Juan Wang

Background. Postoperative delirium (POD) poses a significant risk to patients, and accurate prediction of POD can provide guidance for positive interventions. Although many studies have applied machine learning (ML) to electronic health records to predict POD, there has been a lack of studies utilizing electroencephalogram (EEG) data to accurately predict POD. Methods. This retrospective study included 142 patients from two hospitals, among whom 33 developed POD. We extracted multi-domain features from high-density EEG (HD-EEG), screened 39 key features using recursive feature elimination, and analyzed the feature space structure using linear discriminant analysis (LDA). Four ML models, namely support vector machine (SVM), logistic regression (LR), random forest (RF), and gradient boosting machine, were compared. The dataset was divided using 5-fold cross-validation for model training and testing. Results. Compared to the other three models, SVM (linear kernel) performed best in predicting POD, achieving a classification accuracy of 95.71%. Alpha (8–12 Hz) and theta (4–8 Hz) powers, combined with nonlinear dynamics, critically contributed to the model. Furthermore, the performance of the linear models (SVM and LR) was significantly ( p < 0.001) better than that of the nonlinear models, with LDA confirming strong feature space separability (Jensen-Shannon divergence = 0.69). Conclusions. HD-EEG data can be used to establish high-quality ML models for POD predictions. Multidomain features enable more comprehensive integration of EEG information related to delirium. The feature space structure significantly affects the performance of the prediction model, with the linear model offering distinct advantages in predicting POD.

Brain lateralization is considered evolutionarily adaptive. Impaired functional specialization is thought to cause abnormal lateralization in neurological disorders. However, the dynamic changes in brain laterality in Alzheimer's disease (AD) remain unclear. In this study, resting‐state fMRI data and neuropsychological assessments from 109 participants (49 AD patients and 60 healthy controls [HC]) were used. Dynamic laterality time series were constructed by extracting the dynamic laterality index (DLI) within each sliding window. We assessed two key features: laterality reversal (LR), reflecting intra‐hemispheric processing efficiency, and laterality fluctuation (LF), indicating inter‐hemispheric communication. Group differences in dynamic laterality characteristics were analyzed using statistically rigorous methods, regressing out gender, age, years of education, and head movements. Spearman correlation analyses examined the relationship between laterality characteristics and cognitive performance. Our results showed that AD patients exhibited a more pronounced loss of left lateralization as well as stronger right lateralization, especially in the somatomotor network (SMN) and default mode network (DMN). Additionally, we observed decreased LR as well as increased LF with global trends in AD. These divergent changes disrupted the dynamical balance between intra‐ and inter‐hemispheric information interaction. Notably, this imbalance depended on the degree of lateralization, and the higher order cognitive networks with high‐level lateralization were more vulnerable. Importantly, the observed abnormal lateralization metrics were associated with worse cognitive impairment. Our study highlights a disruption of dynamic lateralization balance in higher order cortical networks in AD patients and reveals its potential role in the disease's pathophysiology.

This paper addresses the $H_{\infty }$ output regulation problem of affine nonlinear systems with unmeasurable states and completely unknown dynamics. The $H_{\infty }$ output regulation problem is formulated into a two-player zero-sum differential game problem with the control input and disturbance being the two adversary players. A model-based global state- feedback optimal solution to the $H_{\infty }$ output regulation problem is presented by solving a Hamilton-Jacobi-Isaacs (HJI) equation, and a state- feedback reinforcement learning (RL) algorithm is proposed to solve the HJI equation without knowledge of system dynamics. A linear data-driven observer is designed to estimate the unmeasurable states and a novel output- feedback RL algorithm is developed based on the state- feedback RL framework by employing a critic-actor-disturbance neural network (NN) with a hidden layer. The salient found of this work is that, under the proposed output- feedback RL algorithm, the closed-loop system can reach globally stable. A numerical example is provided to demonstrate the effectiveness of the proposed solution.

