the depth of anesthesia, thereby improving the reliability of estimating the depth of anesthesia. The EEG signal is divided into multiple segments of equal length after filtering. Ten parameters in the nonlinear domain, frequency domain, and time domain are extracted from each EEG signal segment for constituting the EEG parameter-BIS value data set. Then, a random forest regression model for estimating the depth of anesthesia is established, and the parameter used to assist in the model evaluation is screened out of these EEG parameters. Finally, the performances of the model and the selected parameter in the estimation of the depth of anesthesia are verified on the test set. There are good consistency and correlation between the estimated value on the test set and the true value (Pearson correlation=0.975), and the selected parameter can also achieve a total accuracy of 82.3% on the test set, which shows that the proposed method has a high application value in estimating the depth of anesthesia.