Research Fellow Wayne State University Farmington Hills, Michigan
Rationale: Interictal high-frequency oscillation (HFO) is a promising epilepsy biomarker to localize the epileptogenic zone. Investigators have developed algorithms to compute the occurrence rate of HFO and those to quantify the severity of coupling between the HFO amplitude and a slow-wave phase. The present study investigated the performance of HFO rate and phase-amplitude coupling as epilepsy biomarkers in a cohort of 135 patients who underwent cortical resection following extraoperative intracranial EEG (iEEG) recording. We determined if consideration of an HFO rate would improve the conventional model to predict seizure outcome solely based on the clinical, seizure-onset zone (SOZ), and neuroimaging variables. We likewise determined if incorporation of the phase-amplitude coupling would improve the conventional prediction model. Methods: We computed the HFO rate at given channels with each of the following four detectors incorporated in a toolbox, RIPPLELAB: [1] short-time energy (STE), [2] short-line length (SLL), [3] Hilbert, and [4] Montreal Neurological Institute (MNI) methods. The EEGLAB Toolbox computed the MI quantifying the coupling between HFO and slow-wave at 3-4 Hz. Based on our iEEG normative atlas consisting of 2,477 nonepileptic electrode sites sampled from 47 patients, we computed the statistical deviation (z-score) of the HFO rate from the nonepileptic mean. We finally computed 'subtraction z-HFO' defined as the subtraction of z-score of HFO rate averaged across all preserved sites from that averaged across all resected sites. We hypothesized that larger 'subtraction z-HFO' would be independently associated with a better surgical outcome. We determined if the full logistic regression model incorporating 'subtraction z-HFO' would improve the accuracy of prediction of patients achieving ILAE Class-1 outcome. We likewise computed 'subtraction z-MI' and assessed its performance in the prediction of seizure outcome (Figure 1).
Reference: [1] Staba RJ et al. Quantitative analysis of high-frequency oscillations (80.500 Hz) recorded in human epileptic hippocampus and entorhinal cortex. J Neurophysiol. 2002; 88:1743-1752. [2] Gardner AB et al. Human and automated detection of high-frequency oscillations in clinical intracranial EEG recordings. Clin Neurophysiol. 2007; 118:1134-1143. [3] Crepon B et al. Mapping interictal oscillations greater than 200 Hz recorded with intracranial macroelectrodes in human epilepsy. Brain. 2010; 133: 33-45. [4] Zelmann R et al. A comparison between detectors of high frequency oscillations. Clin Neurophysiol. 2012; 123: 106-116. Results: Ninety-five patients (70.4%) had a Class-1 outcome. ‘Subtraction z-HFO>80 Hz’ improved the performance of outcome prediction (STE: p = 0.078; SLL: p = 0.030; Hilbert: p = 0.066; MNI: p=0.003). Neither 'subtraction z-HFO>150 Hz' nor 'subtraction z-HFO>250 Hz' improved the performance of outcome prediction (p >0.05). ‘Subtraction z-MI’ improved the prediction performance (z-MI>80 Hz: p = 0.012; z-MI>150 Hz: p = 0.014; z-MI>250 Hz: p = 0.014). The conventional model predicted a Class-1 outcome with an accuracy of 0.751. Incorporation of 'subtraction z-HFO>80 Hz' defined by STE, SLL, Hilbert, and MNI detectors improved the accuracy to 0.759, 0.766, 0.757, and 0.799, respectively (Figure 2). 'Subtraction z-MI>80 Hz', 'subtraction z-MI>150 Hz', and 'subtraction z-MI>250 Hz' improved the accuracy to 0.791, 0.818, and 0.810, respectively. Conclusions: The rate of automatically detected HFO>80 Hz and MI modestly improved the performance of outcome prediction in our large patient cohort. The effect size of prediction performance was comparable between HFO and MI measures. Funding: Please list any funding that was received in support of this abstract.: NIH grant NS064033. Click here to view image/table
