ELECTROPHYSIOLOGIC LOCALIZATION OF PATHOLOGICAL BRAIN TISSUE IS ACCOMPLISHED WITH INTRACRANIAL SLEEP STAGING

Abstract Introduction Low Frequency brain rhythms facilitate communication across large spatial regions in the brain and high frequency rhythms are thought to signify local processing among nearby assemblies. A heavily investigated mode by which these low frequency and high frequency phenomenon interact is Phase-Amplitude Coupling )PAC). This phenomenon has recently shown promise as a novel electrophysiologic biomarker, in a number of neurologic diseases. In 10 patients undergoing phase-2 monitoring for the evaluation of surgical resection and in whom temporal depth electrodes were implanted, we investigated electrophysiologic relationships of PAC in epileptogenic (seizure onset zone or SOZ) and non-epileptogenic tissue (non-SOZ). That this biomarker can differentiate pathologic from non-pathologic brain and has been established with ictal and pre-ictal data, but less so with interictal data. Here we show that this biomarker can differentiate interictally. We also show PAC activity is related to interictal epileptiform discharges and high frequency activity. Importantly, we also show a differential level of PAC in slow-wave-sleep from NREM1-2 and awake. And finally we show that localization of pathologic tissue sensitivity and specificity is optimal when utilizing beta or alpha phase onto high-gamma or Ripple with knowledge of the sleep stage. Illustrating some of the physiologic nature of this biomarker in human epilepsy will provide a basis for understanding the mechanism of neurologic disease and normal physiology of brain communication, details which are at this point ready to be utilized in neurotechnological therapies to treat and understand both. Methods Per Institutional Review Board protocol, 10 patients who were under evaluation for resective surgery for MRE at Mayo Clinic in Rochester MN were included in this study. IRB approved the study and necessary consenting procedures were followed prior to any data acquisition. All subjects had bilaterally placed intracranial depth electrodes with usually 8 contacts. In some cases not all contacts could be used for data acquisition (hardware or recording problems). 6 Subjects of this cohort had scalp and EMG recordings concurrently placed for the purposes of sleep scoring. Subject recordings were ignored for POD-1 as anesthetics were dissipating. Subjects then stayed in the ICU ranging from 3-12 days before explanation.Pathological tissue identified as seizure onset zone (SOZ) was determined from phase II monitoring and determined by a trained neurologist. Sleep staging was done with expert-in-the-loop semi automated methods described elsewhere but overseen by a trained neurologist. Behavioral state was determined with scalp EEG signals and verified by a neurologist board certified in sleep medicine. All EEG recordings were bandpass filtered 0.3-75Hz and 60Hz notch filtered for scoring. Visual sleep scoring was in accordance with standard methods with modification for replacing the electrooculogram (EOG) recording with FP1, FP2, FPZ scalp electrodes. Wakefulness was determined by the presence of eye blinks visualized in fronto-parietal scalp leads, accompanied by posteriorly dominant alpha rhythms (8 - 12 Hz) comprising >50% of the epoch. Slow-wave sleep (N3) was scored when high-voltage (>75 uV) delta (0.5 - 3 Hz) frequency scalp EEG activity was present in at least 20% of the epoch (i.e., at least 6 s within a 30 s epoch) in the frontal derivations using conventional International 10-20 System electrode placements (FP1, FP2, FZ, F3, F4, CZ, C3, C4, O1, O2, and Oz). Phase Amplitude Coupling with Coherency Angle CFC. A Hanning taper n points is the length of the sliding time window. Next, the coherency CFC(fmodulating,fmodulated) was estimated between signal {Xt} and the estimate of the time-course of power{Pt(fmodulated)} for a given frequency fmodulating. The coherence was the absolute value of the CFC . The phase difference between the signal at fmodualting and the power at fmodulated is given by the angle of the coherency arg(CFC). In this case γ refer to a 1024 points Hanning window and * to the complex conjugate. This allowed us to characterize the phase-to-power cross-frequency interaction with respect to f and fmodulated sensor by sensor. The spike detection algorithm was utlized to evaluated successive 1 minute blocks of iEEG