Characterizing PTSD Using Electrophysiology: Towards A Precision Medicine Approach

Abstract

Objective. Resting-state EEG measures have shown potential in distinguishing individuals with PTSD from healthy controls. ERP components such as N2, P3, and late positive potential have been consistently linked to cognitive abnormalities in PTSD, especially in tasks involving emotional or trauma-related stimuli. However, meta-analyses have reported inconsistent findings. The understanding of biomarkers that can classify the varied symptoms of PTSD remains limited. This study aimed to develop a concise set of electrophysiological biomarkers, using neutral cognitive tasks, that could be applied across psychiatric conditions, and to identify biomarkers associated with the anxiety and depression dimensions of PTSD. Approach. Continuous simultaneous recordings of EEG and electrocardiogram (ECG) were obtained in veterans with PTSD (n = 29) and healthy controls (n = 62) during computerized tasks. EEG, ERP, and heart rate measures were evaluated in terms of their ability to discriminate between the groups or correlate with psychological measures. Results. The PTSD cohort exhibited faster alpha oscillations, reduced alpha power, and a flatter power spectrum. Furthermore, stronger reduction in alpha power was associated with higher trait anxiety, while a flatter slope was related to more severe depression symptoms in individuals with PTSD. In ERP tasks of visual memory and sustained attention, the PTSD cohort demonstrated delayed and exaggerated early components, along with attenuated LPP amplitudes. The three tasks revealed distinct and complementary EEG signatures PTSD. Significance. Multimodal individualized biomarkers based on EEG, cognitive ERPs, and ECG show promise as objective tools for assessing mood and anxiety disturbances within the PTSD spectrum.

Introduction

Post-traumatic stress disorder (PTSD) causes marked debilitation in affected individuals and is considerably widespread, with some sources estimating a lifetime prevalence between 6.1% and 8.7% in the U.S. population^1–3 and a prevalence as high as 15% in the general population following the COVID-19 pandemic. ⁴ Though frequently linked to combat veterans,^5–7 PTSD is also now linked to survivors of natural disaster and victims of childhood abuse and sexual assault.^8,9 The impact of PTSD is often leading to depression, severe anxiety, and suicidal thoughts^10,11 and presenting a high risk of chronicity and significant somatic complications. ¹² Consequently, early detection and improved symptom assessment are of utmost importance.

For many years, researchers have used electroencephalography (EEG) as a cost-effective, non-invasive, and efficient technique to derive biomarkers for PTSD. However, the heterogeneous etiology and symptomology of PTSD ¹³ as well as the diverse methodology applied by researchers has frustrated efforts to identify a consistent neurophysiological profile^14–18.

Analyses of spectral power during resting state have produced mixed results. ¹⁹ In eyes closed condition, studies have found either no significant differences in EEG power between PTSD subjects and controls,^14,20–22 or both increases (alpha and theta; ²³ beta and theta ²⁴ ) and decreases (theta ²⁵ ). In eyes open condition, a decrease in alpha and theta power in PTSD subjects was observed in some studies (alpha; ²⁶ alpha and theta ²⁷ ).

Event related potential (ERP) components, such as N2, P3 and late positive potential (LPP), have been identified as promising biomarkers of PTSD. The N2 component has been associated with stimulus novelty, cognitive control, threat detection and control over threat response, while the P3 component has been associated with processes relating to attention, working memory and evaluative and task-relevant cognitive processing.^17,28 PTSD has been associated with delayed N2 latency and reduced P3/P300 amplitudes in various paradigms^{17,18,27,29–32} and even for distinguishing PTSD from major depressive disorder (MDD). ³³ However, this decrease in P3 is not universal as there is ample evidence that PTSD patients show increased amplitudes in response to trauma-related stimuli. ¹⁵

Event-related spectral perturbations (ERSP) quantify stimulus-induced alterations in ongoing EEG spectral power. It is computed in time windows following stimulus onset as a relative change from the pre-stimulus baseline, across the broad EEG frequency spectrum. ³⁴ Although ERSP provides complementary information to ERP, ³⁵ particularly in relation to anxiety,^36,37 it is rarely reported in the context of PTSD.

