|
摘要：
目的 探索对丙泊酚药物浓度和意识状态改变敏感的脑功能特征。 方法 选择拟行胸腹部手术治疗的男性患者6例，对所有患者使用靶控输注设备进行丙泊酚输注，初始药物浓度设置为1.0 mg/L，然后每6 min增加0.2 mg/L，直到患者达到无意识状态。同时，采集患者的脑电信号。将采集的脑电信号进行预处理并提取脑功能特征。构建药物浓度相同、意识状态不同和意识状态相同、药物浓度相同两个数据集，并使用线性判别分析（linear discriminant analysis, LDA）、逻辑回归（logistic regression, LR）和支持向量机（support vector machine, SVM）3种机器学习算法模型在不同的数据集上进行分类分析。 结果 特征筛选结果显示功率谱‑δ‑顶区、排列熵（permutation entropy, PE）‑δ‑颞区和相位滞后指数（phase lag index, PLI）‑α‑顶区到顶区为对意识状态改变敏感的特征，使用这3个特征在LDA、LR和SVM机器学习算法模型上得到的意识状态分类正确率分别为（82±5）%、（83±5）%和（84±4）%；功率谱‑β‑额区、功率谱‑β‑颞区和功率谱‑β‑顶区对药物浓度改变敏感，使用这3个特征在LDA、LR和SVM机器学习算法模型上得到的药物浓度分类正确率分别为（77±4）%、（76±4）%和（80±4）%。功率谱‑δ‑顶区在药物浓度变化时没有明显变化（P>0.05），而在意识状态从有意识到无意识明显升高（P<0.001）；PE‑δ‑颞区和功率谱‑β‑额区在从低药物浓度到高药物浓度和从有意识到无意识时均明显降低（P<0.001）；PLI‑α‑顶区到顶区在低药物浓度状态下明显高于其在高药物浓度状态下（P<0.001），但是其在意识状态变化时并没有明显变化（P>0.05）；功率谱‑β‑颞区和功率谱‑β‑顶区在意识状态变化时均没有出现明显变化（P>0.05），但是功率谱‑β‑颞区从低药物浓度到高药物浓度时明显升高（P<0.001），而功率谱‑β‑顶区从低药物浓度到高药物浓度时明显降低（P<0.05）。 结论 功率谱‑δ‑顶区可以很好地表征意识水平的变化同时避免药物浓度的影响，在精确监测麻醉深度上具有潜力。
|
|
Abstract: Objective To explore the brain functional characteristics sensitive to propofol drug concentration and consciousness changes. Methods A total of six male patients undergoing thoracic and abdominal surgery were enrolled. Propofol was infused via a target‑controlled infusion system. The initial drug concentration of propofol was 1.0 mg/L, with an increase of 0.2 mg/L every six min until unconsciousness was achieved. Meanwhile, the electroencephalography data were recorded. Then, the collected electroencephalography data were preprocessed to analyze brain function characteristics. Two data set were established, which were same drug concentration and different consciousness as well as same drug concentration and same consciousness. Then, three machine learning algorithm models including linear discriminant analysis (LDA), logistic regression (LR) and support vector machine (SVM) were used to classify the above data sets. Results According to characteristics screening, the parietal δ‑power, permutation entropy (PE)‑δ in the temporal area, and parietal‑parietal phase lag index (PLI)‑α were sensitive to changes in consciousness, and their accuracy of consciousness classification based on LDA, LR, and SVM models were (82±5)%, (83±5)%, and (84±4)%, respectively. The frontal, temporal and parietal β‑power were sensitive to changes in drug concentration, and their accuracy of drug concentration classification based on LDA, LR and SVM models were (77±4)%, (76±4)%, and (80±4)%, respectively.The parietal δ‑power did not change significantly when the drug concentration changed (P>0.05), but it significantly increased in unconsciousness, as compared with consciousness (P<0.001). The PE‑δ in temporal area and the frontal β‑power were significantly reduced both from low drug concentrations to high drug concentrations and from conscious to unconscious (P<0.001). The parietal‑parietal PLI‑α showed no significant difference between different conscious levels (P>0.05), whereas it significantly enhanced in low drug concentration level, as compared with high drug concentration level (P<0.001). The temporal and parietal β‑power did not significantly change when the conscious state changed (P>0.05), but the temporal β‑power significantly increased (P<0.001) and the parietal β‑power significantly decreased (P<0.05) in low drug concentration level, as compared with high drug concentration level. Conclusions The parietal δ‑power can well characterize the change of consciousness without the influence of drug concentration, which has the potential to accurately monitor the depth of anesthesia.
|
