Epileptic Seizure Type Detection Model Using CNN-Derived Features and Random Forests

Abstract

Epilepsy is a chronic neurological disorder characterized by recurrent and unpredictable seizures, affecting nearly 1% of the global population and profoundly impacting daily life and overall well-being. Accurate detection and classification of seizures are essential for effective patient management, as traditional EEG-based monitoring relies heavily on manual interpretation, which is time-consuming and subject to variability. This study explores computational approaches for both seizure detection and type classification using EEG data. Two complementary strategies are examined: a feature-based machine learning approach, which extracts statistical, temporal, and spectral features to capture distinct seizure patterns, and a deep learning approach that analyses spectrogram representations of EEG signals, allowing for the modelling of complex temporal-frequency interactions associated with different seizure types. These methods incorporate rigorous preprocessing, including artifact removal, normalization, and dimensionality reduction, to ensure high-quality input for model training and to improve interpretability. In addition, feature importance analysis and visual representations of EEG activity provide insights into the distinguishing characteristics of various seizure types, facilitating clinical understanding and potential application. By combining feature-based and image-based modelling, the study demonstrates a flexible and scalable framework for automated seizure detection and classification, offering a non-invasive solution capable of supporting patient-specific clinical decisions. The findings underscore the potential of intelligent computational techniques to enhance monitoring, diagnosis, and management of epilepsy, paving the way for personalized healthcare systems and improved patient outcomes.

Introduction

This study develops an automated EEG-based seizure detection system using two complementary datasets: a controlled dataset from ten epileptic patients in New Delhi, India, and the large-scale, clinically annotated TUSZ corpus from Philadelphia, USA. The New Delhi dataset provides short, well-segmented recordings into pre-ictal, interictal, and ictal phases, while TUSZ offers over 500 hours of multi-patient EEG data covering various seizure types, ensuring both precision and real-world variability.

The system combines feature-based machine learning (Random Forest) and deep learning (CNN on spectrograms) to capture both explicit signal characteristics and complex temporal–spectral patterns. Preprocessing removes artifacts, and time-, frequency-, and time–frequency features are extracted for robust classification. By integrating controlled and clinical datasets, the framework ensures generalizable, patient-specific, and clinically applicable seizure detection, capable of accurately distinguishing multiple seizure types and supporting real-time monitoring.

Conclusion

This research presents a comprehensive framework for automated seizure type detection using EEG signals, effectively combining empirical signal decomposition, multi-domain feature extraction, and hybrid modelling approaches. By integrating both Convolutional Neural Networks (CNNs) and Random Forest (RF) classifiers, the study bridges the strengths of deep learning’s pattern recognition capabilities with the interpretability and robustness of machine learning. The CNN model, trained on time–frequency spectrograms, achieved an accuracy of 93.75%, successfully capturing complex spectral–temporal structures characteristic of seizure activity. In comparison, the RF model attained an even higher accuracy of 95.42%, driven by its ability to leverage well-engineered features such as spectral entropy, Hjorth complexity, and wavelet energy. The results highlight that both models effectively distinguish between normal and seizure EEG signals, as well as across different seizure types, demonstrating strong generalization on both controlled and large-scale datasets. The visual analyses including confusion matrices, OOB feature importance plots, and classification outputs further validated the models’ reliability and interpretability. While the CNN excels at learning intricate data-driven representations, the RF offers clarity in understanding which physiological signal properties contribute most to accurate classification. Together, they form a complementary system that balances precision, transparency, and adaptability. Overall, this study underscores the potential of hybrid EEG-based computational intelligence systems in advancing clinical epilepsy diagnosis and monitoring. The methodology’s flexibility allows it to adapt to diverse EEG datasets and seizure types, making it a viable candidate for real-time implementation in hospital and wearable neuro-monitoring applications. Future work may explore integrating patient-specific adaptation, multimodal bio-signals, and lightweight architectures to enhance scalability and real-world usability.

