癲癇是一種以腦內神經元反覆異常或過度同步放電為主要特徵的神經系統疾病，其中約三成患者無法透過藥物有效控制。對於不適合接受手術或術後治療效果不佳的患者，神經電刺激為可行的治療方式之一。在電刺激系統中，癲癇偵測方法的性能是影響治療效果的重要因素，越早偵測到癲癇發作並進行電刺激，越有機會成功抑制癲癇活動。傳統的癲癇偵測方法多以訊號中的時域或頻域特徵作為判斷依據，較難反映癲癇發作背後的生理機制。相較之下，結合神經群模型與卡爾曼濾波器的模型驅動方法，可透過估測具有生理意義的模型參數來追蹤神經動態，不僅能夠了解癲癇過程中大腦內部不同狀態的轉變，也具備癲癇偵測與量化的潛力。傳統上多使用交流耦合放大器記錄患者的癲癇訊號，但此類記錄方式會造成癲癇發作特徵之一的發作期直流偏移流失。發作期直流偏移可出現在癲癇病灶周圍，因此目前主要應用於病灶定位。此外，其出現時間位於癲癇發作初期，甚至可能早於傳統上用於辨識癲癇發作的節律性放電。因此，保留此類特徵能夠獲得較完整的癲癇訊號，並預期有助於得到較穩定且準確的參數估測結果。基於上述背景，本研究透過動物實驗探討保留發作期直流偏移對神經群模型參數估測與癲癇偵測性能的影響。首先，利用4-AP誘發小鼠產生癲癇，並使用直流耦合放大器記錄其癲癇訊號。接著，經由前處理分別獲得發作期低頻偏移（Ictal slow shift, ISS）去除訊號（ISS-removed signal）與發作期低頻偏移保留訊號（ISS-preserved signal），並使用拘束方根容積卡爾曼濾波器估測神經群模型中的慢速抑制性突觸增益B，以進行癲癇偵測，最後比較兩種訊號的參數估測與癲癇偵測結果。在成功誘發癲癇的小鼠中，皆可觀察到ISS。在保留ISS的情況下，參數B在不同癲癇波形之間呈現階段性的分布，與先前研究的模擬測試結果相符。在連續癲癇偵測結果中，相較於ISS去除訊號，使用ISS保留訊號可提高正確辨識比例，並降低誤判率。此外，偵測癲癇起始點的延遲時間中位數由0.03秒降至−8.48秒，表示可提前約8秒偵測到癲癇發作。綜合而言，在訊號中保留發作期直流偏移可改善神經群模型參數估測的穩定性，並提升癲癇偵測性能。

Epilepsy is a neurological disorder characterized by recurrent abnormal or excessive synchronous neuronal discharges. For patients with drug-resistant epilepsy, neurostimulation provides an alternative treatment, and early seizure detection is important for timely intervention. Model-based seizure detection using neural mass models and Kalman filters can track neural dynamics through physiologically meaningful parameters. However, conventional AC-coupled recordings may remove ictal direct current (DC) shifts, which can occur during the early stage of seizures and may contain information relevant to parameter estimation and seizure detection. This study investigated the effect of preserving ictal DC shifts on neural mass model-based parameter estimation and seizure detection. Seizures were induced in mice using 4-aminopyridine, and electrophysiological signals were recorded with a DC-coupled amplifier. Ictal slow shift (ISS)-preserved and ISS-removed signals were obtained through signal preprocessing. A constrained square-root cubature Kalman filter was then used to estimate the slow inhibitory synaptic gain, parameter B, for seizure detection. ISS was observed in all mice with successfully induced seizures. With ISS preserved, parameter B showed clearer stage-dependent distributions across different epileptic waveforms. ISS-preserved signals also improved seizure detection performance compared with ISS-removed signals. The median detection latency decreased from 0.03 s to −8.48 s, indicating seizure detection approximately 8 s earlier. Overall, preserving ictal DC shifts improved neural mass model-based parameter estimation and enhanced seizure detection performance.

