Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Magnetic seizure therapy and electroconvulsive therapy
increase frontal aperiodic activity
Sydney E. Smith3#*, Eena L. Kosik1#, Quirine van Engen1#, Aron T. Hill5,8, Reza Zomorrodi6, Daniel M.
Blumberger6, Zafiris J. Daskalakis7, Itay Hadas7#, Bradley Voytek1-4#
#These authors contributed equally.
Affiliations:
1Department of Cognitive Science, 2Halıcıoğlu Data Science Institute, 3Neurosciences Graduate Program, 4Kavli Institute for Brain
and Mind, 5Cognitive Neuroscience Unit, School of Psychology, Deakin University, Melbourne, Australia, 6Temerty Centre for
Therapeutic Brain Intervention, Centre for Addiction and Mental Health, University of Toronto, Toronto, Ontario, Canada,
7Department of Psychiatry, University of California, San Diego, La Jolla, CA, USA, 8Department of Psychiatry, Central Clinical
School, Monash University, Melbourne, Australia
*e-mail: s1smith@health.ucsd.edu
Data availability
All code used for all analyses and plots are publicly available on GitHub at
https://github.com/voytekresearch/ect-mst. The data collected in this study is not available at this time.
Acknowledgements
Support: NIH National Institute of General Medical Sciences grant R01GM134363-01 (to B.V.)
Thanks: We thank Andrew Bender, Dillan Cellier, Ryan Hammonds, Blanca Martin-Burgos, Michael
Preston, and Trevor McPherson for their advice and feedback on the manuscript.
Author contributions
S.E.S., I.H., and B.V. conceived of the experiments and developed the analyses. I.H., R.Z., A.T.H., Z.J.D.,
and D.M.B., collected the data. S.E.S., E.L.K., Q.vE., and B.V. wrote analysis code and analyzed the data.
S.E.S., E.L.K., Q.vE., I.H., and B.V. wrote the manuscript, and all authors edited the manuscript.
Competing interests A.T.H. was supported by an Alfred Deakin Postdoctoral Research Fellowship. D.M.B.
receives research support from the Canadian Institutes of Health Research (CIHR), National Institutes of
Health – US (NIH), Brain Canada Foundation and the Temerty Family through the CAMH Foundation and
the Campbell Family Research Institute. He received research support and in-kind equipment support for
an investigator-initiated study from Brainsway Ltd. and he was the site principal investigator for three
sponsor-initiated studies for Brainsway Ltd. He received in-kind equipment support from Magventure for
investigator-initiated studies. He received medication supplies for an investigator-initiated trial from
Indivior. He has participated in an advisory board for Janssen. He has participated in an advisory board
for Welcony Inc. Z.J.D. has received research and equipment in-kind support for an investigator-initiated
study through Brainsway Inc and Magventure Inc and industry-initiated trials through Magnus Inc. He
also currently serves on the scientific advisory board for Brainsway Inc. His work has been supported by
1
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.
the National Institutes of Mental Health (NIMH), the Canadian Institutes of Health Research (CIHR), Brain
Canada and the Temerty Family, Grant and Kreutzcamp Family Foundations.
2
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Abstract
Major depressive disorder (MDD) is a leading cause of disability worldwide. One of the most efficacious
treatments for treatment-resistant MDD is electroconvulsive therapy (ECT). Recently, magnetic seizure
therapy (MST) was developed as an alternative to ECT due to its more favorable side effect profile. While
these approaches have been very successful clinically, the neural mechanisms underlying their
therapeutic effects are unknown. For example, clinical “slowing” of the electroencephalogram has been
observed in both treatment modalities. A recent longitudinal study of a small cohort of ECT patients
revealed that observed clinical slowing was better explained by increases in frontal aperiodic activity, an
emerging EEG signal linked to neural inhibition. Here we investigate the role of aperiodic activity in a
cohort of patients who received ECT and a cohort of patients who received MST treatment. We find that
across treatments, frontal aperiodic activity better explains increases in delta band power associated
with clinical slowing, compared to delta oscillations. Increased aperiodic activity is also linked to
therapeutic efficacy, which is suggestive of a potential shared neural mechanism of action across ECT and
MST: an increase in frontal inhibitory activity.
Introduction
Since its development in 1938, electroconvulsive therapy (ECT) has been used as a treatment for mood
disorders including Major Depressive Disorder (MDD) and particularly, treatment-resistant depression
(TRD)1. During a session of ECT, an electrical current is applied to the scalp of an anesthetized patient
which induces a seizure as it passes through the brain. Despite the remission rates between 50-70%2, it
remains one of the least used treatments for depression. Fewer than 1% of patients with TRD receive
ECT due to a combination of fear, stigma, and concerns about cognitive side effects, such as short-term
amnesia3. The search for alternative, yet comparably effective, therapeutic stimulation techniques have
led to the development of treatments like repetitive transcranial magnetic stimulation (rTMS) and more
recently, magnetic seizure therapy (MST).
MST is a more focal treatment that was developed to mimic the therapeutic effects of ECT while
minimizing the adverse side effects. Specifically, MST involves the application of a magnetic field to
produce a seizure in the brain. The first person to receive MST was treated in 20004. Compared to ECT,
MST can produce remission rates as high as 50%5and patients receiving MST experience fewer cognitive
side effects and recover more quickly after the procedure compared to patients receiving ECT5–8(Fig. 1).
Many of the cognitive side effects of ECT are assumed to result from the induced seizure propagating
widely through the cortex, particularly into the medial temporal lobes and hippocampus7,9. Although
both ECT and MST induce a seizure, the characteristics of the induced seizure may differ between
treatments10 (Fig. 1). Specifically, when tested in nonhuman primates, analysis of intracranial electrodes
implanted in the prefrontal cortex and hippocampus revealed that the seizure induced by MST had less
3
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
robust ictal expression than ECT and does not spread to the hippocampus11 . Furthermore, indicators of
pathological neuroplasticity in the hippocampus, including increased cell proliferation and fewer mossy
fibers, were less pronounced in MST than ECT11 . Anatomically realistic biophysical models of electrical
current proliferation in ECT and MST support the findings in primates. In these simulations, compared to
ECT, MST may avoid medial temporal and hippocampal regions, ultimately stimulating a much smaller
portion of the brain which may minimize the negative side effects associated with bilateral ECT12,13.
Beyond ictal expression, post-treatment electroencephalogram (EEG) recordings of both patients
receiving ECT and MST are characterized by clinical “slowing,” seen in increased spectral power in the
delta (1-4 Hz) and theta (4-8 Hz) frequency ranges compared to baseline14–16 . Sometimes, this increase in
spectral power can be observed as apparent high amplitude delta and theta oscillations in the time
domain17,18 . However, this slowing has not been consistently linked to a mechanism of action or clinical
efficacy for either treatment modality19 .