Heart rate (HR) and heart rate variability (HRV) are simple, easily obtainable metrics that reflect the balance between the sympathetic (fight-or-flight) and parasympathetic (rest-and-digest) branches of the autonomic nervous system. ³⁸ Several studies have found that resting HR is increased^39,40 and HRV is decreased in individuals with PTSD compared to healthy controls,^41–44 a finding that has been correlated with higher mortality rates, in large part due to an increased chance of heart disease. ⁴⁵

The current study aimed to investigate the relationship between PTSD and neuropsychological and physiological measures that are easily obtainable in clinical settings. These measures included EEG and HR/HRV resting state measures, and the N2 and P3/LPP components, ERSP and behavioral measures from ERP tasks involving memory and attention. The tasks were chosen as part of a broader objective to establish a precision psychiatry approach to differentiate between various diagnoses and/or symptoms, including depression, anxiety, and trauma, using the same testbed.

Methods

Participants

32 combat veterans clinically diagnosed with PTSD were recruited for the study in Spokane, WA. A diagnosis of PTSD was validated through standard clinical evaluation as well as through the Clinician-Administered PTSD Scale for DSM-5 (CAPS). All veterans were fully employed. Three participants were excluded due to protocol noncompliance, leaving 29 participants for analysis in the PTSD group. 62 healthy participants were recruited in ***. Of those, 34 individuals scoring below 200 on the Holmes-Rahe Stress Inventory ⁴⁶ were designated as the healthy control group (HC). The remaining 34 healthy participants, scoring 200 or higher on the inventory, comprised the Major Life Stress group (MLS). Participants in both the MLS and HC groups reported no major health problems, including but not limited to diagnosis of neurological, psychiatric, sleep, behavioral, or cardiovascular disorders, or diabetes.

All participants completed several neuropsychological self-report questionnaires at the start of their visit, including the Beck Depression Inventory (BDI) and the State-Trait Anxiety Inventory (STAI). One of the healthy subjects exhibited excessive anxiety scores (54 on STAI-State and 59 on STAI-Trait) and was removed from the subsequent analyses. The demographic and clinical status statistics for the three groups are presented in Table 1.

Table 1.

Demographic Characteristics of HC, MLS and PTSD Groups. The Statistical Significance of Differences in age and Clinical Scores Between HC and PTSD Groups is Indicated in the Last Column.

	HC	MLS	PTSD	p-value
Sample size (Sex)	27 (F = 20, M = 7)	34 (F = 17, M = 17)	29 (F = 2, M = 27)
Age range (min-max)	23–65	19–70	26–76
Mean age (SD)	38.36 (14.13)	35.68 (13.84)	41.23 (13.79)	0.561
Mean STAI Trait (SD)	25.69 (5.77)	36.74 (11.79)	49.54 (9.10)	<0.001
Mean STAI State (SD)	25.34 (4.66)	32.94 (10.25)	44.43 (8.88)	<0.001
Mean BDI (SD)	1.53 (2.03)	6.03 (5.91)	21.67 (9.33)	<0.001

Data Recording

A STAT™ X24 system (Advanced Brain Monitoring, USA), an FDA-cleared wireless EEG system with passive Ag/AgCl electrodes and flexible circuit cables, was used for stimulus presentation and data collection. EEG data were recorded using twenty channels configured in the standard 10–20 montage and referenced to linked mastoids. The data were sampled at a 256 Hz rate using an amplifier with cut-off frequencies of 0.1 and 100 Hz. Electrode impedances were maintained below 40 kOhm. ⁴⁷

An ECG electrode was connected to the STAT™ X24 system and placed on the right clavicle and lower left rib. ECG data were recorded at the same 256 Hz sampling rate and saved within the EEG file as an additional 21^st channel.

Tasks

The test battery began with 5 min of eyes open (EO) and 5 min of eyes closed (EC) resting state, followed by two ERP tasks: the image recognition task (IR) and the three-choice vigilance task (3CVT).