[1] T. A. Milligan, “Epilepsy: A Clinical Overview,” Am. J. Med., vol. 134, no. 7, pp. 840–847, Jul. 2021, doi: 10.1016/j.amjmed.2021.01.038.
[2] R. S. Fisher et al., “Epileptic Seizures and Epilepsy: Definitions Proposed by the International League Against Epilepsy (ILAE) and the International Bureau for Epilepsy (IBE),” Epilepsia, vol. 46, no. 4, pp. 470–472, Apr. 2005, doi: 10.1111/j.0013-9580.2005.66104.x.
[3] E. Beghi, C. Cornaggia, and the RESt‐1 Group, “Morbidity and Accidents in Patients with Epilepsy: Results of a European Cohort Study,” Epilepsia, vol. 43, no. 9, pp. 1076–1083, Sep. 2002, doi: 10.1046/j.1528-1157.2002.18701.x.
[4] L. M. Willems et al., “Incidence, Risk Factors and Consequences of Epilepsy-Related Injuries and Accidents: A Retrospective, Single Center Study,” Front. Neurol., vol. 9, p. 414, Jun. 2018, doi: 10.3389/fneur.2018.00414.
[5] E. Beghi, “The Epidemiology of Epilepsy,” Neuroepidemiology, vol. 54, no. 2, pp. 185–191, 2020, doi: 10.1159/000503831.
[6] J. F. Téllez-Zenteno and L. Hernández-Ronquillo, “A Review of the Epidemiology of Temporal Lobe Epilepsy,” Epilepsy Res. Treat., vol. 2012, pp. 1–5, Dec. 2012, doi: 10.1155/2012/630853.
[7] R. S. Fisher et al., “Operational classification of seizure types by the International League Against Epilepsy: Position Paper of the ILAE Commission for Classification and Terminology,” Epilepsia, vol. 58, no. 4, pp. 522–530, Apr. 2017, doi: 10.1111/epi.13670.
[8] W. O. T. Iv, “Mesial Temporal Lobe Epilepsy,” vol. 29, no. 5, 2012.
[9] L. Kalilani, X. Sun, B. Pelgrims, M. Noack‐Rink, and V. Villanueva, “The epidemiology of drug‐resistant epilepsy: A systematic review and meta‐analysis,” Epilepsia, vol. 59, no. 12, pp. 2179–2193, Dec. 2018, doi: 10.1111/epi.14596.
[10] G. Liu and H. V. M. Associates, “Epilepsy: Treatment Options”.
[11] B. C. Jobst et al., “Brain‐responsive neurostimulation in patients with medically intractable seizures arising from eloquent and other neocortical areas,” Epilepsia, vol. 58, no. 6, pp. 1005–1014, Jun. 2017, doi: 10.1111/epi.13739.
[12] H. Kassiri, A. Muneeb, R. Salahi, and A. Dabbaghian, “Closed-Loop Implantable Neurostimulators for Individualized Treatment of Intractable Epilepsy: A Review of Recent Developments, Ongoing Challenges, and Future Opportunities,” IEEE Trans. Biomed. Circuits Syst., vol. 18, no. 6, pp. 1268–1295, Dec. 2024, doi: 10.1109/TBCAS.2024.3456825.
[13] B. P. Grone and S. C. Baraban, “Animal models in epilepsy research: legacies and new directions,” Nat. Neurosci., vol. 18, no. 3, pp. 339–343, Mar. 2015, doi: 10.1038/nn.3934.
[14] L. Kandratavicius et al., “Animal models of epilepsy: use and limitations,” Neuropsychiatr. Dis. Treat., p. 1693, Sep. 2014, doi: 10.2147/NDT.S50371.
[15] A. J. Becker, “Review: Animal models of acquired epilepsy: insights into mechanisms of human epileptogenesis,” Neuropathol. Appl. Neurobiol., vol. 44, no. 1, pp. 112–129, Feb. 2018, doi: 10.1111/nan.12451.
