Mateusz Kunik, Artur Gramacki (2025)

Deep learning epileptic seizure detection based on the matching pursuit algorithm and its time-frequency graphical representation

---

Electroencephalography (EEG) is the primary diagnostic and an important prognostic clinical tool for epilepsy. The detection of epileptic activity is usually performed by a human expert and is based on finding specific patterns in the multi-channel electroencephalogram. However, the manual inspection of EEG signals is a time-consuming procedure for neurologists. Therefore, various attempts are made to automate it using both conventional and deep learning techniques. In this article, (i) we investigate the possibility of using time-frequency maps of energy derived from the matching pursuit algorithm for accurate detection of epileptic seizures (to the best of our knowledge, such an approach has not been analyzed so far, making this a pilot study); (ii) we show how to build an effective deep convolutional neural network with the so-called (2+1)D convolution technique; (iii) using carefully selected 79 neonatal EEG recordings, we develop a complete framework for seizure detection employing a deep learning approach, (iv) we share a ready to use R and Python codes which allow reproducing all the results presented in the paper.

DOI: 10.61822/amcs-2025-0044