However, previous investigations of band power changes have not considered the contributions of
aperiodic activity to EEG signals. This is important because traditional analysis methods conflate periodic
(oscillatory) activity with aperiodic activity20,21 . This conflation occurs even in the absence of oscillations,
which are not omnipresent, but rather appear infrequently in short bursts22–24. So even if there are no
oscillations, a large aperiodic signal can appear very similar to slowing in the EEG, as we have recently
demonstrated25. This effect occurs because the EEG signal is a mix of oscillations and aperiodic activity,
where oscillations are defined by concentrated power with a specific, narrow frequency band. In
contrast, aperiodic activity manifests as a broadband phenomenon, where power decreases
exponentially as a function of frequency (1/fχscaling). This exponent is parameterized by χ(Fig. 2A),
which naturally arises from the physiology of the EEG signal26,27. The EEG signal is largely composed of
transmembrane postsynaptic currents that are characterized by their double-exponential form, the
Fourier Transform of which naturally gives rise to 1/fχscaling. The aperiodic exponent has been shown to
at least partially capture the relative excitatory and inhibitory contributions to the local field
potential28,29, where an increase in the aperiodic exponent reflects a “steeping” rotation of the power
spectrum. This corresponds to a shift toward greater inhibitory drive, and manifests as a large increase in
low frequency power with a concomitant decrease in high frequency power (Fig. 2B).
Our recent investigation into longitudinal changes in aperiodic activity throughout a course of ECT
revealed that many of the observations of increased frontal delta band power could be better explained
by increases in frontal aperiodic activity25. Specifically, in that report, we found that the aperiodic
exponent significantly increased longitudinally throughout the course of treatment. Furthermore, both
aperiodic exponent at baseline and the magnitude of the change in exponent throughout treatment
were related to treatment response, as measured by the Quick Inventory of Depressive Symptomatology
- Self Report (QIDS-SR). In comparison, there were no longitudinal effects in delta band power or
oscillatory power. This was interpreted as evidence that longitudinal increases in frontal aperiodic
activity may better explain clinical slowing observed in seizure-inducing treatments.
Here, we sought to replicate and extend that smaller (n = 9), longitudinal study in a larger sample of two
independent datasets, one from a study of patients receiving ECT (n = 22)30 and the other from a
4
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
registered clinical trial of patients receiving MST (n = 23)31(Clinicaltrials.gov NCT01596608). We
hypothesized that this recent finding would generalize to interventional findings of slowing in both ECT
and MST. To test this hypothesis we compare measures of aperiodic activity, oscillations, and canonical
band power. For each treatment modality, resting state EEG was collected at baseline and after
completing a standard treatment, and analyses were restricted to frontal channels for replicability25. We
find that in both ECT and MST treatment conditions, aperiodic exponent in frontal regions increases
significantly and this change better explains observed delta (1-4 Hz) band power increases than power of
a delta oscillation. We also find that theta (4-8 Hz) oscillations emerge following both ECT and MST. The
magnitude of both of these effects was greater in ECT than in MST. Post-ECT EEG also contains more
delta (1-4 Hz) oscillations and increases in alpha (8-12 Hz) oscillation power compared to baseline, a
result not found in MST. These results highlight the relevance of changes in frontal aperiodic and
oscillatory activity to stimulation treatments for depression and present promising opportunities for
future research to better understand and differentiate the mechanisms of clinical efficacy and cognitive
side effects.
Results
Clinical effects
For patients that received ECT treatment, the severity of depressive symptomatology as assessed by the
clinician using the Hamilton Depression Rating Scale (HAMD-17) decreased significantly (pre = 24.26,
post = 13.21, t(18) = 5.94, dz= 2.07, p = 1.3 x 10-5) (Fig. 1, left, and Table 1). For patients that received
MST treatment, HAMD-24 scores also decreased significantly (pre = 28.13, post = 21.40, t(14) = 4.14, dz=
0.93, p = 9.97 x 10-4) (Fig. 1, right, and Table 1). Compared to MST, patients receiving ECT demonstrated
greater clinical symptom improvements, as quantified by the relative change from baseline in HAMD-17
or HAMD-24 for ECT and MST, respectively (ECT = 44.02%, MST = 24.0%, t(31.97) = 2.23, dz= 0.75 , p =
0.03).
Treatment-related EEG effects: ECT
Compared to baseline, patients who receive a full course of ECT exhibit significant increases in aperiodic
activity, as measured by the aperiodic exponent of power spectra in frontal electrodes, which become
visibly steeper (pre = 0.89 µV2Hz-1, post = 1.56 µV2Hz-1, t(21) = -8.12, dz= 1.85, p = 6.15 x 10-8) (Fig. 3B).
We also observed concomitant increase in delta band power (pre = -11.91 µV2Hz-1, post = -10.97 µV2Hz-1,
t(21) = -8.03, dz= 1.96, p = 7.69 x 10-8) (Fig. 3C). However, very few patients had detectable delta
oscillation peaks both at baseline and after treatment, which means that it is unlikely that clinical
slowing reflects actual oscillatory processes (Fig. 3D). To determine whether the increases in delta band
power were driven primarily by aperiodic activity or delta oscillation peaks, we computed delta
abundance, defined as the fraction of frontal electrodes exhibiting a delta oscillation peak in their power
5
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
spectra, above the aperiodic signal. We found that delta abundance increases significantly post-ECT (pre
= 0.03, post = 0.26, t(21) = -3.16, dz= 0.79, p = 4.75 x 10-3) (Fig. 3E).
To further investigate how much of the observed changes in frontal delta band power, or slowing, was
driven by aperiodic activity compared to delta oscillations, we performed a multiple linear regression to
examine the contributions of delta oscillations and aperiodic activity to delta band power. This regression
was significant overall (R2= 0.59, F(2, 19) = 13.41, p = 2.33 x 10-4), with the aperiodic exponent more
significantly contributing to observations of increased delta band power (β= 0.83, p = 0.003) compared
to delta abundance (β= 0.48, p = 0.98). That is, traditional band power definitions of delta, which do not
seek to examine whether or not true oscillations are present, are better explained by non-oscillatory
aperiodic activity in this sample. Furthermore, we performed similar multiple linear regressions to see
whether frontal theta and alpha band power are better predicted by the change in aperiodic exponent,
or the change in abundance of the respective frequency bands. Theta band power was also significantly
related to a change in exponent, but not by the change in theta oscillation abundance. In contrast, alpha
band power was not significantly related to the exponent, instead was related to a change in abundance
and a change in oscillation power. This aligns with prior observations that alpha band power changes in
human EEG are largely driven by actual alpha oscillations, whereas power in other bands is more related
to aperiodic activity32. Full details of results are in the supplementary materials.
We also observed changes in frontal power in the theta and alpha ranges in patients who receive ECT.
For patients whose EEG power spectra contained theta oscillation peaks before and after treatment, we
observed significant increases in aperiodic-adjusted theta oscillation power (pre = 0.34 µV2, post = 0.75
µV2, t(9) = -6.03, dz= 2.03, p = 1.94 x 10-4) (Fig. 4A). Similar to what we observed for delta, the fraction of
electrodes that exhibited a theta oscillation peak also increased significantly with treatment (pre = 0.26,
post = 0.63, t(21) = -3.66, dz= 1.00, p = 1.46 x 10-3) (Fig. 4B). When we repeated this analysis in the alpha
band, we found that alpha oscillation peaks decrease in power (pre = 1.21 µV2, post = 0.83 µV2, t(21) =
3.81, dz= 0.96, p = 1.03 x 10-3) and abundance (pre = 1.0, post = 0.92, t(21) = 2.41, dz= 0.73, p = 0.025)
post-ECT compared to baseline (Fig. 4 D-E). This result is notable because alpha band power increases
post-ECT when measured using the canonical bandpass approach (pre = -11.65 µV2Hz-1, post = -11.42
µV2Hz-1, t(21) = -2.30, dz= 0.47, p = 0.031), highlighting the importance of spectral parameterization to
disambiguate the contributions of periodic and aperiodic activity to band power, which separates band
power into the periodic and aperiodic components.