[16] F. Peña and R. Tapia, “Seizures and neurodegeneration induced by 4-aminopyridine in rat hippocampus in vivo: role of glutamate- and GABA-mediated neurotransmission and of ion channels,” Neuroscience, vol. 101, no. 3, pp. 547–561, Nov. 2000, doi: 10.1016/S0306-4522(00)00400-0.
[17] L. E. Gonzalez-Reyes, T. P. Ladas, C.-C. Chiang, and D. M. Durand, “TRPV1 antagonist capsazepine suppresses 4-AP-induced epileptiform activity in vitro and electrographic seizures in vivo,” Exp. Neurol., vol. 250, pp. 321–332, Dec. 2013, doi: 10.1016/j.expneurol.2013.10.010.
[18] A. Bonaccini Calia et al., “Full-bandwidth electrophysiology of seizures and epileptiform activity enabled by flexible graphene microtransistor depth neural probes,” Nat. Nanotechnol., vol. 17, no. 3, pp. 301–309, Mar. 2022, doi: 10.1038/s41565-021-01041-9.
[19] R. J. Gumnit and T. Takahashi, “Changes in direct current activity during experimental focal seizures,” Electroencephalogr. Clin. Neurophysiol., vol. 19, no. 1, pp. 63–74, Jul. 1965, doi: 10.1016/0013-4694(65)90007-6.
[20] A. Ikeda et al., “Subdural Recording of Ictal DC Shifts in Neocortical Seizures in Humans,” Epilepsia, vol. 37, no. 7, pp. 662–674, Jul. 1996, doi: 10.1111/j.1528-1157.1996.tb00631.x.
[21] A. Ikeda, “Focal ictal direct current shifts in humanepilepsy as studied by subdural and scalp recording,” Brain, vol. 122, no. 5, pp. 827–838, May 1999, doi: 10.1093/brain/122.5.827.
[22] S. Wu, H. P. K. Veedu, S. D. Lhatoo, M. Z. Koubeissi, J. P. Miller, and H. O. Lüders, “Role of ictal baseline shifts and ictal high‐frequency oscillations in stereo‐electroencephalography analysis of mesial temporal lobe seizures”, doi: 10.1111/epi.12608.
[23] K. Kanazawa et al., “Intracranially recorded ictal direct current shifts may precede high frequency oscillations in human epilepsy,” Clin. Neurophysiol., vol. 126, no. 1, pp. 47–59, Jan. 2015, doi: 10.1016/j.clinph.2014.05.028.
[24] A. Ikeda et al., “Active direct current (DC) shifts and ‘Red slow’: two new concepts for seizure mechanisms and identification of the epileptogenic zone,” Neurosci. Res., vol. 156, pp. 95–101, Jul. 2020, doi: 10.1016/j.neures.2020.01.014.
[25] M. Nakatani et al., “Ictal direct current shifts contribute to defining the core ictal focus in epilepsy surgery,” Brain Commun., vol. 4, no. 5, p. fcac222, Sep. 2022, doi: 10.1093/braincomms/fcac222.
[26] M. Izumi et al., “Focal ictal direct current shifts by a time constant of 2 seconds were clinically useful for resective epilepsy surgery,” Epilepsia, vol. 64, no. 12, pp. 3294–3306, 2023, doi: 10.1111/epi.17782.
[27] V. K. Jirsa, W. C. Stacey, P. P. Quilichini, A. I. Ivanov, and C. Bernard, “On the nature of seizure dynamics,” Brain, vol. 137, no. 8, pp. 2210–2230, Aug. 2014, doi: 10.1093/brain/awu133.
[28] M. A. Lopes, K.-E. Lee, and A. V. Goltsev, “Neuronal network model of interictal and recurrent ictal activity,” Phys. Rev. E, vol. 96, no. 6, p. 062412, Dec. 2017, doi: 10.1103/PhysRevE.96.062412.