The IR task begins with an encoding phase, where 20 images are presented twice in a pseudo-random order. Participants are instructed to memorize as many images as possible. In the subsequent test phase, the images from the encoding phase (Targets) are mixed with 80 novel images (NonTargets). Participants press the left arrow key for recognized images and the right arrow key for unrecognized images (Supplementary Figure S1).

The 3CVT is a continuous performance task that probes sustained and selective attention. ⁴⁸ Participants are instructed to discriminate frequent stimuli (Targets, 70% of trials) from infrequent stimuli (Supplementary Figure S2).

EEG Preprocessing and Analysis

EEG data were imported into MATLAB (version R2018). Data were bandpass filtered (0.1-50 Hz) using a zero-phase Hamming windowed sinc FIR filter implemented in EEGLAB software v2023.0. ⁴⁹ Independent component analysis (ICA) was performed using EEGLAB and ICLabel plugin v1.3 ⁵⁰ was used to automatically identify the source of independent components. ICLabel employs a classifier that is pretrained by thousands of labeled components obtained through crowdsourcing. Components classified as having sources other than brain (eg, eye blinks, EMG, etc) were automatically removed.

The continuous recording was divided into 1-s segments, and the absolute power spectral density was computed for each. The results were then averaged across segments for each of the two resting state conditions. Relative power was calculated by dividing the absolute power at each frequency by the total power over the 1–40 Hz range. For each participant we calculated individual alpha frequency (IAF) as the dominant frequency in the eyes closed condition. IAF was determined using MATLAB's findpeaks function applied to the absolute power in the 7–14 Hz frequency range.

To calculate spectral power within standard bands, we employed an individualized approach for alpha and theta bands, utilizing central alpha frequency as the anchor. Due to observed group differences in IAF, as shown in the Results, employing standard theta and alpha bands (4-7 Hz and 8-12 Hz) was deemed unsuitable because any disparities in power would be confounded by variations in peak alpha frequency. Individualized alpha band was defined as a 5 Hz range centered at IAF, [IAF-2 Hz, IAF+2 Hz]. Individualized theta band was defined as a 4 Hz range below the alpha range, [IAF-6 Hz, IAF-3 Hz]. Beta and gamma bands were defined using the standard ranges 13–28 Hz, and 28–40 Hz, respectively. Preliminary analyses revealed that patients exhibited a flatter spectrum, characterized by increased power in high frequencies and decreased power in low frequencies. Accordingly, power spectra were split into aperiodic and periodic components. We computed the aperiodic component of the absolute power spectrum using the FOOOF toolbox, ⁵¹ version 1.0.0. The algorithm was applied with recommended default settings, using a fitting range of 1–40 Hz and a ‘fixed’ aperiodic mode. FOOOF algorithm approximates the aperiodic component of the log power spectrum with an exponential function parametrized with offset b and exponent s:

$$AP(f)=b-{\log }_{10}{f}^s$$

where f ranges from 1 to 40 Hz. Offset is related to the total power, while exponent captures the 1/f slope of the EEG spectrum. In the reminder of the manuscript, we denote these two parameters as AP Offset and AP Slope.

Periodic components were calculated by first subtracting the aperiodic component from the log power spectrum, then summing up the residuals across the frequencies within each frequency band. The resultant periodic components were designated as PIAlpha, corresponding to the individualized alpha frequency band, and as PBeta and PGamma, corresponding to the standard beta and gamma bands.

ERP Preprocessing and Analysis

To compute ERP wave forms from the two tasks, the EEG data were segmented into 1.5-s epochs, with a 0.2-s pre-stimulus baseline. Similar to the approach used for the resting state data, the ICA method was applied to remove artifacts from the epoched data. For each participant, artifact-free single trial, correct response data were grouped based on the time-locking stimulus type and averaged across trials. The mean number of trials before and after preprocessing are provided in Supplementary Table 1.

