Abstracts

Rationale: Infantile spasms (IS) is characterized by epileptic spasms, interictal hypsarrhythmia in ~2/3 of cases, and high risk of developmental impairment—especially in patients with delayed diagnosis and treatment. Identification of IS is frequently delayed, in part because routine EEG screening may not capture epileptic spasms, and hypsarrhythmia may be absent or difficult to identify. A reliable method to identify IS without EEG confirmation of epileptic spasms, and especially in the absence of hypsarrhythmia, may improve diagnostic accuracy and minimize lead time to successful treatment. To address this need, we set out to identify EEG-based metrics that distinguish children with IS from normal controls. Methods: Here, we present five candidate computational EEG metrics that were assessed using a cohort of 50 patients treated for IS at the UCLA Mattel Children’s Hospital alongside 37 control subjects. Eight of the fifty IS patients presented with hypsarrhythmia on the baseline EEG. Each video-EEG study was sampled four times (2 awake, 2 sleep) in a blinded fashion. Based on prior study of a separate smaller cohort, we suspected that IS cases and controls would be distinguished on the basis of 1) the amplitude, defined as range of the broadband bandpass-filtered data within one-second windows, 2) the power spectrum and spectral edge frequency (SEF) calculated with the fast Fourier Transform, 3) the Shannon entropy as a measure of signal disorder, 4) the Detrended Fluctuation Analysis (DFA) exponent and intercept, which reflects the long-range temporal structure of the data, and 5) functional connectivity calculated via cross-correlation. Results: Empirical cumulative distribution functions of EEG amplitude revealed significantly higher amplitude values in IS patients than control subjects in both awake and sleep data (Wilcoxon rank-sum test, p-value <0.05). High EEG power dominated the lower frequency bands in IS EEG, and the spectral edge frequency (SEF) was significantly lower in IS patients when compared with controls in awake data (Wilcoxon rank-sum test, corrected via Benjamini-Hochberg (BH) procedure, adj. p-value <0.05). In contrast to previous reports of entropy in epilepsy, we found that EEG entropy was significantly higher in IS data than control data (Wilcoxon rank-sum test, BH corrected, adj. p-value <0.05). The DFA intercept best separated IS data and control data, with significance in all frequency bands in both wakefulness and sleep (Wilcoxon rank-sum test, BH corrected, adj. p-value <0.05). Functional connectivity maps were significantly stronger in IS patients, with high levels of connectivity observed in cross-hemispheric, anterior-posterior channel pairs (Wilcoxon rank-sum test, BH corrected, adj. p-value <0.05). Conclusions: We analyzed EEG in a large cohort of IS patients and identified several computational EEG markers that differ significantly from control subjects. Because only eight of the 50 patients exhibited hypsarrhythmia on baseline EEG, this study aptly describes objective EEG characteristics of IS that are generally independent of the presence of hypsarrhythmia. As prompt successful treatment improves chances of short-term seizure control and favorable long-term outcomes, we believe these computational measures may lay the groundwork for more objective diagnosis and treatment assessment in IS. Funding: This study was accomplished with support from UCB Biopharma, the Elsie and Isaac Fogelman Endowment, the Hughes Family Foundation, and the UCLA Children’s Discovery and Innovation Institute.