[29] P. Boonyakitanont, A. Lek-uthai, K. Chomtho, and J. Songsiri, “A review of feature extraction and performance evaluation in epileptic seizure detection using EEG,” Biomed. Signal Process. Control, vol. 57, p. 101702, Mar. 2020, doi: 10.1016/j.bspc.2019.101702.
[30] R. Esteller, J. Echauz, T. Tcheng, B. Litt, and B. Pless, “Line length: an efficient feature for seizure onset detection,” in 2001 Conference Proceedings of the 23rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Istanbul, Turkey: IEEE, 2001, pp. 1707–1710. doi: 10.1109/IEMBS.2001.1020545.
[31] N. Kannathal, M. L. Choo, U. R. Acharya, and P. K. Sadasivan, “Entropies for detection of epilepsy in EEG,” Comput. Methods Programs Biomed., vol. 80, no. 3, pp. 187–194, Dec. 2005, doi: 10.1016/j.cmpb.2005.06.012.
[32] M. Bandarabadi, C. A. Teixeira, J. Rasekhi, and A. Dourado, “Epileptic seizure prediction using relative spectral power features,” Clin. Neurophysiol., vol. 126, no. 2, pp. 237–248, Feb. 2015, doi: 10.1016/j.clinph.2014.05.022.
[33] Y. Yuan, G. Xun, K. Jia, and A. Zhang, “A Multi-view Deep Learning Method for Epileptic Seizure Detection using Short-time Fourier Transform,” in Proceedings of the 8th ACM International Conference on Bioinformatics, Computational Biology,and Health Informatics, Boston Massachusetts USA: ACM, Aug. 2017, pp. 213–222. doi: 10.1145/3107411.3107419.
[34] S. Naseem, K. Javed, M. Jawad Khan, S. Rubab, M. Attique Khan, and Y. Nam, “Integrated CWT-CNN for Epilepsy Detection Using Multiclass EEG Dataset,” Comput. Mater. Contin., vol. 69, no. 1, pp. 471–486, 2021, doi: 10.32604/cmc.2021.018239.
[35] R. A. Stefanescu, R. G. Shivakeshavan, and S. S. Talathi, “Computational models of epilepsy,” Seizure, vol. 21, no. 10, pp. 748–759, Dec. 2012, doi: 10.1016/j.seizure.2012.08.012.
[36] G. Hocepied, B. Legros, P. Van Bogaert, F. Grenez, and A. Nonclercq, “Early detection of epileptic seizures based on parameter identification of neural mass model,” Comput. Biol. Med., vol. 43, no. 11, pp. 1773–1782, Nov. 2013, doi: 10.1016/j.compbiomed.2013.08.022.
[37] 湯登棋, “基於神經群模型之容積卡爾曼濾波器於慢性癲癇發作之應用.” 碩士論文, 國立成功大學機械工程系, 臺灣臺南市, 2019.
[38] J.-L. Song, Q. Li, M. Pan, B. Zhang, M. B. Westover, and R. Zhang, “Seizure tracking of epileptic EEGs using a model-driven approach,” J. Neural Eng., vol. 17, no. 1, p. 016024, Jan. 2020, doi: 10.1088/1741-2552/ab2409.
[39] X. Liu and Q. Gao, “Parameter estimation and control for a neural mass model based on the unscented Kalman filter,” Phys. Rev. E, vol. 88, no. 4, p. 042905, Oct. 2013, doi: 10.1103/PhysRevE.88.042905.
[40] 呂柏諺, “方根容積卡爾曼濾波器於癲癇光刺激抑制之研究.” 碩士論文, 國立成功大學機械工程系, 臺灣臺南市, 2021.
[41] A. Jafarian and R. C. Wykes, “Impact of DC-Coupled Electrophysiological Recordings for Translational Neuroscience: Case Study of Tracking Neural Dynamics in Rodent Models of Seizures,” Front. Comput. Neurosci., vol. 16, p. 900063, Jul. 2022, doi: 10.3389/fncom.2022.900063.