and removes artifact channels. Individual channels were defined as average slope greater than 10 SD outside of mean slopes of all channels. Second, iEEG was bandpass filtered 20-50Hz to identify possible spikes, where a sharp discharge must last between 20-70ms. Absolute amplitudes of peaks greater than 4SD of channel mean amplitude were noted as potential spike locations for further consideration. Third, raw iEEG was bandpass filtered (2nd order Butterworth) 1-35Hz. A scaling factor is determined by finding a value that will bring the median of all channel amplitudes to 70uV. All channels are multiplied by this scaling factor. Once the data have been scaled, the amplitude and slope of each half-wave of the potential spikes identified previously in step 2 are calculated and the values are compared to static thresholds (Total amplitude of both half-waves > 600μV, slope of each half-wave > 7μV/ms, duration of each half-wave > 10ms). Potential spikes with half-waves that exceed these thresholds are marked as interictal spikes.HFOs were detected using a Hilbert transform-based method, as previously reported. Here, the discrete time series is transformed into an analytic signal, where the real part is the original signal, and the imaginary part is the Hilbert transform of the original, x(t). Results Holding constant the frequency for amplitude to include all high activity (30-175Hz), beta is the best localizing (not statistically significant from alpha) band. Holding the frequency for phase, constant, there is no statistical difference between LG, HG, and ripple in terms of pathological bain tissue localization. Examining different frequencies for phase in varied behavioral/sleep states using 12 minute awake segments were used in all 10 patients, localization is best via AUROC when delta is the frequency band used to calculate PAC in Slow Wave Sleep / NREM3.Interictal Epileptiform spiking is seen with much higher regularity, although correlated and seen in the SOZ. A great deal of the associated elevations in wide-spectrum (0.5-30 modulating 65-175Hz) PAC (here defined as >2 SD across the entire 2 hour period and assessed across all channels) occurred with IEDs. Since 3 second segments were used to calculate PAC values, sometimes 2 IEDs were within a single epoch although in a small minority of observations. PAC elevations are seen with IEDs and with HFO, although to a greater extend with spikes. However, some IEDs are not associated with elevations in PAC. In fact most detected IEDs were not associated with elevated PAC values or with HFOs for that matter. A minority of IEDs are associated with elevated PAC values and HFOs. At the group and individual levels the average spike count per patient, grouping all SOZ and non-SOZ electrodes together, there is not a statistically significant effect, but if the average count is taken per electrode within a single patient, a significant difference was noted at p = .0031. Subjects on average had 15 (SEM 0.15) electrodes and SOZ electrode counts of 3 (SEM 0.71). When analyzing IED + elevated PAC, grouped SOZ electrodes among patients and within individuals show a strongly significant effect (.0004). When taking into account electrode numbers within each group this significance only increases (.0001). At the individual level, significance of p<0.05 is seen for all but one patient in this cohort. Conclusion Sleep Stage is critical in the analysis of pathological brain from non-pathological brain and electrophysiologic biomarkers behave differently in the different behavioral states. Here we found evidence to support his proposition in the following ways: low frequency delta phase modulates a broad high frequency amplitude in N3 and has relevance for brain pathology. Phase-Amplitude Coupling is increased in pathologic tissue. Peaks in PAC occur sporadically and infrequently in these patients. PAC is correlated with Interictal Epileptiform Discharges and High Frequency Oscillations however most Interictal Epileptiform Discharges are unrelated to peaks in Phase-Amplitude Coupling. Low Frequency Activity in Beta-band modulates broad high frequency power across low gamma, high gamma, and ripple bands. Support (If Any) Neurophysiological mapping and stimulation of the human brain for memory enhancement. Funded by EUO - PIS via BioTechMed Center, Department of Multimedia Systems, Faculty of Electronics, Telecommunication and Informatics, Gdansk University of Technology