Averaged ERPs with maximum absolute amplitudes greater than 50 µV were excluded from the statistical analyses (<1%). The ERP wave forms were lowpass filtered using EEGLAB's eegfilt function with a 25 Hz cutoff. ERP peaks were measured from the filtered wave forms using an in-house automated algorithm. The algorithm uses MATLAB's findpeaks function to identify the peak with the highest amplitude in a predefined direction (positive or negative) within a predefined time window. ERP peak amplitudes vary across the scalp and can be indistinguishable in certain electrode sites. In such instances, the automated algorithm might not detect a peak. Reliable channels were previously determined using the entire healthy control database of ABM (N > 200). A channel was considered reliable for a given ERP peak if it was successfully detected in at least 90% of healthy subjects. In the present study, the statistical analyses of ERP peaks were limited to thus determined reliable channels. LPP components were calculated by averaging ERP amplitudes within a specified time window.

In the IR task, ERP waveforms were generated for Target and NonTarget conditions. The ERP morphology was assessed by measuring the P2 peak within the 120–220 ms time window and the LPP within the 350–650 ms time window. The linked mastoid referencing produces a positive P2 polarity corresponding to the traditional N2 peak (Supplementary Figure S4). The memory-related increase in late positive amplitudes was measured as the within-subject difference in LPP between the Target and NonTarget conditions.

In the 3CVT task, ERP waveforms were generated for Targets, NonTargets, and Distractors. The morphology of each waveform was evaluated by measuring the N1 peak within the 130–240 ms time window and the LPP within the 350–550 ms time window. For subsequent statistical analysis, we primarily focused on the Target condition, as it produced the most robust ERPs, with an average of over 100 trials.

ERSP Analysis

To evaluate event-related changes in spectral power, we applied EEGLAB's newtimef function ³⁴ to the same artifact-corrected trials grouped by stimulus type, as used for calculating ERPs. For each electrode, ERSP was calculated across the frequency-time domain, covering frequencies from 1 to 40 Hz in 1 Hz increments and times from 0 to 871 ms in 7.8 ms increments. ERSP was computed as the deviation from the pre-stimulus baseline period of [-500 0] ms. The newtimef function generated normalized values in decibels (dB).

Behavioral Measures

In the image recognition task, mean response time and accuracy were computed from trials with correct responses occurring between 300 and 2000 ms after stimulus onset. This range was selected to exclude the influence of accidental early responses or significantly delayed responses.

In the sustained attention task, the mean response time and accuracy were determined using the same methodology. The only difference was in defining the admissible range to be 300 to 1500 ms, accounting for the faster response times in the 3CVT task.

Heart Rate Variability Measures

Heart rate data collected during the resting state session were divided into eyes open and eyes closed conditions. Preprocessing of the ECG data was carried out in MATLAB using the HRVTool toolbox. ⁵² The data were filtered to remove artifacts using the default thresholds provided by the toolbox. The automated artifact rejection and R-peak detection were visually verified on a random sample of 10 recordings, confirming their accuracy.

Five HRV measures were computed: root mean square of successive differences (RMSSD), standard deviation normal to normal inter-beat intervals (SDNN), percentage of successive normal intervals of more than 50 ms (pNN50), low to high-frequency ratio (LF/HF), and the median distance to the center of the RR interval return map (rrHRV), a measure shown to be robust against signal artifacts, making it well-suited for unsupervised automatic HRV estimation. ⁴⁸

Statistical Analysis

We compared EEG and ECG based measures between the PTSD and HC groups using the Student's t-test. Independent two-sample t-tests were used at each EEG channel to examine between-group differences while controlling for unequal variances. To adjust for multiple comparisons, we set the false detection rate (FDR) threshold, ⁵³ computed using EEGLAB's fdr function, at 0.05. We report significance for channel-wise p-values before and after FDR correction.