Treatment-related EEG effects: MST
Similar to ECT, patients who were treated with MST (see Methods) exhibit a significant increase in the
frontal aperiodic exponent compared to baseline (pre = 0.97 µV2Hz-1, post = 1.16 µV2Hz-1, t(22) = -3.17, dz
= 0.80, p = 4.42 x 10-3) (Fig. 3G). We also observe significant increases in delta band power (pre = -11.87
µV2Hz-1, post = -11.64 µV2Hz-1, t(22) = -2.39, dz= 0.58, p = 0.03), but no significant change in delta
abundance (pre = 0.01, post = 0.03, t(22) = -2.05, dz= 0.23, p = 0.18) compared to baseline (Fig. 3 H, J).
Furthermore, the multiple linear regression to relate delta band power to the aperiodic exponent and
delta abundance was overall significant (R2= 0.72, F(2, 20) = 26.21, p = 2.58 x 10-6). Increases in aperiodic
6
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
activity (β= 1.27, p = 5.0 x 10-6) better explain increases in delta band power than do changes in delta
abundance (β= -1.66, p = 0.24), indicating that aperiodic activity is likely driving observed clinical
slowing in MST as well. Similar to our ECT analysis, we performed multiple linear regressions to relate
theta and alpha band power to the change in aperiodic exponent, or to the change in abundance of the
respective frequency bands. We found that the aperiodic exponent was a significant predictor for both
theta and alpha band power, while theta and alpha abundance did not significantly predict band power
in each respective frequency band.
Although we do not observe significant changes in delta abundance, there were significant changes in
theta oscillations. Specifically, both theta oscillation peak power (pre = 0.43 µV2, post = 0.63 µV2, t(12) =
-4.71, dz= 1.10, p = 6.39 x 10-4) and abundance (pre = 0.31, post = 0.57, t(22) = -2.74, dz= 0.65, p = 0.012)
increase significantly post-MST compared to baseline (Fig. 4 F-G). This effect was similar to what we
observe in ECT. However, unlike ECT, we observe no significant changes in alpha oscillation peak power
(pre = 1.03 µV2, post = 1.02 µV2, t(22) = 0.21, dz= 0.04, p = 0.83), nor in alpha abundance (pre = 1, post =
0.89, t(22) = 2.00, dz= 0.59, p = 0.058), though the effect is marginal, post-MST (Fig. 4 I-J).
Lastly, we performed a t-test on the change in exponent between ECT and MST. The exponent change in
ECT was significantly higher than for MST (ECT = 0.67 µV2Hz-1, MST = 0.19 µV2Hz-1, t(39.01) = 4.66, dz=
1.40, p = 3.7 x 10-4).
Clinical improvement and the spectral features in ECT and MST
To determine whether the observed changes in frontal aperiodic and oscillatory activity were related to
overall treatment response independent of treatment modality, we computed a multiple linear
regression to relate the relative magnitude of clinical improvement, as measured by the HAM-D, from
the features that changed significantly after both ECT and MST, theta abundance and aperiodic
exponent. We also included a term to account for a statistical interaction between change in aperiodic
exponent and the number of treatments received. The overall regression was significant (R2= 0.24, F(3,
29) = 3.01, p = 0.046). We found that changes in aperiodic exponent (β= 0.67, p = 0.007) were predictive
of treatment response (Fig. 5) but that theta abundance was not (β= -0.02, p = 0.91). In this model,
patients who exhibit the greatest increases—or “steepening”—in aperiodic exponent after either ECT or
MST treatment showed the greatest improvements in depression symptom severity, as measured by
HAMD relative to baseline.
Furthermore, to determine whether any of the absolute, baseline EEG features could be identified as
potential biophysical indicators of clinical treatment response, we performed a similar multiple linear
regression using baseline measures of aperiodic exponent and theta abundance. Although trending, the
model did not significantly predict treatment outcome from baseline measures (R2= 0.13, F(2, 31) = 2.34,
p = 0.11). Within the model, baseline aperiodic exponent (β= -0.43, p = 0.040) significantly predicted
clinical improvement (Fig. 5), but baseline theta abundance did not (β= 0.09, p = 0.55). In addition, we
do find a significant correlation between baseline aperiodic exponent and treatment response across
MST and ECT datasets (spearman’s r = -0.36, p = 0.039), though this relationship should be interpreted
with caution, as it is a simple correlation and has not been corrected for multiple comparisons. This
7
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
relationship indicates that patients who begin treatment with smaller aperiodic exponents—or “flatter”
spectra—tend to be more responsive to treatment.
Discussion
The observation that frontal aperiodic activity increases after a course of ECT supports our recent finding
from a smaller, longitudinal study25. Furthermore, similar observations after a course of MST identify the
increase in aperiodic activity as a potentially informative physiological change shared by both of these
seizure-based treatments for depression. For both ECT and MST, this increase in aperiodic activity is a
more parsimonious explanation for observations of clinical slowing than delta band power or delta
oscillations. Also, MST and ECT both induce increases in the power and abundance of theta oscillations.
In addition to the changes in aperiodic activity and theta oscillations, ECT is associated with increases in
the abundance of delta oscillations and decreases in the power of alpha oscillations. These effects are
not observed in MST.
Aperiodic activity has been widely associated with behavioral and disease states, such as cognitive and
perceptual task performance33–36, development37, aging38, anesthesia39, ADHD40, and schizophrenia41.
Furthermore, changes in aperiodic activity, like those observed in these two populations with MDD, have
been associated with the physiological effects of deep brain stimulation as a treatment for MDD42. It is
hypothesized that these aperiodic changes in the brain are related to the balance of excitation (E) and
inhibition (I) based on simple computational models of the local field potential28, complex microcircuit
models43 , and experimental manipulations of EI balance using optogenetics29. The changes in aperiodic
activity seen in patients undergoing ECT and MST, specifically increases in aperiodic exponent visible as a
“steepening” of the power spectrum, are associated with relative increases in inhibitory activity.
Therapeutic interventions that potentially increase inhibitory activity are particularly relevant in light of
the cortical inhibition theory of depression. According to this theory, patients with MDD have insufficient
inhibitory activity in various brain regions, particularly frontal cortices44. Post-mortem tissue analyses
have revealed that these patients have pathologically reduced numbers of inhibitory, GABAergic
neurons45. Specifically, these patients have reduced somatostatin-expressing (SST) interneurons in
prefrontal cortices46–48. Simulated biophysical models of human microcircuits with reduced SST activity
produce LFP signals with flatter power spectra and lower aperiodic exponents compared to control
simulations of microcircuits with healthy SST populations49. Deficits in these inhibitory interneuron
populations could cause pathological dysfunction in prefrontal EI balance. Because prefrontal cortices
play an essential role in regulating EI balance throughout distributed networks in the brain50,
dysfunctional inhibition in prefrontal regions could lead to widespread disruptions, including in limbic
structures51 and the serotonergic and noradrenergic systems targeted by antidepressant medications44.
Moreover, transcranial magnetic stimulation (TMS) concurrent with EEG also demonstrates decreased
activation localized to frontal regions after successful neuromodulatory treatments for depression, and
this attenuation is attributed to strengthening of inhibitory circuits9,52.