References

[1] Santushti Santosh Betgeri, Madhu Shukla, Dinesh Kumar, Surbhi B. Khan, Muhammad Attique Khan, Nora A. Alkhaldi, “Enhancing seizure detection with hybrid XGBoost and recurrent neural networks”, Neuroscience Informatics, Volume 5, Issue 2, 2025, 100206, ISSN 2772-5286, https://doi.org/10.1016/j.neuri.2025.100206. [2] AUTHOR=Shah Vinit , von Weltin Eva , Lopez Silvia , McHugh James Riley , Veloso Lillian , Golmohammadi Meysam , Obeid Iyad , Picone Joseph ,”The Temple University Hospital Seizure Detection Corpus” ,Frontiers in Neuroinformatics ,Volume 12 – 2018,2018, DOI=10.3389/fninf.2018.00083,ISSN=1662-5196 [3] M. N. A. Tawhid, S. Siuly and T. Li, \"A Convolutional Long Short-Term Memory-Based Neural Network for Epilepsy Detection From EEG,\" in IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1-11, 2022, Art no. 4010211, doi: 10.1109/TIM.2022.3217515. [4] P. Boonyakitanont, A. Lek-Uthai, K. Chomtho, and J. Songsiri, “A review of feature extraction and performance evaluation in epileptic seizure detection using EEG,” Biomedical Signal Processing and Control, vol. 57, pp. 1–28, 2020. [5] K. Rasheed, A. Qayyum, J. Qadir, S. Sivathamboo, and P. Kwan, “Machine learning for predicting epileptic seizures using EEG signals: a review,” IEEE Reviews in Biomedical Engineering, vol. 14, pp. 139–155, 2020. [6] S. Toraman, “Preictal and interictal recognition for epileptic seizure prediction using pre-trained 2D-CNN models,” Traitement du Signal, vol. 37, no. 6, pp. 1045–1054, 2020. [7] B. Behnoush, E. Bazmi, S. H. Nazari, S. Khodakarim, M. A.Looha, and H. Soori, “Machine learning algorithms to predict seizure due to acute tramadol poisoning,” Human & Experimental Toxicology, vol. 40, no. 8, pp. 1225–1233, 2021. [8] Y. S. Chandu and S. Fathimabi, “Epilepsy prediction using deep learning,” International Journal of Engineering Research & Technology (IJERT), vol. 9, no. 5, pp. 211–219, 2021 [9] M. Wu, H. Qin, X. Wan, and Y. Du, “HFO detection in epilepsy: a stacked denoising autoencoder and sample weight adjusting factors-based method,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 29, pp. 1965–1976, 2021. [10] T. Iešmantas and R. Alzbutas, “Convolutional neural network for detection and classi?cation of seizures in clinical data,” Medical & Biological Engineering & Computing, vol. 58 no. 9, pp. 1919–1932, 2020. [11] M. K. Siddiqui, R. Morales-Menendez, X. Huang, and N. Hussain, ‘‘A review of epileptic seizure detection using machine learning classifiers,’’ Brain Informatics , vol. 7, no. 1, p. 5, May 2020. [12] K.Rasheed, A.Qayyum ,J.Qadir, S.Sivathamboo ,P.Kwan, L.Kuhlmann, T. O’Brien, and A. Razi, ‘‘Machine learning for predicting epileptic seizures using EEG signals: A review,’’ IEEE Review in Biomedical Engineering, vol. 14, pp. 139–155, 2020. [13] A. Shoeibi, M. Khodatars, N. Ghassemi, M. Jafari, P. Moridian, R. Alizadehsani, M. Panahiazar, F. Khozeimeh, A. Zare, H. Hosseini-Nejad, A. Khosravi, A. F. Atiya, D. Aminshahidi, S. Hussain, M. Rouhani, S. Nahavandi, and U. R. Acharya, ‘‘Epileptic seizures detection using deep learning techniques: A review,’’ International Journal of Environmental Research on Public Health, vol. 18, no. 11, p. 5780, May 2021. [14] W.Lin,Q.Chen,M.Jiang,J.Tao,Z.Liu,X.Zhang,L.Wu,S.Xu,Y. Kang, and Q. Zeng, ‘‘Sitting or walking? Analyzing the neural emotional indica tors of urban green space behavior with mobile EEG,’’ Journal of Urban Health, vol. 97, no. 2, pp. 191–203, Apr. 2020. [15] E. Q. Wu, ‘‘Detecting fatigue status of pilots based on deep learning network using EEG signals,’’ IEEE Transactions on Cognitive Developmental Systems, vol. 13, no. 3, pp. 575–585, Sep. 2021. [16] V. R. Carvalho, M. F. D. Moraes, A. P. Braga, and E. M. A. M. Mendes, ‘‘Evaluating five different adaptive decomposition methods for EEG signal seizure detection and classification,’’ Biomedical Signal Processing and Control, vol. 62, Sep. 2020, Art. no. 102073. [17] M. Chen, X. Shi, Y. Zhang, D. Wu, and M. Guizani, ‘‘Deep feature learning