We regarded a multi-channel EEG measure as a biomarker if at least three spatially adjacent electrodes exhibited significance at p < 0.05. For resting state measures we added an additional requirement of having a consistent spatial pattern across EO and EC conditions, ie the same direction of the effects (PTSD < HC or opposite), comparable effect size and significance across scalp electrodes. This requirement was verified through visual inspection. For ERP measures, the t-test was carried out for reliable channels only. A consistent spatial pattern was required for at least two different stimulus types within the same task. Each channel was also tested for significance after FDR correction. However, the resulting significance levels did not serve as a requirement for selecting biomarkers. Normalized effect sizes (ES) were calculated using Hedge's G. ⁵⁴

We examined the correlations within the patient group between EEG/ECG based measures and scores from CAPS Frequency, CAPS Severity, STAI, and BDI. To eliminate the influence of demographic factors, we regressed age and sex from the EEG/ECG measures prior to calculating the correlations. Pearson correlation coefficients (r-values) and corresponding p-values were computed using MATLAB's corrcoef function. Univariate correlations that were significant at a level of p < 0.05 were considered as severity biomarkers. For multivariate endpoints, the same criteria of significance as stated above were applied to determine their eligibility as multivariate severity biomarkers.

Statistical analysis of ERSP measures was performed using cluster-based permutations ⁵⁵ as implemented in FieldTrip ⁵⁶ on decibel-normalized power values across the electrode × time × frequency domain for each ERP task and stimulus type, comparing HC and PTSD groups.

While the primary aim of the analyses was to contrast PTSD patients with healthy controls who have not experienced exceptionally stressful events, the MLS group offered a chance to investigate the entire spectrum from health to illness. Consequently, the MLS group was assessed against both HC and PTSD groups using identical methods as described above.

Results

Resting State Biomarkers

Figure 1 displays the group means of the global power spectra and corresponding aperiodic components. Mean IAF was 9.24 (SD = 0.89) for the HC group and 10.07 (SD = 0.7) for the PTSD group. In comparison to the HC group, the PTSD group showed a significant increase in IAF, with a mean increase of 0.83 Hz (p < 0.001, ES = 1.02). IAF did not exhibit significant correlations with clinical measures among the patients. In the EO condition, the PTSD group exhibited reduced AP slope, PIAlpha and PBeta, and increased PGamma (Figure 1C). These effects were significant after FDR correction, with strong effect sizes across the scalp (max ES > 0.8), except for PBeta decrease, which was significant across the posterior and occipital regions only, with moderate effect sizes (0.3 < ES < 0.8). Results concerning the MLS group are provided in the Supplementary Results.

Figure 1.

Resting state biomarkers of PTSD. Panels A and B show group grand averages of the global power spectrum with corresponding fitted aperiodic components in eyes closed and eyes open conditions, respectively. Panel C. Horizontal rows present scalp maps of aperiodic slope and periodic power measures, PIAlpha, PBeta and PGamma for the HC and PTSD groups during eyes open condition. Each row includes a third scalp map representing the normalized effects sizes of the difference between group means, overlaid with markers of statistical significance. Panel D. Scalp maps of significant correlations between measures and symptom severity scores. Beneath each scalp map, scatterplots show associations at a site with the maximum significant absolute correlation.

Across the PTSD group, aperiodic slope exhibited an inverse association with depression severity (maximum r = −0.47, p = 0.01 at C3), while both PIAlpha and PBeta displayed negative relationships with anxiety severity was found for PIAlpha (maximum r = -0.46, p = 0.01at T6) and PBeta (maximum r = –0.55, p = 0.002 at O2) (Figure 1D). PGamma did not exhibit any significant correlations with measures of symptom severity.

Behavioral Results

T-test results did not reveal any statistically significant differences between the PTSD and HC groups. Group means and standard deviations of performance measures are presented in the Supplementary Table 2. One participant scored below 50% in accuracy on the IR task, possibly due to misunderstood or forgotten task instructions, and was excluded from the analysis of IR performance.

ERP Biomarkers

The grand average ERP waveforms from the IR and 3CVTs tasks can be found in Supplementary Figures S4 and S5.

In the IR task, the P2 peak was delayed in the PTSD group for both Target and NonTarget conditions (Figure 2). The maximum delay (16 ms) and statistical significance were observed in the NonTarget condition (p = 0.014, ES = 1.75 at electrode Fp1). There was a weak trend indicating higher P2 amplitudes in PTSD, but it was not statistically significant. The second biomarker from the IR task was the Target-NonTarget difference in LPP (Figure 2B). The maximum difference of 3.2 µV was observed at Cz (p = 0.02, ES = 0.76).

