Representasi Distribusi Energi EEG Multi-Channel untuk Deteksi Aktivitas Epilepsi Menggunakan SVM

Siswandari Noertjahjani, Ratih Sari Wardani, Aris Kiswanto


Abstract


Deteksi epilepsi berbasis electroencephalogram (EEG) menghadapi tantangan berupa ketidakseimbangan kelas, data leakage, dan variasi sinyal antarpasien. Penelitian ini mengusulkan representasi distribusi energi EEG multikanal menggunakan Support Vector Machine (SVM) dengan kernel Radial Basis Function (RBF). Dataset CHB-MIT dipreproses dan disegmentasi menjadi window berdurasi 2 detik dengan overlap 25%. Setiap window direpresentasikan oleh 23 fitur energi domain waktu dan 23 fitur Power Spectral Density (PSD) metode Welch, sedangkan energy map berukuran 23 × 16 digunakan sebagai visualisasi distribusi energi kanal–waktu dan bukan sebagai masukan model. Evaluasi dilakukan menggunakan independent file-wise testing, grouped file-wise 5-fold cross-validation, dan patient-wise Leave-One-Subject-Out (LOSO). Untuk mencegah data leakage, pembagian data dilakukan pada tingkat file atau pasien sebelum segmentasi, sedangkan undersampling dan standardisasi hanya diterapkan pada data pelatihan. Hasil file-wise testing menunjukkan akurasi 93,247%, sensitivitas 98,387%, F1-score 42,958%, dan ROC-AUC 0,9880. Pada evaluasi LOSO, akurasi, sensitivitas, dan F1-score menurun menjadi 78,071%, 51,909%, dan 20,550%, yang menunjukkan adanya inter-subject variability. Uji Mann–Whitney menunjukkan perbedaan distribusi energi yang signifikan antara kelas seizure dan non-seizure (p = 2,1804 × 10⁻¹⁴⁰). Metode ini efektif untuk deteksi antarfile pada pasien yang sama, namun masih memerlukan validasi lintas pasien dan dataset eksternal untuk meningkatkan kemampuan generalisasi.
Kata kunci: CHB-MIT, deteksi epilepsi, distribusi energi EEG, patient-wise LOSO, SVM.

Keywords


CHB-MIT, deteksi epilepsi, distribusi energi EEG, patient-wise LOSO, SVM.

References


. Cai S, Lin Z, Liu X, Wei W, Wang S, Zhang M, et al. Spiking neural networks for EEG signal analysis: From theory to practice. Neural Networks. 2026;194:108127.

. Fernández-Martín R, Feys O, Juvené E, Aeby A, Urbain C, De Tiège X, et al. Towards the automated detection of interictal epileptiform discharges with magnetoencephalography. J Neurosci Methods. 2024;403:110052.

. Cai Q, Hu D, Gao F, Lou X, Cao J. Multiple epileptiform waves detection algorithm based on improved VMD and multidimensional feature fusion. J Neurosci Methods. 2026;429:110703.

. Ghosh S, Chandrasekaran V, Rohan NR, Chakravarthy VS. Electroencephalogram classification using a bio-inspired deep oscillatory neural network. Biomed Signal Process Control. 2025;103:107379.

. Zhang L, Li B, Cao M, Peng C, Wang H. Classification of EEG-fNIRS bimodal brain signals for motor imagery tasks based on wavelet transform and spatio-temporal domain processing. Neuroscience. 2026;594:42–57.

. Li C, Li H, Dong X, Zhong X, Cui H, Ji D, et al. CNN-Informer: A hybrid deep learning model for seizure detection on long-term EEG. Neural Networks. 2025;181:106855.

. Wang J, Wei L, Qian Z, Shi C, Liu Y, Xu Y. An explainable multi-view representation fusion learning framework with hybrid MetaFormer for EEG-based epileptic seizure detection. Neurocomputing. 2026;675:132929.

. Sharma M, Acharya UR. Automated detection of epileptic seizures in EEG signals using STFT and texture features. Comput Methods Programs Biomed. 2024;238:107604.

. Zhou T, Liu Y, Wang H, Zhang Z. Multi-view EEG representation learning for epileptic seizure detection. IEEE Trans Neural Syst Rehabil Eng. 2025;33:1452–1461.

. Zhang Y, Chen X, Liu J, Wang H. Explainable transformer-based model for EEG signal classification using multi-view features. Neurocomputing. 2026;670:120–132.

. Li Y, Zhang Q, Wang S, Chen X. Graph neural network-based spatio-temporal modeling for EEG seizure detection. IEEE Trans Biomed Eng. 2025;72(4):1123–1132.

. Subasi A, Ercelebi E. Classification of EEG signals using energy-based features for epileptic seizure detection. Expert Syst Appl. 2023;213:119012.

. Polat K, Güneş S. Detection of epileptic seizures in EEG using wavelet-based energy and entropy features. Appl Soft Comput. 2022;120:108689.

. Roy Y, Banville H, Albuquerque I, Gramfort A, Falk TH, Faubert J. Deep learning-based electroencephalography analysis: A systematic review. J Neural Eng. 2024;21(2):021001.

. Abbas W, Khan S, Ahmad M. Hybrid deep learning frameworks for epileptic seizure detection using EEG signals. Biomed Signal Process Control. 2025;98:106021.




DOI: https://doi.org/10.30743/jet.v11i2.13713

Refbacks

  • There are currently no refbacks.


Copyright (c) 2026 Siswandari - Noertjahjani

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.