8
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Furthermore, we found that the degree to which a patient’s frontal aperiodic activity increases
post-treatment is related to the magnitude of therapeutic response, when controlling for the interaction
of the aperiodic exponent with the number of treatments received. It should be noted that for the MST
dataset, 61% of patients received exactly 24 treatments, the maximum under the study protocol, leading
to an imbalanced number of treatments. However, the errors of the multiple linear regression model are
normally distributed, thus mitigating the statistical effect of this imbalance. Previous analysis of the same
ECT dataset has shown that the number of treatments correlates to more power in the slower
frequencies. However, based on our re-analysis using spectral parameterization, we can interpret this
slowing as an increase in the aperiodic activity16.Although interventional, this observation is in keeping
with our recent longitudinal study and provides further evidence that aperiodic activity might be
relevant to the therapeutic efficacy of ECT and MST. This clinical finding, combined with the
computational and experimental evidence that identify aperiodic activity as an indicator of EI balance,
suggest a promising mechanism of action for these stimulation treatments for depression. Specifically,
we posit that ECT and MST ameliorate depressive symptoms by restoring healthy levels of inhibition in
frontal regions, as measured by changes in aperiodic activity. Increasing levels of inhibition as measured
through aperiodic activity align with observations of the anticonvulsant effects of ECT, with seizure
induction threshold progressively increasing throughout a course of treatment53.
Beyond identifying the physiological mechanism of ECT and MST, it is also vital to consider differences in
the cognitive and physical side effects of these treatments. Previous studies suggest that the magnitude
of therapeutic effects of ECT are unrelated to the severity of cognitive side effects54, suggesting that
these two phenomena are potentially dissociable. For instance, some theorists suggest that the
therapeutic mechanisms of ECT are related to increases in delta power—an effect we argue is better
explained by increases in aperiodic activity— and that cognitive side effects are driven mostly by theta
power55. The observation that both ECT and MST produce increases in aperiodic activity that are related
to clinical response supports the idea of a shared therapeutic mechanism. This effect is especially
notable because the emergence of theta oscillations does not predict clinical efficacy. The differences we
observed between ECT and MST in theta, as well as in delta and alpha oscillations, provide promising
avenues for future investigation into the differential cognitive effects of these treatments. Theta
oscillations, classically linked to memory56, are of special interest here. The difference between ECT and
MST in the post-treatment emergence of delta oscillations is also notable7and might be related to the
amplitude and spatial distribution of the ictal activity during the induced seizure (Fig. 1). However,
further investigation is needed to more precisely describe the contributions of oscillations and periodic
activity to the therapeutic and cognitive effects of ECT and MST.
Unfortunately, the patients in the dataset included in this study were not thoroughly evaluated for
cognitive and physical side effects so our analyses are limited to investigating therapeutic mechanisms.
Furthermore, the data in this interventional study and the preceding longitudinal study can only speak to
the short term effects of ECT and MST. More evidence is needed to determine how long changes in
aperiodic and oscillatory activity persist post-treatment and if the longevity of these changes in EEG are
linked to rates of MDD relapse. Furthermore, although we found similarities across ECT and MST, the
increases in aperiodic exponent and theta oscillatory power were stronger in the patients who
9
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
underwent ECT. One potential explanation for this difference in magnitude could be that the ECT
patients collected post-treatment EEG within 48 hours after the last session, while the MST patients
waited longer on average to receive their post-treatment EEG, with an average of 3.81 days after the last
session.
Future investigations can also explore the relationship of hemispheric differences and stimulation
laterality to therapeutic efficacy and cognitive side effects. Because the vast majority of ECT and MST
patients included in the study received bilateral stimulation, our analyses combined effects across frontal
hemispheres. However, measures of hemispheric differences have been relevant to studies of
depression, especially frontal alpha asymmetry, despite this measure being called into question by
several recent meta analyses57,58. More evidence is needed to uncover the role of hemispheric
differences in aperiodic and periodic activity in ECT and MST.
The current study shows that therapy-induced changes in frontal EEG aperiodic activity predicts clinical
improvements in MDD patients undergoing ECT and MST treatment. Although not yet a viable clinical
biomarker, the observed changes in aperiodic activity here provide further support for the cortical
inhibition theory of depression47. Our results hint at potential physiological mechanisms of depression,
and why ECT and MST are useful treatments, though much more physiological evidence is required.
Furthermore, we show some important differences in the two treatments: Although MST may be less
efficacious compared to ECT in terms of HAM-D improvement, MST still provides a significant
therapeutic benefit, with potentially fewer adverse side effects.
10
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Fig. 1| ECT vs. MST. This figure highlights the most important similarities and differences between
electroconvulsive therapy (ECT), and magnetic seizure therapy (MST). Both treatments are typically only
used on patients with treatment-resistant depression and involve inducing a seizure, either with an
electrical current or a magnetic field. The main difference is that ECT has a more global spread to
subcortical structures and hippocampus, whereas MST affects more local cortical structures. However,
both treatment types significantly reduce depression ratings, as measured by the HAMD-17 for ECT (pre
= 24.26, post = 13.21, t(18) = 5.94, dz= 2.07, p = 1.3 x 10-5) and the HAMD-24 for MST (pre = 28.13, post =
21.40, t(14) = 4.14, dz= 0.93, p = 9.97 x 10-4).
11
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Fig. 2| Using spectral parameterization to disambiguate periodic and aperiodic contributions to delta
band power. (A) Simulated power spectrum illustrating parameterized spectra. Unlike traditional band
power measures that conflate periodic and aperiodic activity, spectral parameterization defines
oscillatory power as relative power above the aperiodic component (pink dashed line). (B) Increases in
the aperiodic exponent can cause apparent increases in total (T) band power, while power relative (R) to
the aperiodic component remains unchanged. We see this here in the power spectrum averaged over
frontal electrodes from one patient who exhibits an increase in exponent with no delta oscillation
changes after treatment. (C) True increases in oscillatory power show increases in both total power and
relative power. We see this here in the power spectrum averaged over frontal electrodes from one
patient who exhibits an increase in delta oscillation power after treatment in addition to an increase in
exponent. (D) Delta in the EEG trace vs. aperiodic activity. EEG with delta oscillations (where a delta peak
is present in the spectra) is visibly different from EEG with only aperiodic activity in the delta band.
12
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Fig. 3| EEG results - Aperiodic vs. delta band power slowing
Spectral differences in aperiodic exponent and delta oscillations in ECT (top) and MST (bottom). (A) Raw
power spectra averaged across channels for each patient pre- and post-ECT. Bolded spectra represent
average across patients. (B) Comparison of aperiodic exponent pre- and post-treatment (pre = 0.89
µV2Hz-1, post = 1.56 µV2Hz-1, t(21) = -8.12, dz= 1.85, p = 6.15 x 10-8), (C) total power in the delta band (pre
= -11.91 µV2Hz-1, post = -10.97 µV2Hz-1, t(21) = -8.03, dz= 1.96, p = 7.69 x 10-8), (D) aperiodic-adjusted
oscillatory power in the delta band (pre = 0.16 µV2Hz-1, post = 0.61 µV2Hz-1), and (E) abundance of delta
oscillations (pre = 0.03, post = 0.26, t(21) = -3.16, dz= 0.79, p = 4.75 x 10-3). (F) Raw power spectra
averaged across channels for each patient pre- and post-MST. Bolded spectra represent average across
patients. (G) Comparison of aperiodic exponent pre- and post-treatment (pre = 0.97 µV2Hz-1, post = 1.16
µV2Hz-1, t(22) = -3.17, dz= 0.80, p = 4.42 x 10-3), (H) total power in the delta band (pre = -11.87 µV2Hz-1,
post = -11.64 µV2Hz-1, t(22) = -2.39, dz= 0.58, p = 0.03), (I) aperiodic-adjusted oscillatory power in the
delta band (pre = 0.63 µV2Hz-1, post = 0.54 µV2Hz-1), and (J) abundance of delta oscillations (pre = 0.01,
post = 0.03, t(22) = -2.05, dz= 0.23, p = 0.18).
