Abstract:Brain Electroencephalography (EEG) classification is widely applied to analyze cerebral diseases in recent years. Unfortunately, invalid/noisy EEGs degrade the diagnosis performance and most previously developed methods ignore the necessity of EEG selection for classification. To this end, this paper proposes a novel maximum weight clique-based EEG selection approach, named mwcEEGs, to map EEG selection to searching maximum similarity-weighted cliques from an improved Fréchet distance-weighted undirected EEG graph simultaneously considering edge weights and vertex weights. Our mwcEEGs improves the classification performance by selecting intra-clique pairwise similar and inter-clique discriminative EEGs with similarity threshold $\delta$. Experimental results demonstrate the algorithm effectiveness compared with the state-of-the-art time series selection algorithms on real-world EEG datasets.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the impact of invalid or noisy electroencephalogram (EEG) on diagnostic performance in EEG classification. Specifically, invalid or noisy EEG will reduce the discrimination of target features, thereby affecting diagnostic accuracy. Most of the existing methods ignore the importance of EEG selection and directly perform subsequent analyses such as denoising, feature extraction and classification. Therefore, this paper proposes an EEG selection method based on maximum weight clique (MWC), named mwcEEGs, aiming to select EEGs with high intra - class similarity and strong inter - class discrimination by searching for the maximum similarity - weighted clique in the improved Fréchet distance - weighted undirected EEG graph, thereby improving classification performance. ### Main contributions 1. **Proposed a new EEG selection method mwcEEGs**: This method maps the EEG selection problem to searching for the maximum weight clique in the similarity - weighted EEG graph, considering both edge weights and vertex weights. 2. **Verified the effectiveness of mwcEEGs**: Through detailed experiments using multiple popular and the latest classifiers on real - world EEG datasets, it is proved that mwcEEGs is superior to existing time - series selection methods. ### Method overview - **Edge weight**: Use the improved Fréchet distance (IFD) to calculate the similarity between EEGs as the edge weight. - **Vertex weight**: Represents the importance of the vertex in the potential maximum similarity - weighted clique, which is determined by calculating the average similarity of each EEG with other EEGs. - **Maximum weight clique search**: By setting the similarity threshold δ, select EEGs with high intra - class similarity and strong inter - class discrimination to form the maximum weight clique. ### Experimental results - **Dataset**: Used the slow cortical potential (SCPs) dataset provided by the Institute of Medical Psychology and Behavioral Neurobiology of Tübingen University. - **Evaluation metrics**: Used metrics such as Rand index (RI), F - score and Fleiss' kappa to evaluate classification performance. - **Baseline methods**: Compared with several state - of - the - art EEG time - series selection methods (such as lwEEGs, gwEEGs, lrtEEGs and grtEEGs). ### Parameter setting - **Similarity threshold δ**: Select the optimal δ value according to the discrete probability distribution of EEG similarity to balance the discrimination and universality of the selected EEGs. Through these methods and experiments, the paper demonstrates the effectiveness and superiority of mwcEEGs in improving EEG classification performance.

Brain EEG Time Series Selection: A Novel Graph-Based Approach for Classification

Assisting Schizophrenia Diagnosis Using Clinical Electroencephalography and Interpretable Graph Neural Networks: a Real-World and Cross-Site Study

Classify EEG and Reveal Latent Graph Structure with Spatio-Temporal Graph Convolutional Neural Network

Toward Optimal Feature and Time Segment Selection by Divergence Method for EEG Signals Classification

Deep Convolutional Neural Network-Based Epileptic Electroencephalogram (EEG) Signal Classification

Time-Series Anomaly Detection Based on Dynamic Temporal Graph Convolutional Network for Epilepsy Diagnosis

Two Heads are Better than One: A Bio-inspired Method for Improving Classification on EEG-ET Data

An EEG based real-time epilepsy seizure detection approach using discrete wavelet transform and machine learning methods

Learning EEG topographical representation for classification via convolutional neural network

ssvEEGc: An efficient EEG clustering method using balance-constrained voting mechanism

Dynamic Frequency Feature Selection Based Approach for Classification of Motor Imageries.

Clustering Brain Signals: A Robust Approach Using Functional Data Ranking

The Random Electrode Selection Ensemble For Eeg Signal Classification

Classification of EEG Using Adaptive SVM Classifier with CSP and Online Recursive Independent Component Analysis

EEG Signal Epilepsy Detection With a Weighted Neighbor Graph Representation and Two-Stream Graph-Based Framework

Classification of EEG signals based on time-frequency analysis and spiking neural network

Ensemble Fusion Models Using Various Strategies and Machine Learning for EEG Classification

Unsupervised Multivariate Feature-Based Adaptive Clustering Analysis of Epileptic EEG Signals

Effective and Extensible Feature Extraction Method Using Genetic Algorithm-Based Frequency-Domain Feature Search for Epileptic EEG Multi-classification

Markov Chain-Guided Graph Construction and Sampling Depth Optimization for EEG-Based Mental Disorder Detection

Hierarchy Graph Convolution Network and Tree Classification for Epileptic Detection on Electroencephalography Signals