Deep Learning for Interictal Epileptiform Spike Detection from scalp EEG frequency sub bands
Summary
This study introduces a deep learning approach using convolutional neural networks (CNNs) to automatically detect interictal epileptiform discharges (IEDs) in electroencephalogram (EEG) signals, improving epilepsy diagnosis.
Area of Science:
- Clinical Neuroscience
- Biomedical Engineering
- Artificial Intelligence in Medicine
Background:
- Epilepsy diagnosis relies on identifying interictal epileptiform discharges (IEDs) in electroencephalogram (EEG) signals, a process often challenging and time-consuming for clinicians.
- Automated detection of IEDs using deep learning offers a potential solution to improve diagnostic efficiency and accuracy.
Purpose of the Study:
- To develop and evaluate a novel convolutional neural network (CNN) based method for the automatic detection of IEDs in scalp EEG signals.
- To assess the performance of the proposed CNN model using a large, expert-annotated EEG dataset and external clinical data.
Main Methods:
- A 1D CNN model was designed, utilizing a combination of raw EEG data and its frequency sub-bands (delta, theta, alpha, beta) as input features.
- The model was trained and validated using five-fold cross-validation on a dataset of 554 scalp EEGs containing 18,164 expertly marked IEDs.
- Performance was evaluated using metrics such as sensitivity, false positive rate per minute, and precision, and further validated on external datasets from three different clinics.
Main Results:
- The 1D CNN detector achieved a false positive rate of 0.23 per minute and a precision of 0.79 at 90% sensitivity.
- Features extracted from the CNN outputs demonstrated significant discrimination (p < 0.05) between EEGs with and without IEDs across multiple clinical datasets.
- The proposed method showed superior performance compared to existing approaches in the literature.
Conclusions:
- The developed CNN-based system provides an effective and optimized method for automatic IED detection in EEG signals.
- This automated approach has the potential to assist clinicians in expediting epilepsy diagnosis and informing treatment strategies.
- The study highlights the efficacy of deep learning in analyzing complex neurophysiological signals for clinical applications.