13
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
*
Fig. 4| EEG results - Emergence of theta and alpha oscillations
Changes in theta and alpha oscillations in ECT (top) and MST (bottom). (A) Comparison between theta
oscillation power (4-7Hz) pre- and post-treatment (pre = 0.34 µV2, post = 0.75 µV2, t(9) = -6.03, dz= 2.03,
p = 1.94 x 10-4),(B) theta abundance (pre = 0.26, post = 0.63, t(21) = -3.66, dz= 1.00, p = 1.46 x 10-3). (C)
Power spectrum from frontal electrode F8 in a patient who received ECT. (D) Comparison between pre-
and post-ECT alpha oscillation power (7-12 Hz) (pre = 1.21 µV2, post = 0.83 µV2, t(21) = 3.81, dz= 0.96, p =
1.03 x 10-3), and (E) alpha abundance (pre = 1.0, post = 0.92, t(21) = 2.41, dz= 0.73, p = 0.025). (F)
Comparison between theta oscillation power (4-7Hz) pre- and post-treatment (pre = 0.43 µV2, post =
0.63 µV2, t(12) = -4.71, dz= 1.10, p = 6.39 x 10-4), (G) theta abundance (pre = 0.31, post = 0.57, t(22) =
-2.74, dz= 0.65, p = 0.012). (H) Power spectrum from frontal electrode F8 in a patient who received MST.
(I) Comparison between pre- and post-MST alpha oscillation power (7-12 Hz) (pre = 1.03 µV2, post = 1.02
µV2, t(22) = 0.21, dz= 0.04, p = 0.83), and (J) alpha abundance (pre = 1, post = 0.89, t(22) = 2.00, dz=
0.59, p = 0.058).
14
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Fig. 5| Partial regression analysis - Exponent predicting treatment outcome
Change in aperiodic exponent and baseline aperiodic exponent significantly predict clinical outcome. (A)
Partial regression showing a positive relationship between patients’ change in aperiodic exponent, and
clinical improvement, as measured by the ratio of pre- and post-treatment HAMD-17 and HAMD-24, for
ECT and MST, respectively (β= 0.67, p = 0.007). Here, patients whose aperiodic exponents change the
most (greatest spectral steepening) respond best to treatment. (B) Partial regression showing a negative
relationship between patients’ baseline aperiodic exponent, and clinical improvement, also measured by
the ratio of pre- and post-treatment HAMD-17 and HAMD-24, for ECT and MST, respectively (β= -0.43, p
= 0.040). Here, patients who begin treatment with lower aperiodic exponents (flatter spectra) at baseline
respond best to treatment.
Table 1 | ECT and MST dataset details for patients included in this paper
ECT dataset details were from REF #16 and further correspondence. MST dataset details were from REF
#9 and further correspondence.
ECT_n = 22; MST_n = 23
ECT: Mean and SD
MST: Mean and SD
Age
47.29 ± 16.75
46.13 ± 11.04
Gender (Male/Female)
9/14
16/15
HAM-D pre
24.25 ± 3.67 (HAMD-17)
28.52 ± 5.50 (HAMD-24)
HAM-D post
13.21 ± 6.56 (HAMD-17)
21.40 ± 8.14 (HAMD-24)
Number of treatments received
13.87 ± 5.32
18.80 ± 7.40
Responders
60.87%
45.16%
Medications
15
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Antidepressant
91%
61%
Antipsychotic
32%
26%
Anxiolytics
45%
17%
Stimulants
9%
13%
Sedative
27%
35%
Methods
Participants
Twenty-two patients who received a diagnosis of MDD as per the Diagnostic and Statistical Manual
(DSM-IV) were included in this study for ECT. Twenty-three patients with the same diagnosis were
included in this study for MST. Patients were considered as having treatment-resistant depression.
Written informed consent was provided by all patients. Ethical approval was granted from the Centre for
Addiction and Mental Health (CAMD) research ethics committee in accordance with the Declaration of
Helsinki. A complete list of inclusions and exclusions criteria is provided in REF #30.
Electroconvulsive therapy
Patients received ECT 2-3 times per week. Square wave pulses were delivered using an open label
protocol with a brief-pulse device (MECTA Corporation, Lake Eswego, OR). Patients started their
treatment with either right unilateral ultra-brief pulse width ECT, or brief pulse width (1.0 msec)
bi-temporal ECT based on the preference of the treating physician and the patient. Electrode placement
was in accordance with American Psychiatric Association guidelines. Patients who received unilateral
treatment were later switched to bi-temporal ECT if they showed an initial poor response to treatment.
Anesthesia was induced by administering methohexital for sedation, and succinylcholine for muscle
relaxation. Treatment completion was based on clinical factors, patient response, the patient's desire to
discontinue treatment, or the most responsible physician's clinical judgment. More details about this
process can be found in REF #30.
Magnetic seizure therapy
Patients received MST 2-3 times per week. A twin coil (Twin Coil-XS) was used with a MagPro MST
stimulator (Magvenure, Denmark). The two coils were placed bilaterally over the prefrontal cortex,
approximating F3 and F4 locations (international 10-20 system). Anesthesia was induced by
administering methohexital sodium or methohexital plus remifentanil for sedation, and succinylcholine
for muscle relaxation. Treatment completion was based on clinical factors, defined as a remission if the
HAMD-24 score < 10 and greater than 60% reduction in depressive symptoms using HAMD-24 scale, or a
16
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
total of 24 treatments were administered. 61% of patients in this dataset received the maximum 24
treatments. More details can be found in REF #5,15.
Data acquisition
EEG data were collected within a week before patients started their treatment. For the ECT dataset, post
treatment resting state EEG was collected within 48h after their last treatment. Whereas for the MST
dataset, post treatment resting state EEG were collected on average 3.81 days (SD 3.86)15 after their last
treatment. A total of 10 minutes of resting state data with eyes closed were collected pre- and
post-treatment. A 64-electrode cap (Neuroscan Quik-Cap) containing sintered Ag/AgCl electrodes
connected to a SynAmpsˆ2 amplifier (Neuroscan, Compumedics, USA) was used for all recordings (online
reference and ground electrodes located at the vertex, and just posterior to Fz, respectively).
Impedances were maintained below 5kΩ throughout the recordings.
Clinical measures
Demographic and medication information were recorded at baseline during clinical interview (Table 1).
For the ECT dataset, the primary clinical measure was the 17-item Hamilton Depression Rating Scale
(HAMD-17), which was completed before the first treatment sessions, and within 48h after the last. For
the MST dataset, the 24-item HAM-D was used. Because these datasets were collected independently,
we assessed clinical improvement as a ratio of the change in depression severity relative to baseline as
measured by the HAMD-17 or HAMD-24 for the ECT and MST datasets, respectively.
