Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The . gov means it’s official. Federal government websites often end in . gov or VSports app下载. mil. Before sharing sensitive information, make sure you’re on a federal government site. .

Https

The site is secure V体育官网. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely. .

. 2025 Sep 25;15(1):32820.
doi: 10.1038/s41598-025-18479-9.

Epileptic seizure detection from electroencephalogram signals based on 1D CNN-LSTM deep learning model using discrete wavelet transform

Homa Kashefi Amiri  1   2 , Masoud Zarei  2 , Mohammad Reza Daliri  3
Affiliations

Epileptic seizure detection from electroencephalogram signals based on 1D CNN-LSTM deep learning model using discrete wavelet transform

Homa Kashefi Amiri et al. Sci Rep. .

Abstract

Excessive electrical activity in the brain causes epileptic seizures which can be detected through Electroencephalogram (EEG) signals. The research aims to identify epileptic seizures using EEG records automatically VSports手机版. Firstly, EEG bands are extracted using Discrete Wavelet Transform (DWT) and concatenated. Secondly, the resulting feature vector is fed into a 1-dimensional Convolutional Neural Network (CNN) to extract spatial information. The Long-Short Term Memory (LSTM) layer then receives the feature maps in order to extract the temporal information. Ultimately, a fully connected layer will use the generated spatiotemporal features as input to categorize the signal. Results show that the suggested model performs well on the following datasets: the TUSZ corpus, which has 94. 32% accuracy, 86. 08% Kappa value, and 79. 01% GDR; the BONN dataset, which has 97. 24% accuracy, 97. 92% Kappa value, and 99. 18% GDR; and the CHB-MIT dataset, which has 96. 94% accuracy, 94. 33% Kappa value, and 96. 36% GDR. The computational complexity for BONN, CHB-MIT, and TUSZ datasets are [Formula: see text], [Formula: see text] and [Formula: see text] respectively. The performance of several popular machine learning classifiers is compared with the proposed model. The results show that the model outperforms existing approaches. The model's strong performance is largely due to the CNN's ability to effectively extract meaningful spatial features. .

Keywords: Convolutional neural network; Discrete wavelet transform; Electroencephalogram; Epileptic seizure detection; Long-short term memory V体育安卓版. .

PubMed Disclaimer

Conflict of interest statement

Declarations. Ethics approval and consent to participate: This study utilizes three publicly available and fully de-identified EEG datasets: the CHB-MIT Scalp EEG Database, the University of Bonn EEG dataset, and the Temple University Hospital Seizure Corpus (TUSZ) V体育ios版. These datasets were collected initially under appropriate institutional ethical approvals and informed consent protocols. As this research involved only secondary analysis of anonymized, publicly accessible data, no additional ethics approval or participant consent was required for this study. Competing interests: The authors declare no competing interests.

VSports app下载 - Figures

Fig. 1
Fig. 1
The Bonn EEG dataset contains five distinct kinds of epileptic EEG signals. A (Healthy, eyes open); B (Healthy, eyes closed); C (Epileptic, seizure-free (opposite hemisphere)); D (Epileptic, seizure-free (epileptogenic zone)), and E (Epileptic, during seizure (ictal) are the sets. The horizontal axis displays the sample number, and the vertical axis shows the amplitude.
Fig. 2
Fig. 2
Raw EEG signal of the healthy subject (above) and epileptic seizure subject (bottom) of the CHB-MIT dataset. The vertical line indicates channels, and the horizontal line indicates time (s). Y-axis labels mean each channel is the difference between two electrodes. For example, channel 1 is the difference between electrodes FP1 and F7.
Fig. 3
Fig. 3
One second snapshot of FNSZ seizure class of TUSZ corpus. The vertical axis shows the Amplitude, and the horizontal axis shows the sample number.
Fig. 4
Fig. 4
(A) EEG signal decomposition with 3-Level DWT. formula imagemeans down-sampling by half. (B) Concatenation of decomposed EEG signal using DWT to create a one-dimensional feature vector as an input for the 1D CNN-LSTM model. formula imagemeans down-sampling by half.
Fig. 5
Fig. 5
The proposed 1D CNN-LSTM model’s block diagram.
Fig. 6
Fig. 6
The internal Structure of the LSTM layer.
Fig. 7
Fig. 7
Bar accuracy plot of the proposed network and other machine learning classifiers on Bonn and CHB-MIT and TUSZ datasets.
Fig. 8
Fig. 8
SHAP summary plot for extracted features using DWT for CHB-MIT dataset.
See this image and copyright information in PMC

References

    1. Farooq, M. S., Zulfiqar, A. & Riaz, S. Epileptic Seizure Detection Using Machine Learning: Taxonomy, Opportunities, and Challenges, Diagnostics, vol. 13, no. 6, p. 1058, Mar. (2023). 10.3390/diagnostics13061058 - PMC - PubMed
    1. Holland, K. Epilepsy: Facts, Statistics, and You. [Online]. Available: https://www.healthline.com/health/epilepsy/facts-statistics-infographic
    1. Chakrabarti, S., Swetapadma, A. & Pattnaik, P. K. A review on epileptic seizure detection and prediction using soft computing techniques, in Smart Techniques for a Smarter Planet, vol. 374, (eds Mishra, M. K., Mishra, B. S. P., Patel, Y. S. & Misra, R.) in Studies in Fuzziness and Soft Computing, vol. 374., Cham: Springer International Publishing, 37–51. doi: 10.1007/978-3-030-03131-2_3. (2019).
    1. Valentinuzzi, M. E. Bioelectrical signal processing in cardiac and neurological applications and electromyography: physiology, engineering, and noninvasive applications. Biomed. Eng. OnLine. 6 (1), 27. 10.1186/1475-925X-6-27 (2007).
    1. Yogarajan, G. et al. EEG-based epileptic seizure detection using binary dragonfly algorithm and deep neural network. Sci. Rep.13 (1), 17710. 10.1038/s41598-023-44318-w (Oct. 2023). - PMC - PubMed