In the realm of complex network analysis, identifying crucial nodes represents a formidable challenge that has attracted considerable research focus. Historically, research efforts often neglected the significance of incorporating spatial node information, thereby constraining the accurate identification of key nodes. Addressing this gap, this study introduces an advanced centrality model named Degree-K-shell-Betweenness centrality (DKBC), leveraging gravitational principles. The DKBC model integrates node degree, spatial positioning, and betweenness centrality attributes to enhance the precision of crucial node identification for complex networks. This innovative approach surpasses conventional gravity-based methodologies in efficacy. Evaluation of the proposed algorithm employs the Susceptible-Infectious-Recovered (SIR) Epidemic model and assesses correlation using the Kendall coefficient τ. Comparative analysis against a standard benchmark algorithm underscores the superior performance of the DKBC model. Empirical validation conducted across twelve real-world networks substantiates the exceptional accuracy of DKBC model in pinpointing critical nodes. This research significantly advances the field by demonstrating the efficacy of integrating spatial information into centrality metrics for enhanced network analysis and application.

Objective. General anesthesia (GA) can induce reversible loss of consciousness. Nonetheless, the electroencephalography (EEG) characteristics of patients with minimally consciousness state (MCS) during GA are seldom observed. Approach. We recorded EEG data from nine MCS patients during GA. We used the permutation Lempel–Ziv complexity (PLZC), permutation fluctuation complexity (PFC) to quantify the type I and II complexities. Additionally, we used permutation cross mutual information (PCMI) and PCMI-based brain network to investigate functional connectivity and brain networks in sensor and source spaces. Main results. Compared to the preoperative resting state, during the maintenance of surgical anesthesia state, PLZC decreased (p < 0.001), PFC increased (p < 0.001) and PCMI decreased (p < 0.001) in sensor space. The results for these metrics in source space are consistent with sensor space. Additionally, node network indicators nodal clustering coefficient (NCC) (p < 0.001) and nodal efficiency (NE) (p < 0.001) decreased in these two spaces. Global network indicators normalized average path length ( L a v e / L r ) (p < 0.01) and modularity (Q) (p < 0.05) only decreased in sensor space, while the normalized average clustering coefficient ( C a v e / C r ) and small-world index ( σ ) did not change significantly. Moreover, the dominance of hub nodes is reduced in frontal regions in these two spaces. After recovery of consciousness, PFC decreased in the two spaces, while PLZC, PCMI increased. NCC, NE, and frontal region hub node dominance increased only in the sensor space. These indicators did not return to preoperative levels. In contrast, global network indicators L a v e / L r and Q were not significantly different from the preoperative resting state in sensor space. Significance. GA alters the complexity of the EEG, decreases information integration, and is accompanied by a reconfiguration of brain networks in MCS patients. The PLZC, PFC, PCMI and PCMI-based brain network metrics can effectively differentiate the state of consciousness of MCS patients during GA.

Moxibustion is under active research as a complementary and alternative treatment for various diseases such as pain. “Heat-sensitization” responses have been reported during suspended moxibustion, whose occurrence is associated with significantly better therapeutic effects. The present study aimed to investigate the cortical activities of this interesting phenomenon by a standardized low-resolution brain electromagnetic tomography. We performed electroencephalography recording in a group of patients with chronic low back pain before, during, and after moxibustion treatment at Yaoyangguan (DU3) areas. 11 out of 21 subjects experienced strong heat-sensitization during moxibustion, which were accompanied with significant decreases of current densities in the beta frequency bands in prefrontal, primary and second somatosensory, and cingulate cortices, as well as increased current densities in the alpha2 band in the left insula. No changes were detected in patients without sensitization responses, or in the post-moxibustion phase of either group. These data indicated widespread activity changes across different frequency bands during heat-sensitization. Cortical oscillatory activities could be used to evaluate the “heat-sensitization” responses during suspended moxibustion.