EEG pre-processing
EEG data was first downsampled to 1kHz. Then bad electrodes were identified and removed based on
the presence of excess noise by inspecting the raw time series and the power spectra per electrode.
After this, the data was re-referenced with the average method. Then, a FIR high-pass filter of 0.5 Hz was
applied with a Hamming window. Fast ICA was used with 15 components to remove eye movements, eye
blinks, and other non-neural artifacts. Lastly, the bad electrodes were interpolated.
EEG Spectral Parameterization
Power spectra were computed per patient for each electrode from the continuous EEG data using
Welch's method, with a Hamming window of 2 seconds, and 1 second overlap between windows.
The spectral parameterization model was fit to each power spectrum between 1 and 30 Hz,
without a knee, and oscillatory peaks were defined as peaks that surpassed a threshold of 0.05 µV2
above the aperiodic component. A maximum of 12 peaks per fit with a minimum band width of 1 Hz and
maximum of 8 Hz. On average, 3 peaks were found per power spectra, both pre- and post-treatment.
Furthermore, the three frequency ranges of interest were delta (1 - 4 Hz), theta (4 - 7 Hz), and alpha (7 -
12 Hz). The peak with the highest power was selected within each frequency range. Model fits (R2) below
0.8 were excluded from further analysis. One patient in the ECT dataset was completely removed from
further analysis, due to excessive noise over the majority of the electrodes, which caused model fits
below the threshold.
17
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Canonical band power was calculated in addition to the metrics extracted using spectral
parameterization methods for the purpose of comparing methodological approaches to quantifying
spectral power. Band power was computed as the mean of spectral power in each frequency band (delta,
theta, and alpha).
Electrodes of interest for further analysis were frontal (FP1, FPZ, FP2, AF3, AF4, F7, F5, F3, F1, FZ,
F2, F4, F6, F8, FC5, FC3, FC1, FCZ, FC2, FC4, FC6). Further analysis was restricted to these electrodes
because previous investigations found strong effects in frontal regions25 and because both ECT and MST
treatments target frontal cortical regions for stimulation. These electrodes were chosen to maintain
consistency between the analyses in this study and the exploratory study in REF #25, and because prior
research shows EEG effects of MDD treatment are strongest on frontal electrodes59,60. Prior to statistical
analysis and visualization, features of interest were averaged across all included EEG electrodes for each
patient.
Predicting band power based on aperiodic exponent and oscillation power
Band power is traditionally used to compute power within certain frequency ranges. Therefore, we
wanted to include a regression to see whether it is the aperiodic exponent or actual oscillation power
within a frequency range that predicts these band power measures. We did this for three frequency
ranges of interest: delta, theta and alpha. For these regressions we used the change in exponent and the
change in oscillation abundance. Furthermore for alpha power, because the abundance is so high, we
also included oscillation power as a predictor. Thus, the formula for predicting delta band power is
ε
∆𝑑𝑒𝑙𝑡𝑎 𝑏𝑎𝑛𝑑 𝑝𝑜𝑤𝑒𝑟= β0+β1∆𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡+β2∆𝑑𝑒𝑙𝑡𝑎 𝑎𝑏𝑢𝑛𝑑𝑎𝑛𝑐𝑒+
The formula for predicting theta band power is
ε
∆𝑡ℎ𝑒𝑡𝑎 𝑏𝑎𝑛𝑑 𝑝𝑜𝑤𝑒𝑟= β0+β1∆𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡+β2∆𝑡ℎ𝑒𝑡𝑎 𝑎𝑏𝑢𝑛𝑑𝑎𝑛𝑐𝑒+
Last, the formula for predicting alpha band power is
ε
∆𝑎𝑙𝑝ℎ𝑎 𝑏𝑎𝑛𝑑 𝑝𝑜𝑤𝑒𝑟= β0+β1∆𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡+β2∆𝑎𝑙𝑝ℎ𝑎 𝑎𝑏𝑢𝑛𝑑𝑎𝑛𝑐𝑒+β2∆𝑎𝑙𝑝ℎ𝑎 𝑜𝑠𝑐𝑖𝑙𝑙𝑎𝑡𝑖𝑜𝑛 𝑝𝑜𝑤𝑒𝑟+
Predicting clinical outcome based on EEG spectral parameters
We used ordinary least squares regression to test whether and which changes in EEG spectral
parameters predict clinical outcome. The dependent variable is the ratio of improvement on HAM-D
scores. The predictors were the difference between pre- and post-treatment for the aperiodic exponent,
theta abundance, and the interaction between exponent and the number of treatments:
ε
𝐻𝐴𝑀𝐷𝑝𝑟𝑒−𝐻𝐴𝑀𝐷𝑝𝑜𝑠𝑡
𝐻𝐴𝑀𝐷𝑝𝑟𝑒 = β0+β1∆𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡+β2∆𝑡ℎ𝑒𝑡𝑎 𝑎𝑏𝑢𝑛𝑑𝑎𝑛𝑐𝑒+β3(∆𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡×𝑁𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑)+
18
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Furthermore, we also looked into EEG spectral parameters at baseline that could predict treatment
outcome. For this we also looked at the ratio of improvement on HAMDscores, but without the
interaction between exponent and number of treatments received:
ε
𝐻𝐴𝑀𝐷𝑝𝑟𝑒−𝐻𝐴𝑀𝐷𝑝𝑜𝑠𝑡
𝐻𝐴𝑀𝐷𝑝𝑟𝑒 = β0+β1𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑝𝑟𝑒+β2𝑡ℎ𝑒𝑡𝑎 𝑎𝑏𝑢𝑛𝑑𝑎𝑛𝑐𝑒𝑝𝑟𝑒+
Statistical analysis
The dependent variables considered for analysis from the EEG signal are the aperiodic exponent, delta
oscillation power, delta band power, theta oscillation power, theta band power, alpha oscillation power,
and alpha band power. Furthermore, we calculated the fraction of frontal electrodes containing an
oscillation per frequency band of interest (abundance).
All these parameters were calculated before the treatment started (pre), and after the treatment
was completed (post). Normality was checked using the Shapiro-Wilk test. A paired t-test was performed
on the pre- vs post-variables if the data was normally distributed, otherwise, a permutation test was
used to determine statistical significance. The permutation null distribution was created by randomly
changing the sign of the post- minus pre-features for 10,000 iterations. Additionally, t-tests were
performed on the clinical outcome, and the change in exponent between ECT and MST, with a correction
on the degrees of freedom for unequal sample sizes where appropriate.
Software
All EEG (pre-)processing, statistical analyses, and plotting was performed in Python (3.9.7), using MNE
0.24.1)61, Spectral Parameterization (FOOOF; 1.0.0)20, Pandas (1.3.2)62, Pingouin (0.5.0)63, Statsmodels
(OSL) (0.13.1)64, Matplotlib (3.4.2)65, and Seaborn (0.11.2)66.
References
1. The Practice of Electroconvulsive Therapy: Recommendations for Treatment, Training, and
Privileging. (American Psychiatric Association, 2001).
2. UK ECT Review Group. Efficacy and safety of electroconvulsive therapy in depressive
disorders: a systematic review and meta-analysis. The Lancet 361, 799–808 (2003).
3. Wilkinson, S. T., Agbese, E., Leslie, D. L. & Rosenheck, R. A. Identifying Recipients of
19
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Electroconvulsive Therapy: Data From Privately Insured Americans. Psychiatr. Serv. 69,
542–548 (2018).
