Related Experiment Video
Updated: Aug 15, 2026
07:47
A Protocol For Uncovering Neural Mechanisms Of Neurotherapeutic Effects On Electroencephalography Using The Human Neocortical Neurosolver
Published on: May 19, 2026
Leednet: a lightweight network for event detection in EEG signals
Mohammad Ali Alqarni1, Hira Masood2,3, Hassan Aqeel Khan4
1College of Computer Science and Engineering, University of Jeddah, Jeddah, Saudi Arabia.
Journal on Advances in Signal Processing
|August 14, 2026
Summary
A new lightweight neural network, LEEDNet, efficiently detects critical events in EEG recordings. It achieves high accuracy using raw EEG data and a 2-second window, making it ideal for resource-limited healthcare settings.
Area of Science:
- Computational Neuroscience
- Medical Signal Processing
- Artificial Intelligence in Healthcare
Background:
- Analyzing long-duration EEG recordings for diagnostically relevant events demands significant computational resources.
- This computational load hinders the deployment of EEG triage and decision support systems, particularly in resource-constrained environments.
- Efficient processing is crucial for real-time clinical applications and field deployments.
Purpose of the Study:
- To develop and evaluate a computationally efficient method for detecting critical events in EEG.
- To explore the impact of architecture, decision window length, and feature extraction on computational efficiency and performance.
- To optimize EEG event detection for deployment in resource-limited healthcare systems.
Main Methods:
- Development of LEEDNet (Lightweight EEG Event Detection Network), a novel convolutional neural network architecture.
- Evaluation of LEEDNet on the NMT-Events dataset using subject-wise splits and comparison against four state-of-the-art baselines.
- Ablation studies to determine optimal decision window length (2s) and analysis of different input modalities (raw EEG, FFT spectra, wavelets).
Main Results:
- LEEDNet achieved superior performance on three out of four evaluation metrics compared to baselines, with significantly fewer parameters (0.25M) and FLOPs (37M).
- Inference time was less than 5 ms on a standard CPU, demonstrating high computational efficiency.
- A 2-second decision window and raw EEG data as input yielded the best performance (81.2% accuracy, 77.5% macro-F1), outperforming complex feature extraction methods.
Conclusions:
- LEEDNet offers a highly efficient and accurate solution for real-time EEG event detection.
- A data-driven approach using raw EEG waveforms is more effective and computationally advantageous than handcrafted feature engineering for this architecture.
- The developed method is well-suited for deployment in resource-constrained healthcare settings, enabling advanced EEG analysis.
