Abstract:Deep learning methods are at the forefront of automated epileptic seizure detection and onset zone localization using scalp-EEG. However, the performance of deep learning methods rely heavily on the quality of annotated training datasets. Scalp EEG is susceptible to high noise levels, which in turn leads to imprecise annotations of the seizure timing and characteristics. This label noise presents a significant challenge in model training and generalization. In this paper, we introduce a novel statistical framework that informs a deep learning model of label ambiguity, thereby enhancing the overall seizure detection performance. Our Bayesian UncertaiNty-aware Deep Learning, BUNDL, strategy offers a straightforward and model-agnostic method for training deep neural networks with noisy training labels that does not add any parameters to existing architectures. By integrating domain knowledge into the statistical framework, we derive a novel KL-divergence-based loss function that capitalizes on uncertainty to better learn seizure characteristics from scalp EEG. Additionally, we explore the impact of improved seizure detection on the task of automated onset zone localization. We validate BUNDL using a comprehensive simulated EEG dataset and two publicly available datasets, TUH and CHB-MIT. BUNDL consistently improves the performance of three base models on simulated data under seven types of label noise and three EEG signal-to-noise ratios. Similar improvements were observed in the real-world TUH and CHB-MIT datasets. Finally, we demonstrate that BUNDL improves the accuracy of seizure onset zone localization. BUNDL is specifically designed to address label ambiguities, enabling the training of reliable and trustworthy models for epilepsy evaluation.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of the impact of annotation noise in electroencephalogram (EEG) data of epilepsy on the training and generalization performance of deep - learning models. Specifically, the author points out: 1. **Challenges of annotation noise**: - EEG data is easily affected by high noise levels, resulting in inaccurate annotation of seizure time and characteristics. - This "label noise" poses a significant challenge to model training and generalization. 2. **Limitations of existing methods**: - Most of the current deep - learning models for epilepsy detection rely on the supervised learning paradigm, assuming that the provided "true labels" are accurate. However, the data annotated by clinicians may have manual errors and low inter - rater consistency (about 60%). - Ignoring these label uncertainties may cause the model to learn misleading features, thus affecting its generalization ability. 3. **Proposed solution**: - The author introduces a new statistical framework - **Bayesian Uncertainty - aware Deep Learning (BUNDL)** to inform the deep - learning model of label uncertainties, thereby improving the overall performance of epilepsy detection. - BUNDL can handle instance - dependent label noise through Bayesian modeling and optimize the existing deep network architecture without any parameter overhead. - This method uses Monte Carlo dropout to estimate uncertainties and adjusts the posterior probability of label flipping to make predictions more reliable. 4. **Experimental verification**: - The author uses a comprehensive simulated EEG data set and two publicly available real - world data sets (Temple University Hospital, TUH and Boston Children’s Hospital, CHB - MIT) to verify the effectiveness of BUNDL. - The results show that BUNDL can significantly improve the performance of three benchmark models under various types of label noise and different signal - to - noise ratios. 5. **Application prospects**: - BUNDL not only improves the accuracy of epilepsy seizure detection but also improves the performance of the task of automatically locating the seizure onset area. - This method can be seamlessly integrated into existing deep - learning models and is suitable for complex EEG data analysis in clinical practice. In conclusion, this paper is committed to developing a robust deep - learning framework that can deal with annotation noise in EEG data to improve the reliability and credibility of epilepsy detection and analysis.

BUNDL: Bayesian Uncertainty-aware Deep Learning with Noisy training Labels for Seizure Detection in EEG

Novel deep learning framework for detection of epileptic seizures using EEG signals

Delving into α-stable distribution in noise suppression for seizure detection from scalp EEG.

Estimating Patient-Level Uncertainty in Seizure Detection Using Group-Specific Out-of-Distribution Detection Technique

Incorporating Hand-crafted Features to Deep Neural Networks for Seizure Prediction

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

From Epilepsy Seizures Classification to Detection: A Deep Learning-based Approach for Raw EEG Signals

Epileptic Seizure Detection by Cascading Isolation Forest-Based Anomaly Screening and EasyEnsemble

End-to-end model for automatic seizure detection using supervised contrastive learning

A novel end-to-end approach for epileptic seizure classification from scalp EEG data using deep learning technique

Semisupervised Seizure Prediction in Scalp EEG Using Consistency Regularization

Residual Deep Convolutional Neural Network for EEG Signal Classification in Epilepsy

A deep learning framework for epileptic seizure detection based on neonatal EEG signals

Towards trustworthy seizure onset detection using workflow notes

Unsupervised EEG-Based Seizure Anomaly Detection with Denoising Diffusion Probabilistic Models

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

A DEEP LEARNING FRAMEWORK TO CHARACTERIZE NOISY LABELS IN EPILEPTOGENIC ZONE LOCALIZATION USING FUNCTIONAL CONNECTIVITY

Automatic Seizure Detection Using Multi-Input Deep Feature Learning Networks for EEG Signals

Benchmarking the Impact of Noise on Deep Learning-based Classification of Atrial Fibrillation in 12-Lead ECG

Deep extreme learning machine with knowledge augmentation for EEG seizure signal recognition