for medical image analysis with convolutional autoencoder neural network,’’ IEEE Transactions on Big Data, vol. 7, no. 4, pp. 750–758, Oct. 2021. [18] S. Chakrabarti, A. Swetapadma, A. Ranjan, and P. K. Pattnaik, ‘‘Time domain implementation of pediatric epileptic seizure detection system for enhancing the performance of detection and easy monitoring of pediatric patients,’’ Biomedical Signal Processing and Control, vol. 59, May 2020, Art. no. 101930. [19] R. Sari?, D. Joki?, N. Beganovi?, L. G. Pokvi?, and A. Badnjevi?, ‘‘FPGA-based real-time epileptic seizure classification using artificial neural network,’’ Biomedical Signal Processing and Control, vol. 62, Sep. 2020, Art. no. 102106. [20] K. Zhang, N. Robinson, S.-W. Lee, and C. Guan, ‘‘Adaptive transfer learning for EEG motor imagery classification with deep convolutional neural network,’’ Neural Networks, vol. 136, pp. 1–10, Apr. 2021. [21] G. Placidi, L. Cinque, and M. Polsinelli, ‘‘A fast and scalable framework for automated artifact recognition from EEG signals represented in scalp topographies of independent components,’’ Computers in Biology and Medicine, vol. 132, May 2021, Art. no. 104347. [22] D. Borra, S. Fantozzi, and E. Magosso, ‘‘Interpretable and lightweight convolutional neural network for EEG decoding: Application to movement execution and imagination,’’ Neural Networks, vol. 129, pp. 55–74, Sep. 2020. [23] R. Y. Aminabadi, O. Ruwase, M. Zhang, Y. He, J.-M. Arnau, and A. Gonazalez, ‘‘SHARP: An adaptable, energy-efficient accelerator for recurrent neural networks,’’ ACM Transactions on Embedded Computing Systems, vol. 22, no. 2, pp. 1–23, Mar. 2023. [24] G. Paulin, F. Conti, L. Cavigelli, and L. Benini, ‘‘Vau da muntanialas: Energy-efficient multi-die scalable acceleration of RNN inference,’’ IEEE Transactions on Circuits and Systems I, Reg. Papers, vol. 69, no. 1, pp. 244–257, Jan. 2022. [25] G. Muhammad, M. S. Hossain, and N. Kumar, ‘‘EEG-based pathology detection for home health monitoring,’’ IEEE Journal on Selected Areas in Communications, vol. 39, no. 2, pp. 603–610, Feb. 2021. [26] A.Kurapa, D. Rathore, D. R. Edla, A. Bablani, and V. Kuppili, ‘‘A hybrid approach for extracting EMG signals by filtering EEG data for IoT applications for immobile persons,’’ Wireless Personal Communications, vol. 114, no. 4, pp. 3081–3101, Oct. 2020. [27] T.Siddharth, P. Gajbhiye, R. K. Tripathy, and R. B. Pachori, ‘‘EEG-based detection of focal seizure area using FBSE-EWT rhythm and SAE-SVM network,’’ IEEE Sensors Journal, vol. 20, no. 19, pp. 11421–11428, Oct. 2020. [28] M. Sameer and B. Gupta, ‘‘Detection of epileptical seizures based on alpha band statistical features,’’ Wireless Personal Communications, vol. 115, no. 2, pp. 909–925, Nov. 2020. [29] M. Chakraborty and D. Mitra, ‘‘Epilepsy seizure detection using kurtosis based VMD’s parameters selection and bandwidth features,’’ Biomedical Signal Processing and Control, vol. 64, Feb. 2021, Art. no. 102255. [30] H. Choubey and A. Pandey, ‘‘A combination of statistical parameters for the detection of epilepsy and EEG classification using ANN and KNN classifier,’’ Signal, Image Video Processing, vol. 15, no. 3, pp. 475–483, Apr. 2021. [31] H. Akbari, S. S. Esmaili, and S. F. Zadeh, ‘‘Detection of seizure EEG signals based on reconstructed phase space of rhythms in EWT domain and genetic algorithm,’’ Signal Processing and Renewable Energy, vol. 4, no. 2, pp. 23–36, 2020. [32] C. Xiao, S. Wang, L. Iasemidis, S. Wong, and W. A. Chaovalitwongse, ‘‘An adaptive pattern learning framework to personalize online seizure prediction,’’ IEEE Transactions on Big Data, vol. 7, no. 5, pp. 819–831, Nov. 2021. [33] J. M.-T. Wu, M.-H. Tsai, C.-T. Hsu, H.-C. Huang, and H.