Comparison of Empirical Mode Decomposition, Wavelets, and Different Machine Learning Approaches for Patient-Specific Seizure Detection Using Signal-Derived Empirical Dictionary Approach

Application of empirical mode decomposition, particle swarm optimization, and support vector machine methods to predict stream flows

Modeling stream flows is vital for water resource planning and flood and drought management. In this study, the performance of hybrid models constructed by combining least square support vector machines (LSSVM), empirical model decomposition (EMD), and particle swarm optimization (PSO) methods in modeling monthly streamflow was evaluated. For establishing the models, 42 years of monthly average streamflow data was used in two hydrometer stations located in the Konya Closed Basin, covering 1964 to 2005. Lagged streamflow values ​​were selected as inputs according to partial autocorrelation values ​​in establishing the models. The dataset was divided into 70% training and 30% testing. Model performances were evaluated according to mean square error, root mean square error, correlation coefficients, scatter plot, and Taylor and Violin diagrams. As a result of the analysis, it was determined that the PSO-LSSVM and EMD-LSSVM models were slightly more successful than the single LSSVM model, and the best model was obtained with the EMD-PSO-LSSVM. In addition, in estimating monthly stream flows, 1-, 9-, 10-, 11-, and 12-month lagged streamflow values were the input combination that gave the best results in semi-arid climatic regions. This result demonstrated that EMD improved the performance of both LSSVM and PSO-LSSVM models by 1% to 5% based on correlation coefficient (R) values.

Capturing the power of seizures: an empirical mode decomposition analysis of epileptic activity in the mouse hippocampus

Introduction

Various methods have been used to determine the frequency components of seizures in scalp electroencephalography (EEG) and in intracortical recordings. Most of these methods rely on subjective or trial-and-error criteria for choosing the appropriate bandwidth for filtering the EEG or local field potential (LFP) signals to establish the frequency components that contribute most to the initiation and maintenance of seizure activity. The empirical mode decomposition (EMD) with the Hilbert-Huang transform is an unbiased method to decompose a time and frequency variant signal into its component non-stationary frequencies. The resulting components, i.e., the intrinsic mode functions (IMFs) objectively reflect the various non-stationary frequencies making up the original signal.

Materials and methods

We employed the EMD method to analyze the frequency components and relative power of spontaneous electrographic seizures recorded in the dentate gyri of mice during the epileptogenic period. Epilepsy was induced in mice following status epilepticus induced by suprahippocampal injection of kainic acid. The seizures were recorded as local field potentials (LFP) with electrodes implanted in the dentate gyrus. We analyzed recording segments that included a seizure (mean duration 28 s) and an equivalent time period both before and after the seizure. Each segment was divided into non-overlapping 1 s long epochs which were then analyzed to obtain their IMFs (usually 8-10), the center frequencies of the respective IMF and their spectral root-mean-squared (RMS) power.

Results

Our analysis yielded unbiased identification of the spectral components of seizures, and the relative power of these components during this pathological brain activity. During seizures, the power of the mid frequency components increased while the center frequency of the first IMF (with the highest frequency) dramatically decreased, providing mechanistic insights into how local seizures are generated.

Discussion

We expect this type of analysis to provide further insights into the mechanisms of seizure generation and potentially better seizure detection.

A hybrid unsupervised and supervised learning approach for postictal generalized EEG suppression detection

Sudden unexpected death of epilepsy (SUDEP) is a catastrophic and fatal complication of epilepsy and is the primary cause of mortality in those who have uncontrolled seizures. While several multifactorial processes have been implicated including cardiac, respiratory, autonomic dysfunction leading to arrhythmia, hypoxia, and cessation of cerebral and brainstem function, the mechanisms underlying SUDEP are not completely understood. Postictal generalized electroencephalogram (EEG) suppression (PGES) is a potential risk marker for SUDEP, as studies have shown that prolonged PGES was significantly associated with a higher risk of SUDEP. Automated PGES detection techniques have been developed to efficiently obtain PGES durations for SUDEP risk assessment. However, real-world data recorded in epilepsy monitoring units (EMUs) may contain high-amplitude signals due to physiological artifacts, such as breathing, muscle, and movement artifacts, making it difficult to determine the end of PGES. In this paper, we present a hybrid approach that combines the benefits of unsupervised and supervised learning for PGES detection using multi-channel EEG recordings. A K-means clustering model is leveraged to group EEG recordings with similar artifact features. We introduce a new learning strategy for training a set of random forest (RF) models based on clustering results to improve PGES detection performance. Our approach achieved a 5-second tolerance-based detection accuracy of 64.92%, a 10-second tolerance-based detection accuracy of 79.85%, and an average predicted time distance of 8.26 seconds with 286 EEG recordings using leave-one-out (LOO) cross-validation. The results demonstrated that our hybrid approach provided better performance compared to other existing approaches.