Predictive modeling of EEG time series for evaluating surgery targets in epilepsy patients

Abstract

During the last 20 years, predictive modeling in epilepsy research has largely been concerned with the prediction of seizure events, whereas the inference of effective brain targets for resective surgery has received surprisingly little attention. In this exploratory pilot study, we describe a distributional clustering framework for the modeling of multivariate time series and use it to predict the effects of brain surgery in epilepsy patients. By analyzing the intracranial EEG, we demonstrate how patients who became seizure free after surgery are clearly distinguished from those who did not. More specifically, for 5 out of 7 patients who obtained seizure freedom (= Engel class I) our method predicts the specific collection of brain areas that got actually resected during surgery to yield a markedly lower posterior probability for the seizure related clusters, when compared to the resection of random or empty collections. Conversely, for 4 out of 5 Engel class III/IV patients who still suffer from postsurgical seizures, performance of the actually resected collection is not significantly better than performances displayed by random or empty collections. As the number of possible collections ranges into billions and more, this is a substantial contribution to a problem that today is still solved by visual EEG inspection. Apart from epilepsy research, our clustering methodology is also of general interest for the analysis of multivariate time series and as a generative model for temporally evolving functional networks in the neurosciences and beyond. Hum Brain Mapp 38:2509-2531, 2017. © 2017 Wiley Periodicals, Inc.

Evaluating resective surgery targets in epilepsy patients: A comparison of quantitative EEG methods

“…el. Patient label in Steimer et al (2017) Rummel et al (2015) 1 I MTLE (R) Non-lesional 64 20 49 9 I-1 2 I MTLE (L) Hippocampal sclerosis 64 13 51 6 I-2 3 I LTLE (L) Cluster of dysplastic neurons 56 5 56 – I-3 4 I PLE (L) Low-grade glioma 74 13 2 4 I-4 5 I MTLE (L) Hippocampal sclerosis 42 11 40 2 I-5 6 I FLE (R) Non-lesional 98 11 98 1 I-6 7 I TLE (L) …”

“…el.). For easier comparison with earlier publications the labels used in Steimer et al (2017) and Rummel et al (2015) are also given (hyphen means this patient was not used in the respective publication). Abbreviations: MTLE: mesial temporal lobe epilepsy, LTLE: lateral temporal lobe epilepsy, PLE: parietal lobe epilepsy, FLE: frontal lobe epilepsy, TLE: temporal lobe epilepsy, FTE: fronto-temporal epilepsy, R: right, L: left.…”

“…Based on the global network ictogenicity they classified correctly 14 out of 16 patients (AUC = 0.87). Steimer et al presented a distributional, soft clustering model for the predictive modeling of multivariate, peri-ictal iEEG time series ( Steimer et al, 2017 ). This model permits patient-specific predictions about seizure propensity under arbitrary simulated resections of brain tissue.…”

“…For this cross-method verification of two fundamentally differing methods we focus on comparing two methods that have recently been developed and tested at our institute and have shown convincing performances by quantitative comparison with the actual resection and outcome in patients undergoing surgery. That is, we directly compare the assessments of hypothetical resections by the nonlinear interrelation measure examined by Rummel et al (2015) with the resections’ seizure suppressing efficiencies as estimated by the model of Steimer et al (2017) . Both methods have shown promise in the prediction of tissue resection in epilepsy surgery.…”

Quantification and selection of ictogenic zones in epilepsy surgery

“…Computational studies of seizure generation in large-scale brain networks with the aim to inform epilepsy surgery have recently begun to emerge (Goodfellow et al, 2016, Sinha et al, 2017, Steimer et al, 2017. In an early such study we introduced a quantitative framework for prospectively evaluating the effect that surgical removal of tissue would have (Goodfellow et al, 2016).…”