-C. Chen, ‘‘Intelligent signal classifier for brain epileptic EEG based on decision tree, multilayer perceptron and over-sampling approach,’’ in Advances in Information and Communication,2020, pp. 11–24. [34] S.Raghu,N.Sriraam,Y.Temel,S.V.Rao,andP.L.Kubben,‘‘EEG-based multi-class seizure type classification using convolutional neural network and transfer learning,’’ Neural Networks, vol. 124, pp. 202–212, Apr. 2020. [35] I. Ahmad, X. Wang, M. Zhu, C. Wang, Y. Pi, J. A. Khan, S. Khan, O. W. Samuel,S.Chen,andG.Li,‘‘EEG-based epileptic seizure detection via machine/deep learning approaches: A systematic review,’’ Computational Intelligence and Neuroscience, vol. 2022, pp. 1–20, Jun. 2022. [36] D.P. Dash, M. Kolekar, C. Chakraborty, M.R. Khosravi, “Review of machine and deep learning techniques in epileptic seizure detection using physiological signals and sentiment analysis”, ACM Transactions on Asian Low-Resource Language Information Processing, 23 (1) (2024), pp. 1-29. [37] Y. Gao, A. Liu, L. Wang, R. Qian, X. Chen, “A self-interpretable deep learning model for seizure prediction using a multi-scale prototypical part network”, IEEE Transactions on Neural Systems and Rehabilitation Engineering, 31 (2023), pp. 1847-1856. [38] R. Saemaldahr, M. Ilyas, “Patient-specific preictal pattern-aware epileptic seizure prediction with federated learning”, Sensors, 23 (14) (Jul. 2023) [39] G.M. Sigsgaard, Y. Gu, “Comparison of patient non-specific seizure detection using multi-modal signals”, Neuroscience Informatics, 4 (2024), Article 100152. [40] R.K. Bhadra, P.K. Singh, M. Mahmud, “HyEpiSeiD: a hybrid convolutional neural network and gated recurrent unit model for epileptic seizure detection from electroencephalogram signals”, Brain Informatics, 11 (1) (Aug. 2024), pp. 1-12. [41] Mohammed E. Seno, Niladri Maiti, Maulik Patel, Mihirkumar M. Patel, Kalpesh B. Chaudhary, Ashish Pasaya, Babacar Toure, “EEG–fNIRS signal integration for motor imagery classification using deep learning and evidence theory”, Neuroscience Informatics, Volume 5, Issue 3, September 2025, 100214. [42] S. A. Karthik, K. N. Bharath, B. R. Ramji, Kiran Puttegowda, B. Aruna, D. S. Sunil Kumar, “Enhanced EEG Signal Processing for Accurate Epileptic Seizure Detection”, SN Computer Science, June 2025, 6:608 [43] Hany F. Atlam ,Gbenga Ebenezer Aderibigbe and Muhammad Shahroz Nadeem, “Effective Epileptic Seizure Detection with Hybrid Feature Selection and SMOTE-Based Data Balancing Using SVM Classifier”, Applied Sciences, April 2025, 15(9), 4690; [44] Wang, Z.; Liu, F.; Shi, S.; Xia, S.; Peng, F.; Wang, L.; Ai, S.; Xu, Z. Automatic epileptic seizure detection based on persistent homology”, Frontiers in Physiology, 2023, 14, 1227952. [45] Indu Dokare , Sudha Gupta, “A hybrid metaheuristic framework for epileptic seizure detection in healthcare decision support systems”, Acta Epileptologica, 2025 Sep 1;7:42 [46] Dokare I, Gupta S., “Optimized seizure detection leveraging band-specific insights from limited EEG channels”, Health Information Science and Systems, 2025;13(1):30. [47] Pushpa Balakrishnan, S. Hemalatha, Dinesh Nayak Shroff Keshav, “Detection of Startle-Type Epileptic Seizures using Machine Learning Technique”, International Journal of Epilepsy, 2018;5:92–98. [48] Qi Li, Wei Cao, Anyuan Zhang, “Multi-stream feature fusion of vision transformer and CNN for precise epileptic seizure detection from EEG signals”, Journal of Translational Medicine, 2025 Aug 6;23:871. [49] Georgis-Yap Z, Popovic MR, Khan SS. Supervised and unsupervised deep learning approaches for EEG seizure prediction. Journal of Healthcare Informatics Research 2024;8:286–312. 10.1007/s41666-024-00160-x. [50] V. N. Vaddi and P. Kodali, \"Enhanced Epilepsy Sensitivity and Detection Rate With Improved Specificity by Integration of Modified LSTM Networks,\" in IEEE Sensors Journal, vol. 25, no. 7, pp. 11963-11970, 1 April1, 2025, doi: 10.1109/JSEN.2025.3543182. [51] Amudhan S, Dhaksha Muthukumaran and Dr. G Vijaya Kumar, “Real-Time Seizure Prediction from EEG Signals Using AI and Wearable Technology: An Integrated Approach for Proactive Epilepsy Management”, Open Access Research Journal of Science and Technology, June 2025, 14(01), 062-075.

Copyright

Copyright © 2025 Abinaya R. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.