[42] C.-H. Huang, P.-H. Wang, M.-S. Ju, and C.-C. K. Lin, “Using Constrained Square-Root Cubature Kalman Filter for Quantifying the Severity of Epileptic Activities in Mice,” Biomedicines, vol. 10, no. 7, p. 1588, Jul. 2022, doi: 10.3390/biomedicines10071588.
[43] F. Wendling, P. Benquet, F. Bartolomei, and V. Jirsa, “Computational models of epileptiform activity,” J. Neurosci. Methods, vol. 260, pp. 233–251, Feb. 2016, doi: 10.1016/j.jneumeth.2015.03.027.
[44] B. H. Jansen and V. G. Rit, “Electroencephalogram and visual evoked potential generation in a mathematical model of coupled cortical columns”.
[45] P. J. Karoly, L. Kuhlmann, D. Soudry, D. B. Grayden, M. J. Cook, and D. R. Freestone, “Seizure pathways: A model-based investigation,” PLOS Comput. Biol., vol. 14, no. 10, p. e1006403, Oct. 2018, doi: 10.1371/journal.pcbi.1006403.
[46] L. Kuhlmann et al., “Neural mass model-based tracking of anesthetic brain states,” NeuroImage, vol. 133, pp. 438–456, Jun. 2016, doi: 10.1016/j.neuroimage.2016.03.039.
[47] K. Jung, S. B. Eickhoff, J. Caspers, UKD-PD team, and O. V. Popovych, “Simulated brain networks reflecting progression of Parkinson’s disease,” Netw. Neurosci., vol. 8, no. 4, pp. 1400–1420, Dec. 2024, doi: 10.1162/netn_a_00406.
[48] H. Zhang et al., “Bidirectional long short-term memory attention neural network to estimate neural mass model parameters,” Chaos Solitons Fractals, vol. 192, p. 116051, Mar. 2025, doi: 10.1016/j.chaos.2025.116051.
[49] F. Wendling, F. Bartolomei, J. J. Bellanger, and P. Chauvel, “Epileptic fast activity can be explained by a model of impaired GABAergic dendritic inhibition,” Eur. J. Neurosci., vol. 15, no. 9, pp. 1499–1508, May 2002, doi: 10.1046/j.1460-9568.2002.01985.x.
[50] F. Wendling, A. Hernandez, J.-J. Bellanger, P. Chauvel, and F. Bartolomei, “Interictal to Ictal Transition in Human Temporal Lobe Epilepsy: Insights From a Computational Model of Intracerebral EEG,” J. Clin. Neurophysiol., vol. 22, no. 5, 2005.
[51] R. J. Moran et al., “Bayesian estimation of synaptic physiology from the spectral responses of neural masses,” NeuroImage, vol. 42, no. 1, pp. 272–284, Aug. 2008, doi: 10.1016/j.neuroimage.2008.01.025.
[52] A. Hartoyo, P. J. Cadusch, D. T. J. Liley, and D. G. Hicks, “Parameter estimation and identifiability in a neural population model for electro-cortical activity,” PLOS Comput. Biol., vol. 15, no. 5, p. e1006694, May 2019, doi: 10.1371/journal.pcbi.1006694.
[53] R. E. Kalman, “A New Approach to Linear Filtering and Prediction Problems,” J. Basic Eng., vol. 82, no. 1, pp. 35–45, Mar. 1960, doi: 10.1115/1.3662552.
[54] S. F. Schmidt, “Application of State-Space Methods to Navigation Problems,” in Advances in Control Systems, vol. 3, Elsevier, 1966, pp. 293–340. doi: 10.1016/B978-1-4831-6716-9.50011-4.
[55] X.-B. Jin, R. J. Robert Jeremiah, T.-L. Su, Y.-T. Bai, and J.-L. Kong, “The New Trend of State Estimation: From Model-Driven to Hybrid-Driven Methods,” Sensors, vol. 21, no. 6, p. 2085, Mar. 2021, doi: 10.3390/s21062085.
