Abstract:Over the past few decades, electroencephalography (EEG) monitoring has become a pivotal tool for diagnosing neurological disorders, particularly for detecting seizures. Epilepsy, one of the most prevalent neurological diseases worldwide, affects approximately the 1 \% of the population. These patients face significant risks, underscoring the need for reliable, continuous seizure monitoring in daily life. Most of the techniques discussed in the literature rely on supervised Machine Learning (ML) methods. However, the challenge of accurately labeling variations in epileptic EEG waveforms complicates the use of these approaches. Additionally, the rarity of ictal events introduces an high imbalancing within the data, which could lead to poor prediction performance in supervised learning approaches. Instead, a semi-supervised approach allows to train the model only on data not containing seizures, thus avoiding the issues related to the data imbalancing. This work proposes a semi-supervised approach for detecting epileptic seizures from EEG data, utilizing a novel Deep Learning-based method called SincVAE. This proposal incorporates the learning of an ad-hoc array of bandpass filter as a first layer of a Variational Autoencoder (VAE), potentially eliminating the preprocessing stage where informative band frequencies are identified and isolated. Results indicate that SincVAE improves seizure detection in EEG data and is capable of identifying early seizures during the preictal stage as well as monitoring patients throughout the postictal stage.

What problem does this paper attempt to address?

The main problem this paper attempts to address is the improvement of anomaly detection in electroencephalogram (EEG) data, particularly the detection of epileptic seizures. Specifically, the authors propose a novel deep learning method—SincVAE, aimed at overcoming the challenges faced by existing supervised learning methods when dealing with epileptic EEG data, such as low label accuracy and data imbalance. SincVAE integrates SincNet into the Variational Autoencoder (VAE) framework as the first layer to process EEG data, thereby enabling seizure detection in a semi-supervised setting. ### Main Problems 1. **Low Label Accuracy**: Existing supervised learning methods rely on accurately labeled epileptic EEG waveforms, but this labeling process is complex and prone to errors. 2. **Data Imbalance**: Epileptic seizure events are relatively rare, leading to a severe imbalance between normal and abnormal data in the dataset, which affects the predictive performance of supervised learning methods. 3. **Cumbersome Preprocessing Steps**: Traditional EEG data processing often requires manual selection and isolation of information-rich frequency bands, a time-consuming and error-prone step. ### Solution - **SincVAE Model**: Combines SincNet and Variational Autoencoder (VAE), capable of automatically learning and applying custom band-pass filters, thereby reducing or eliminating preprocessing steps. - **Semi-Supervised Learning**: The model is trained only on data without seizures, avoiding the issues brought by data imbalance. - **Early Detection and Monitoring**: SincVAE can not only detect seizures but also monitor preictal and postictal stages, improving overall detection accuracy. ### Experimental Evaluation - **Datasets**: Used the Bonn dataset and CHB-MIT dataset, which are widely used in epilepsy seizure detection research. - **Experimental Design**: Employed a semi-supervised learning approach, training the model only on non-seizure data, and classified by analyzing the reconstruction error generated by the model. - **Comparison Methods**: Compared with traditional VAE models to verify the superiority of SincVAE in anomaly detection. ### Conclusion By introducing SincNet as the first layer of VAE, SincVAE improves the detection performance of epileptic seizures in EEG data, particularly in handling data imbalance and reducing preprocessing steps. This method is expected to play a significant role in the healthcare field, especially in the daily monitoring of epilepsy patients.

SincVAE: a New Approach to Improve Anomaly Detection on EEG Data Using SincNet and Variational Autoencoder

Automatic Epileptic Seizures Joint Detection Algorithm Based on Improved Multi-Domain Feature of cEEG and Spike Feature of aEEG.

Real-Time Epileptic Seizure Detection Based on Deep Learning.

Semi-supervised automatic seizure detection using personalized anomaly detecting variational autoencoder with behind-the-ear EEG

hvEEGNet: exploiting hierarchical VAEs on EEG data for neuroscience applications

VAEEG: Variational Auto-encoder for Extracting EEG Representation

A sustainable artificial-intelligence-augmented digital care pathway for epilepsy: Automating seizure tracking based on electroencephalogram data using artificial intelligence

Robust Detection of Brain Stimulation Artifacts in iEEG Using Autoencoder-Generated Signals and ResNet Classification

vEEGNet: learning latent representations to reconstruct EEG raw data via variational autoencoders

A novel epilepsy detection approach using intrinsic multiscale entropy analysis and DSEAM-enhanced 1D-ResNets

A Multi-View Deep Learning Framework for EEG Seizure Detection

Detection of Epileptic Seizures using Unsupervised Learning Techniques for Feature Extraction

A Multi-view Deep Learning Method for Epileptic Seizure Detection Using Short-time Fourier Transform

Deep Classification of Epileptic Signals

A novel universal deep learning approach for accurate detection of epilepsy

Effective Detection of Epileptic Seizures through EEG Signals Using Deep Learning Approaches

Advanced framework for epilepsy detection through image-based EEG signal analysis

Automatic detection of epileptic seizure based on one dimensional cascaded convolutional autoencoder with adaptive window-thresholding

Epileptic Seizure Detection Using Brain-Rhythmic Recurrence Biomarkers and ONASNet-Based Transfer Learning

Parallel Ictal-Net, a Parallel CNN Architecture with Efficient Channel Attention for Seizure Detection

Adversarial Representation Learning for Robust Patient-Independent Epileptic Seizure Detection