4. Lisanby, S. H., Schlaepfer, T. E., Fisch, H.-U. & Sackeim, H. A. Magnetic Seizure Therapy of
Major Depression. Arch. Gen. Psychiatry 58, 303 (2001).
5. Daskalakis, Z. J. et al. Magnetic seizure therapy (MST) for major depressive disorder.
Neuropsychopharmacology 45, 276–282 (2020).
6. Lisanby, S. H. Update on Magnetic Seizure Therapy: A Novel Form of Convulsive Therapy:
J. ECT 18, 182–188 (2002).
7. Lisanby, S. H., Luber, B., Schlaepfer, T. E. & Sackeim, H. A. Safety and Feasibility of
Magnetic Seizure Therapy (MST) in Major Depression: Randomized Within-Subject
Comparison with Electroconvulsive Therapy. Neuropsychopharmacology 28, 1852–1865
(2003).
8. McClintock, S. M., Tirmizi, O., Chansard, M. & Husain, M. M. A systematic review of the
neurocognitive effects of magnetic seizure therapy. Int. Rev. Psychiatry 23, 413–423 (2011).
9. Hadas, I. et al. Subgenual cingulate connectivity and hippocampal activation are related to
MST therapeutic and adverse effects. Transl. Psychiatry 10, 392 (2020).
10. Backhouse, F. A. et al. Characteristics of ictal EEG in Magnetic Seizure Therapy at various
stimulation frequencies. Clin. Neurophysiol. 129, 1770–1779 (2018).
11. Lisanby, S. H. et al. Chapter 9 Neurophysiological characterization of magnetic seizure
therapy (MST) in non-human primates. in Supplements to Clinical Neurophysiology vol. 56
81–99 (Elsevier, 2003).
12. Lee, W. H., Lisanby, S. H., Laine, A. F. & Peterchev, A. V. Comparison of electric field
strength and spatial distribution of electroconvulsive therapy and magnetic seizure therapy
in a realistic human head model. Eur. Psychiatry 36, 55–64 (2016).
20
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
13. Deng, Z.-D., Lisanby, S. H. & Peterchev, A. V. Electric field strength and focality in
electroconvulsive therapy and magnetic seizure therapy: a finite element simulation study. J.
Neural Eng. 8, 016007 (2011).
14. Fink, M. Relation of Electroencephalographic Delta Activity to Behavioral Response in
Electroshock: Quantitative Serial Studies. AMA Arch. Neurol. Psychiatry 78, 516 (1957).
15. Hill, A. T. et al. Resting-state electroencephalographic functional network alterations in major
depressive disorder following magnetic seizure therapy. Prog. Neuropsychopharmacol. Biol.
Psychiatry 108, 110082 (2021).
16. Hill, A. T. et al. Modulation of functional network properties in major depressive disorder
following electroconvulsive therapy (ECT): a resting-state EEG analysis. Sci. Rep. 10,
17057 (2020).
17. Sackeim, H. A. et al. The Effects of Electroconvulsive Therapy on Quantitative
Electroencephalograms: Relationship to Clinical Outcome. Arch. Gen. Psychiatry 53, 11
(1996).
18. Levy, N. A., Serota, H. M. & Grinker, R. R. Disturbances in brain function following
convulsive shock therapy: Electroencephalographic and clinical studies. Arch. Neurol.
Psychiatry 47, 1009 (1942).
19. Krystal, A. D. & Weiner, R. D. EEG correlates of the response to ECT: A possible
antidepressant role of brain-derived neurotrophic factor. J. ECT 15, 27–38 (1999).
20. Donoghue, T. et al. Parameterizing neural power spectra into periodic and aperiodic
components. Nat. Neurosci. 23, 1655–1665 (2020).
21. Donoghue, T., Schaworonkow, N. & Voytek, B. Methodological considerations for studying
neural oscillations. Eur. J. Neurosci. 55, 3502–3527 (2022).
22. Cole, S. & Voytek, B. Cycle-by-cycle analysis of neural oscillations. J. Neurophysiol. 122,
21
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
849–861 (2019).
23. Lundqvist, M. et al. Gamma and Beta Bursts Underlie Working Memory. Neuron 90,
152–164 (2016).
24. Sherman, M. A. et al. Neural mechanisms of transient neocortical beta rhythms: Converging
evidence from humans, computational modeling, monkeys, and mice. Proc. Natl. Acad. Sci.
113, (2016).
25. Smith, S. E. et al. Clinical EEG slowing induced by electroconvulsive therapy is better
described by increased frontal aperiodic activity.
http://medrxiv.org/lookup/doi/10.1101/2022.04.15.22273811 (2022)
doi:10.1101/2022.04.15.22273811.
26. Buzsáki, G., Anastassiou, C. A. & Koch, C. The origin of extracellular fields and currents —
EEG, ECoG, LFP and spikes. Nat. Rev. Neurosci. 13, 407–420 (2012).
27. Lindén, H., Pettersen, K. H. & Einevoll, G. T. Intrinsic dendritic filtering gives low-pass power
spectra of local field potentials. J. Comput. Neurosci. 29, 423–444 (2010).
28. Gao, R., Peterson, E. J. & Voytek, B. Inferring synaptic excitation/inhibition balance from
field potentials. NeuroImage 158, 70–78 (2017).
29. Chini, M., Pfeffer, T. & Hanganu-Opatz, I. L. Developmental increase of inhibition drives
decorrelation of neural activity. http://biorxiv.org/lookup/doi/10.1101/2021.07.06.451299
(2021) doi:10.1101/2021.07.06.451299.
30. Voineskos, D. et al. The Relationship Between Cortical Inhibition and Electroconvulsive
Therapy in the Treatment of Major Depressive Disorder. Sci. Rep. 6, 37461 (2016).
31. Daskalakis, Z. J. Efficacy and Tolerability of Magnetic Seizure Therapy (MST) as an
Alternative to Electroconvulsive Therapy (ECT) for Treatment Resistant Depression,
Schizophrenia, and Obsessive Compulsive Disorder.
22
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
https://clinicaltrials.gov/ct2/show/NCT01596608 (2020).
32. Donoghue, T., Dominguez, J. & Voytek, B. Electrophysiological Frequency Band Ratio
Measures Conflate Periodic and Aperiodic Neural Activity. eneuro 7,
ENEURO.0192-20.2020 (2020).
33. Podvalny, E. et al. A unifying principle underlying the extracellular field potential spectral
responses in the human cortex. J. Neurophysiol. 114, 505–519 (2015).
34. Gyurkovics, M., Clements, G. M., Low, K. A., Fabiani, M. & Gratton, G. Stimulus-Induced
Changes in 1/ f-like Background Activity in EEG. J. Neurosci. 42, 7144–7151 (2022).
35. He, B. J., Zempel, J. M., Snyder, A. Z. & Raichle, M. E. The Temporal Structures and
Functional Significance of Scale-free Brain Activity. Neuron 66, 353–369 (2010).
36. Preston, M., Schaworonkow, N. & Voytek, B. Oscillations and aperiodic activity: Evidence for
dynamic changes in both during memory encoding.
http://biorxiv.org/lookup/doi/10.1101/2022.10.04.509632 (2022)
doi:10.1101/2022.10.04.509632.
37. He, W. et al. Co-Increasing Neuronal Noise and Beta Power in the Developing Brain.
http://biorxiv.org/lookup/doi/10.1101/839258 (2019) doi:10.1101/839258.