[56] S. J. Julier and J. K. Uhlmann, “Unscented Filtering and Nonlinear Estimation,” Proc. IEEE, vol. 92, no. 3, pp. 401–422, Mar. 2004, doi: 10.1109/JPROC.2003.823141.
[57] I. Arasaratnam and S. Haykin, “Cubature Kalman Filters,” IEEE Trans. Autom. Control, vol. 54, no. 6, pp. 1254–1269, Jun. 2009, doi: 10.1109/TAC.2009.2019800.
[58] I. Arasaratnam, S. Haykin, and T. R. Hurd, “Cubature Kalman Filtering for Continuous-Discrete Systems: Theory and Simulations,” IEEE Trans. Signal Process., vol. 58, no. 10, pp. 4977–4993, Oct. 2010, doi: 10.1109/TSP.2010.2056923.
[59] A. Isaksson, A. Wennberg, and L. H. Zetterberg, “Computer analysis of EEG signals with parametric models,” Proc. IEEE, vol. 69, no. 4, pp. 451–461, Apr. 1981, doi: 10.1109/PROC.1981.11988.
[60] A. Galka, O. Yamashita, T. Ozaki, R. Biscay, and P. Valdés-Sosa, “A solution to the dynamical inverse problem of EEG generation using spatiotemporal Kalman filtering,” NeuroImage, vol. 23, no. 2, pp. 435–453, Oct. 2004, doi: 10.1016/j.neuroimage.2004.02.022.
[61] P. A. Valdes, J. C. Jimenez, J. Riera, R. Biscay, and T. Ozaki, “Nonlinear EEG analysis based on a neural mass model,” Biol. Cybern., vol. 81, no. 5–6, pp. 415–424, Nov. 1999, doi: 10.1007/s004220050572.
[62] H. Yokoyama and K. Kitajo, “A data assimilation method to track time-varying changes in the excitation-inhibition balance using scalp EEG,” Dec. 17, 2022, Neuroscience. doi: 10.1101/2022.12.16.520838.
[63] A. López-Cuevas, B. Castillo-Toledo, L. Medina-Ceja, and C. Ventura-Mejía, “State and parameter estimation of a neural mass model from electrophysiological signals during the status epilepticus,” NeuroImage, vol. 113, pp. 374–386, Jun. 2015, doi: 10.1016/j.neuroimage.2015.02.059.
[64] S. Chang et al., “A Data Driven Experimental System for Individualized Brain Stimulation Design and Validation,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 29, pp. 1848–1857, 2021, doi: 10.1109/TNSRE.2021.3110275.
[65] K. L. Navarro, M. Huss, J. C. Smith, P. Sharp, J. O. Marx, and C. Pacharinsak, “Mouse Anesthesia: The Art and Science,” ILAR J., vol. 62, no. 1–2, pp. 238–273, Dec. 2021, doi: 10.1093/ilar/ilab016.
[66] A. Ranft et al., “Isoflurane modulates glutamatergic and GABAergic neurotransmission in the amygdala,” Eur. J. Neurosci., vol. 20, no. 5, pp. 1276–1280, Sep. 2004, doi: 10.1111/j.1460-9568.2004.03603.x.
[67] M. H. Verhaegen, “Round-off error propagation in four generally-applicable, recursive, least-squares estimation schemes,” Automatica, vol. 25, no. 3, pp. 437–444, May 1989, doi: 10.1016/0005-1098(89)90013-7.
[68] R. Kandepu, B. Foss, and L. Imsland, “Applying the unscented Kalman filter for nonlinear state estimation,” J. Process Control, vol. 18, no. 7–8, pp. 753–768, Aug. 2008, doi: 10.1016/j.jprocont.2007.11.004.
[69] 王鵬翔, “深腦限頻亂數電刺激抑制小鼠顳葉癲癇發作之成效” 碩士論文, 國立成功大學機械工程系, 臺灣臺南市, 2020.