A Novel Framework for Estimating Time-Varying Multivariate Autoregressive Models and Application to Cardiovascular Responses to Acute Exercise

Measuring Spinal Cord Potentials and Cortico-Spinal Interactions After Wrist Movements Induced by Neuromuscular Electrical Stimulation

Electroencephalographic (EEG) correlates of movement have been studied extensively over many years. In the present work, we focus on investigating neural correlates that originate from the spine and study their connectivity to corresponding signals from the sensorimotor cortex using multivariate autoregressive (MVAR) models. To study cortico-spinal interactions, we simultaneously measured spinal cord potentials (SCPs) and somatosensory evoked potentials (SEPs) of wrist movements elicited by neuromuscular electrical stimulation. We identified directional connections between spine and cortex during both the extension and flexion of the wrist using only non-invasive recording techniques. Our connectivity estimation results are in alignment with various studies investigating correlates of movement, i.e., we found the contralateral side of the sensorimotor cortex to be the main sink of information as well as the spine to be the main source of it. Both types of movement could also be clearly identified in the time-domain signals.

Significance of event related causality (ERC) in eloquent neural networks

Neural activity emerges and propagates swiftly between brain areas. Investigation of these transient large-scale flows requires sophisticated statistical models. We present a method for assessing the statistical confidence of event-related neural propagation. Furthermore, we propose a criterion for statistical model selection, based on both goodness of fit and width of confidence intervals. We show that event-related causality (ERC) with two-dimensional (2D) moving average, is an efficient estimator of task-related neural propagation and that it can be used to determine how different cognitive task demands affect the strength and directionality of neural propagation across human cortical networks. Using electrodes surgically implanted on the surface of the brain for clinical testing prior to epilepsy surgery, we recorded electrocorticographic (ECoG) signals as subjects performed three naming tasks: naming of ambiguous and unambiguous visual objects, and as a contrast, naming to auditory description. ERC revealed robust and statistically significant patterns of high gamma activity propagation, consistent with models of visually and auditorily cued word production. Interestingly, ambiguous visual stimuli elicited more robust propagation from visual to auditory cortices relative to unambiguous stimuli, whereas naming to auditory description elicited propagation in the opposite direction, consistent with recruitment of modalities other than those of the stimulus during object recognition and naming. The new method introduced here is uniquely suitable to both research and clinical applications and can be used to estimate the statistical significance of neural propagation for both cognitive neuroscientific studies and functional brain mapping prior to resective surgery for epilepsy and brain tumors.

Message Passing-Based Inference for Time-Varying Autoregressive Models

Time-varying autoregressive (TVAR) models are widely used for modeling of non-stationary signals. Unfortunately, online joint adaptation of both states and parameters in these models remains a challenge. In this paper, we represent the TVAR model by a factor graph and solve the inference problem by automated message passing-based inference for states and parameters. We derive structured variational update rules for a composite “AR node” with probabilistic observations that can be used as a plug-in module in hierarchical models, for example, to model the time-varying behavior of the hyper-parameters of a time-varying AR model. Our method includes tracking of variational free energy (FE) as a Bayesian measure of TVAR model performance. The proposed methods are verified on a synthetic data set and validated on real-world data from temperature modeling and speech enhancement tasks.

Parasympathetic-Sympathetic Causal Interactions Assessed by Time-Varying Multivariate Autoregressive Modeling of Electrodermal Activity and Heart-Rate-Variability

OBJECTIVE

Most of the bodily functions are regulated by multiple interactions between the parasympathetic (PNS) and sympathetic (SNS) nervous system. In this study, we propose a novel framework to quantify the causal flow of information between PNS and SNS through the analysis of heart rate variability (HRV) and electrodermal activity (EDA) signals.

METHODS

Our method is based on a time-varying (TV) multivariate autoregressive model of EDA and HRV time-series and incorporates physiologically inspired assumptions by estimating the Directed Coherence in a specific frequency range. The statistical significance of the observed interactions is assessed by a bootstrap procedure purposely developed to infer causalities in the presence of both TV model coefficients and TV model residuals (i.e., heteroskedasticity). We tested our method on two different experiments designed to trigger a sympathetic response, i.e., a hand-grip task (HG) and a mental-computation task (MC).

RESULTS

Our results show a parasympathetic driven interaction in the resting state, which is consistent across different studies. The onset of the stressful stimulation triggers a cascade of events characterized by the presence or absence of the PNS-SNS interaction and changes in the directionality. Despite similarities between the results related to the two tasks, we reveal differences in the dynamics of the PNS-SNS interaction, which might reflect different regulatory mechanisms associated with different stressors.

CONCLUSION

We estimate causal coupling between PNS and SNS through MVAR modeling of EDA and HRV time-series.

SIGNIFICANCE

Our results suggest promising future applicability to investigate more complex contexts such as affective and pathological scenarios.

Proposed Requirements for Cardiovascular Imaging-Related Machine Learning Evaluation (PRIME): A Checklist: Reviewed by the American College of Cardiology Healthcare Innovation Council

Machine learning (ML) has been increasingly used within cardiology, particularly in the domain of cardiovascular imaging. Due to the inherent complexity and flexibility of ML algorithms, inconsistencies in the model performance and interpretation may occur. Several review articles have been recently published that introduce the fundamental principles and clinical application of ML for cardiologists. This paper builds on these introductory principles and outlines a more comprehensive list of crucial responsibilities that need to be completed when developing ML models. This paper aims to serve as a scientific foundation to aid investigators, data scientists, authors, editors, and reviewers involved in machine learning research with the intent of uniform reporting of ML investigations. An independent multidisciplinary panel of ML experts, clinicians, and statisticians worked together to review the theoretical rationale underlying 7 sets of requirements that may reduce algorithmic errors and biases. Finally, the paper summarizes a list of reporting items as an itemized checklist that highlights steps for ensuring correct application of ML models and the consistent reporting of model specifications and results. It is expected that the rapid pace of research and development and the increased availability of real-world evidence may require periodic updates to the checklist.