38. Voytek, B. et al. Age-Related Changes in 1/f Neural Electrophysiological Noise. J. Neurosci.
35, 13257–13265 (2015).
39. Colombo, M. A. et al. The spectral exponent of the resting EEG indexes the presence of
consciousness during unresponsiveness induced by propofol, xenon, and ketamine.
NeuroImage 189, 631–644 (2019).
40. Robertson, M. M. et al. EEG power spectral slope differs by ADHD status and stimulant
medication exposure in early childhood. J. Neurophysiol. 122, 2427–2437 (2019).
41. Molina, J. L. et al. Memantine effects on EEG measures of putative excitatory/inhibitory
23
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
balance in schizophrenia. 15 (2021).
42. Veerakumar, A. et al. Field potential 1/ factivity in the subcallosal cingulate region as a
candidate signal for monitoring deep brain stimulation for treatment-resistant depression. J.
Neurophysiol. 122, 1023–1035 (2019).
43. Brake, N. et al. Aperiodic EEG activity masks the dynamics of neural oscillations during loss
of consciousness from propofol. http://biorxiv.org/lookup/doi/10.1101/2021.10.12.464109
(2021) doi:10.1101/2021.10.12.464109.
44. Luscher, B., Shen, Q. & Sahir, N. The GABAergic deficit hypothesis of major depressive
disorder. Mol. Psychiatry 16, 383–406 (2011).
45. Rajkowska, G., O’Dwyer, G., Teleki, Z., Stockmeier, C. A. & Miguel-Hidalgo, J. J. GABAergic
Neurons Immunoreactive for Calcium Binding Proteins are Reduced in the Prefrontal Cortex
in Major Depression. Neuropsychopharmacology 32, 471–482 (2007).
46. Prévot, T. & Sibille, E. Altered GABA-mediated information processing and cognitive
dysfunctions in depression and other brain disorders. Mol. Psychiatry 26, 151–167 (2021).
47. Fuchs, T. et al. Disinhibition of somatostatin-positive GABAergic interneurons results in an
anxiolytic and antidepressant-like brain state. Mol. Psychiatry 22, 920–930 (2017).
48. Lin, L.-C. & Sibille, E. Reduced brain somatostatin in mood disorders: a common
pathophysiological substrate and drug target? Front. Pharmacol. 4, (2013).
49. Mazza, F., Valiante, T. A., Griffiths, J. D. & Hay, E. EEG biomarkers of reduced inhibition in
human cortical microcircuits in depression.
http://biorxiv.org/lookup/doi/10.1101/2021.07.18.452836 (2021)
doi:10.1101/2021.07.18.452836.
50. Knight, R. T., Richard Staines, W., Swick, D. & Chao, L. L. Prefrontal cortex regulates
inhibition and excitation in distributed neural networks. Acta Psychol. (Amst.) 101, 159–178
24
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
(1999).
51. Drevets, W. C., Price, J. L. & Furey, M. L. Brain structural and functional abnormalities in
mood disorders: implications for neurocircuitry models of depression. Brain Struct. Funct.
213, 93–118 (2008).
52. Hadas, I. et al. Association of Repetitive Transcranial Magnetic Stimulation Treatment With
Subgenual Cingulate Hyperactivity in Patients With Major Depressive Disorder: A
Secondary Analysis of a Randomized Clinical Trial. JAMA Netw. Open 2, e195578 (2019).
53. Sackeim, H. A. Convulsant and anticonvulsant properties of electroconvulsive therapy:
towards a focal form of brain stimulation. Clin. Neurosci. Res. 4, 39–57 (2004).
54. McElhiney, M. C. et al. Autobiographical Memory and Mood: Effects of Electroconvulsive
Therapy. Neuropsychology 9, 501–517 (1995).
55. Sackeim, H. A. et al. Electrophysiological Correlates of the Adverse Cognitive Effects of
Electroconvulsive Therapy: J. ECT 16, 110–120 (2000).
56. Fuentemilla, L., Barnes, G. R., Düzel, E. & Levine, B. Theta oscillations orchestrate medial
temporal lobe and neocortex in remembering autobiographical memories. NeuroImage 85,
730–737 (2014).
57. van der Vinne, N., Vollebregt, M. A., van Putten, M. J. A. M. & Arns, M. Frontal alpha
asymmetry as a diagnostic marker in depression: Fact or fiction? A meta-analysis.
NeuroImage Clin. 16, 79–87 (2017).
58. Kołodziej, A., Magnuski, M., Ruban, A. & Brzezicka, A. No relationship between frontal
alpha asymmetry and depressive disorders in a multiverse analysis of five studies. eLife 10,
e60595 (2021).
59. Cook, I. A., Hunter, A. M., Korb, A. S. & Leuchter, A. F. Do prefrontal midline electrodes
provide unique neurophysiologic information in Major Depressive Disorder? J. Psychiatr.
25
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
Res. 53, 69–75 (2014).
60. Watts, D. et al. Predicting treatment response using EEG in major depressive disorder: A
machine-learning meta-analysis. Transl. Psychiatry 12, 332 (2022).
61. Gramfort, A. et al. MEG and EEG data analysis with MNE-Python. Front. Neurosci. 7,
(2013).
62. The pandas development team. pandas-dev/pandas: Pandas. (2020).
63. Vallat, R. Pingouin: statistics in Python. J. Open Source Softw. 3, 1026 (2018).
64. Seabold, S. & Perktold, J. Statsmodels: Econometric and Statistical Modeling with Python.
in 92–96 (2010). doi:10.25080/Majora-92bf1922-011.
65. Hunter, J. D. Matplotlib: A 2D Graphics Environment. Comput. Sci. Eng. 9, 90–95 (2007).
66. Waskom, M. seaborn: statistical data visualization. J. Open Source Softw. 6, 3021 (2021).
Supplementary material
ECT: The regression to predict theta band power was overall significant (R2= 0.8, F(2, 19) = 5.91, p =
0.010). An increase in aperiodic exponent significantly predicts theta band power measures (β= 0.66, p =
0.022), but not by theta abundance (β= 0.40, p = 0.081). For alpha band power, we were able to include
the aperiodic adjusted alpha power, due to alpha's high abundance, which makes it a more reliable
measurement. The overall regression was significant (R2= 0.50, F(3, 18) = 6.00, p = 5.11 x 10-3). Alpha
band power can be significantly predicted by both alpha aperiodic adjusted power (β= 0.52, p = 0.005),
and by alpha abundance (β= -1.16, p = 0.043), but not by a change in exponent (β= 0.17, p = 0.42).
MST: The regression to predict theta band power was overall significant (R2= 0.70, F(2, 20) = 23.78, p =
5.16 x 1010-56). A change in exponent significantly predicts theta band power (β= 1.39, p = 7.0 x 10-6),
but theta abundance did not (β= 0.13, p = 0.32). The regression for alpha band power was overall
significant (R2= 0.46, F(3, 19) = 5.49, p = 6.88 x 10-3). Both aperiodic exponent (β= 0.77, p = 0.003) and
aperiodic adjusted alpha power (β= 0.59, p = 0.007) significantly predict alpha band power, but not
alpha abundance (β= 0.22, p = 0.37).
26
. CC-BY-NC-ND 4.0 International licenseIt is made available under a
is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted January 12, 2023. ; https://doi.org/10.1101/2023.01.11.23284450doi: medRxiv preprint
