Recursive model identification for the analysis of cardiovascular autonomic modulation during epileptic seizures

Quentin Gillardin1, Virginie Le Rolle2, Anca Nica3, Arnaud Biraben3, Benoît Martin4, Alfredo Hernandez5
1LTSI, 2LTSI - INSERM U1099 - Université de Rennes 1, 3CHU Rennes, 4LTSI U1099, 5INSERM - LTSI U 1099


Background: Significant tachy- and bradycardia are often observed during epileptic seizures. These cardiac modifications, associated with apnea, are considered to be involved on sudden and unexpected death in epilepsy (SUDEP). We hypothesize that these acute cardio-respiratory events are mediated by specific dynamics of the autonomic nervous system (ANS). However, the evaluation of ANS during seizures remains particularly challenging, mainly due to the lack of observability. Computational modelling could help to override these limitations, access the ANS modulation and evaluate this hypothesis. Aim of the study is to evaluate the dynamics of sympathetic and parasympathetic activities during ictal tachy-bradycardia events. Methods: A recursive identification method was proposed and applied to a computational, system-level model of the autonomic modulation of the sinoatrial node, in order to estimate the time-varying vagal and sympathetic contributions to heart rate modulation during ictal period. This method was evaluated on heart rate (HR) data from one epileptic patient during four seizures, including 90s before to 90s after the onset of each seizure. The cost function was defined as the root mean squared error between simulated and observed HR, regularized with two terms taking into account past variations of the identified parasympathetic and sympathetic modulations. The identification method has been optimized through a local sensitivity analysis. Results: After parameter identification, a close match between experimental and simulated signals was observed, with a median RMSE=1.7bpm. The time-varying parasympathetic and sympathetic modulations, estimated from the model-based method, show, for the 4 events analyzed, a co-activation of the sympathetic and parasympathetic components at seizure onset, followed by a massive vagal activation, leading to bradycardia in a second phase. Conclusion: The proposed patient-specific model-based method shows reproducible autonomic dynamics for ictal tachy-bradycardia events, involving a first phase of sympathetic and parasympathetic co-activation, followed by a massive vagal activation.