International Journal of Innovative Research in Computer and Communication Engineering
ISSN Approved Journal | Impact factor: 8.771 | ESTD: 2013 | Follows UGC CARE Journal Norms and Guidelines
| Monthly, Peer-Reviewed, Refereed, Scholarly, Multidisciplinary and Open Access Journal | High Impact Factor 8.771 (Calculated by Google Scholar and Semantic Scholar | AI-Powered Research Tool | Indexing in all Major Database & Metadata, Citation Generator | Digital Object Identifier (DOI) |
| TITLE | Hybrid CNN-LSTM Framework for Harmful Brain Activity Detection Using EEG Signals |
|---|---|
| ABSTRACT | EEG is a primary modality for detecting harmful neurological patterns such as seizures and cortical dysregulation; however, continuous manual examination of EEG recordings is time-consuming and susceptible to variability. To remedy this limitation, this study proposes HMS, an automated harmful EEG activity detection framework built upon a hybrid CNN-LSTM architecture designed for efficient classification of EEG signals. It performs a comprehensive processing pipeline that includes artifact-mitigation filtering, frequency-domain feature extraction, epoch segmentation, and deep spatiotemporal feature learning. Afterwards, in order to increase the SNR and decrease interference, a 0.5–45 Hz Butterworth bandpass filter was first applied to all segments, and Z-score normalization was used. The resulting dataset was comprised of balanced harmful and non-harmful EEG epochs. In the framework of the deep learning model, CNN layers extracted spatial and spectral representations related to epileptiform patterns, while LSTM components captured temporal dependencies that reflect state transitions among normal, pre-ictal, and ictal-like states. The experiments showed that the proposed framework achieved an accuracy of 92%, a precision of 94.41%, a recall of 92%, and an F1-score of 92.48%, demonstrating a strong capability in distinguishing harmful from non-harmful EEG segments with reduced misclassifications. A front-end visualization interface was integrated in order to represent the EEG signals with automated predictions; these allow a better interpretation of the model's outcomes. Future extensions of the framework will explore explainable AI techniques for improved transparency, expansion toward multi-class seizure-type categorization, and broader experimental validation across diverse datasets for enhanced robustness, interpretability, and overall generalizability of harmful EEG activity detection. |
| AUTHOR | PRIYANKA A S, N SUSHMA, SUSHMITHA S, PREETHAM P, D R NAGAMANI UG Student, Dept. of CSE, Bangalore Institute of Technology, VTU, Bengaluru, Karnataka, India Assistant Professor, Dept. of CSE, Bangalore Institute of Technology, VTU, Bengaluru, Karnataka, India |
| VOLUME | 177 |
| DOI | DOI: 10.15680/IJIRCCE.2025.1312110 |
| pdf/110_Hybrid CNN-LSTM Framework for Harmful Brain Activity Detection Using EEG Signals.pdf | |
| KEYWORDS | |
| References | [1] A. Roy, A. Ghosh, and A. K. Saha, “Deep convolutional neural network method to detect seizures and characteristic frequencies using epileptic electroencephalogram (EEG) data,” Procedia Computer Science, vol. 167, pp. 2101–2110, 2020. [2] A. L. Goldberger, L. A. N. Amaral, L. Glass et al., “Behavioral classification of sequential neural activity using time-varying recurrent neural networks,” Neural Computation, vol. 25, no. 3, pp. 732–757, 2013. [3] J. Khosrowabadi and C. Quek, “Brain-Shapelet: A framework for capturing instantaneous abnormalities in brain activity for autism spectrum disorder diagnosis,” Brain Informatics, vol. 6, no. 14, pp. 1–13, 2019. [4] E. Aghabozorgi, Y. Wang, and P. Lu, “Decoding inter-individual relations from spatial similarity of brain activity,” IEEE Trans. Neural Systems and Rehabilitation Engineering, vol. 28, no. 7, pp. 1521–1530, 2020. [5] F. Meng, D. Zhang, Q. Chen, and X. Zhang, “Decoding multi-brain motor imagery from EEG using coupling feature extraction and few-shot learning,” IEEE Trans. Neural Systems and Rehabilitation Engineering, vol. 30, pp. 318–328, 2022. [6] S. Hassan, M. Ahmad, and R. Naqvi, “Decoding spontaneous brain activity from fMRI using Gaussian processes: Tracking brain reactivation,” NeuroImage, vol. 223, p. 117355, 2020. [7] M. Sharma and U. R. Acharya, “Deep C-LSTM neural network for epileptic seizure and tumor detection using high-dimension EEG signals,” Computer Methods and Programs in Biomedicine, vol. 187, p. 105305, 2020. [8] A. Shahin and P. A. Dutilleul, “EEGG: An analytic brain-computer interface algorithm,” IEEE Access, vol. 7, pp. 160759–160771, 2019. [9] J. Lee, M. Kim, S. Jin et al., “Impact of generation rate of speech imagery on neural activity and BCI decoding performance: An fNIRS study,” IEEE J. Biomedical and Health Informatics, vol. 24, no. 7, pp. 1955–1964, 2020. [10] Y. Chen, X. Liu, H. Qiu, and Z. He, “Interpretable dynamic brain network analysis with functional and structural priors,” IEEE Trans. Medical Imaging, vol. 41, no. 4, pp. 899–910, 2022. [11] Z. Zhang and W. Wang, “Joint sparse representation of brain activity patterns related to perceptual and cognitive components of a speech comprehension task,” Neurocomputing, vol. 435, pp. 141–154, 2021. [12] X. Li, K. Wang, and R. Zhang, “Prediction of human brain activity using likelihood-ratio based score fusion,” IEEE Access, vol. 8, pp. 228260–228270, 2020. [13] L. Deng et al., “Temporal dynamic synchronous functional brain network for schizophrenia classification and lateralization analysis,” IEEE Trans. Affective Computing, vol. 12, no. 3, pp. 783–794, 2021. [14] Project Dataset Source (internal clinical dataset, anonymized EEG recordings), 2025. |
