High-density EEG during electroconvulsive therapy-induced seizures and other brain states: frequency evolution through time

Introduction

The seizure induced during electroconvulsive therapy (ECT) is a central part of its strong clinical effect. Part of the typically used quality parameters of a therapeutic seizure is the postictal suppression of EEG activity. Typical clinical setups however, are limited in their spatial resolution and frequency range. Previous works have described the time course of the frequency development throughout the seizure [1]: After the electrical stimulation, the initial frequency often is described as being of broadband higher frequency, which then slows down over the course of the seizure until it typically and optimally ends abruptly.

Methods:

We recruited 33 depressed patients (17 male, 16 female, median age 38 years, 31 unipolar, 2 bipolar depressed), who were treated with ECT for clinical reasons (n=32 bifrontal stimulation, n=3 right unilateral stimulation, restimulation in 2 cases). During the 4th or 5th treatment of the index series, 60-channel EEG (1000 Hz) was measured in five different states: prior to anaesthesia (resting state with eyes open and eyes closed), immediately before the stimulation (during anaesthesia), during the seizure, and throughout the subsequent postictal suppression period. Anaesthesia was induced either with propofol (n=26, median dose 1.5 mg/kg, range 80-200 mg) or etomidate (n=7, median dose 0.2 mg/kg, range 10-20 mg) and add-on Remifentanil (median total dose 165 ug), and muscle relaxation using succinylcholine (median dose 1.7 mg/kg, range 80-200 mg). The EEG were high-pass filtered @ 0.125Hz and notch filters were applied. Common quantitative definitions based on global field power were used to locate segments corresponding to the 5 states. For electrode quality control, instantaneous time-derivative methods were used. Morlet wavelets were used for multi-resolution time-frequency decomposition.

Results:

During the seizure, within each subject, the frequencies across the scalp were highly homogeneous, with a median Wasserstein distance W=0.19 (20-80% quantiles 0.11-0.27) across the cohort, approximately 3 times lower than in resting state. The spectrum was narrow-band, indicated by the entropy fraction of the spectrum across the cohort of 0.23 (quantiles 0.16-0.27), with rhythmic oscillations starting at approximately 4Hz with a non-monotonic decline over the course of the seizure, typically terminating at around 1.5Hz. In 24/28 cases, the broadband, low-frequency (<1hz) band dominating during anaesthesia emptied compared to post-ictally. For i) 9/28 cases, the emptying occurred already before the end of the stimulus, ii) for 12/28 cases the emptying occurred during the seizure, and iii) for 7/28 the post-ictal emptying was not significant. Postictal voltage amplitudes above 3 Hz were significantly lower compared to all other states.

Conclusion:

The transition into the ECT-evoked seizure is marked by a homogenisation of frequencies and a climb in intensity. Contrasting the preictal state to the postictal, ECT downward shifts intensities across all frequencies, a well-observed phenomenon with reports of links to ECT outcomes. At frequencies lower than 1Hz, which are dominant pre-ictally, the onset of suppression begins already during the seizure, in most cases.

References

1. Brumback R, Staton R (1982): The Electroencephalographic Pattern during Electroconvulsive Therapy. Clinical Electroencephalography, Vol 13(3):145-153