Figure 2.

ERP biomarkers from the image recognition task. A. ERP wave forms for the healthy controls (HC) and PTSD groups in the NonTarget condition. In the PTSD group, the P2 peak is delayed, with the scalp map showing the distribution of these delays represented as the difference in P2 latencies between PTSD and HC. Statistical significance is indicated by overlayed markers at each electrode site. B. Memory-related late positive potential amplitudes (350-650 ms) measured by the within-subject subtraction condition (Target-NonTarget) were higher in HC compared to PTSD. The scalp map shows the spatial distribution of this group difference. Statistical significance is indicated by overlayed markers at each electrode site.

In the 3CVT task, the PTSD group showed a delayed and enlarged N1 peak, and decreased LPP, compared to the HC group (Figure 3). The maximum N1 latency delay of 15 ms was observed at the C4 site in the Target condition (p = 0.001, ES = 1.05). The maximum N1 amplitude increase of 1.6 µV in the negative direction was found at the P3 site in the Target condition (p = 0.04, ES = 0.82). The third biomarker was the LPP amplitude, with the maximum PTSD-related reduction of 2.7 µV observed at P4 (p = 0.001, ES = 0.99).

Figure 3.

ERP-based biomarkers from the three-choice vigilance task. Line plot shows grand average ERP wave forms for the healthy controls (HC) and PTSD groups in the Target condition. PTSD patients exhibited delayed and enlarged N1 peak and reduced late positive potential (LPP). The spatial distribution of the group differences between PTSD and HC is depicted with scalp maps overlaid with markers of statistical significance at each electrode site.

ERP measures did not reveal any biomarkers of symptom severity.

ERSP Biomarkers

We identified one significant cluster for the 3CVT Target condition (p = 0.005 with cluster-based correction), distinguishing the HC and PTSD groups (Figure 4, Supplementary Figure S6). The cluster captures heightened event-related alpha synchronization (ERS) approximately 200 ms post-stimulus and diminished alpha and beta event-related desynchronization (ERD) between 300 and 500 ms post-stimulus in the PTSD group compared to healthy controls.

Figure 4.

Grand average time-frequency plots representing event-related spectral perturbations (ERSP) across participant groups (HC, MLS, and PTSD) recorded at the Pz electrode during the Target condition of the three-choice vigilance task. The boundary of the significant cluster, that differentiates the PTSD group from the HC group is overlaid.

Heart Rate Biomarkers

Table 2 displays the statistical results for heart rate and heart rate variability during eyes open condition. There was no significant difference in heart rate between the PTSD and HC groups. However, heart rate variability was decreased in the PTSD group across all time-domain HRV measures, with only pNN50 and rrHRV showing significant differences between the two groups. In the eyes closed condition, similar trends were observed, but none achieved significance. HR and HRV measures did not correlate with symptom severity.

Table 2.

Group Comparison of Heart Rate and Hearty Rate Variability Measures, including p-values and Normalized Effect Sizes of the HC-PTSD Difference in the Last Two Columns. Asterisks (*) Indicate Significant Differences (p< = 0.05) between the PTSD and HC Groups using t-test. Overall, PTSD Patients Displayed Decreased Time-Domain Heart rate Variability Measures, with Two Measures, RMSSD and pNN50, Achieving Significance. Healthy Participants from the Major Life Stress Group (MLS) showed no Statistically Significant Differences Compared to the Healthy Controls (HC).

	HC	MLS	PTSD	p-value	ES
Heart rate (HR)	69.9(12.7)	70.6 (12.4)	70.2(11.6)	0.933	−0.02
Root mean square of successive differences (RMSSD)	39.7(24.1)	35.8 (20.9)	25.6(10.0)	0.01*	0.75
Standard deviation normal to normal inter-beat intervals (SDNN)	45.9(16.2)	48.2 (24.1)	45.9(21.1)	0.996	0
Percentage of successive normal intervals of more than 50 ms (pNN50)	19.6(19.8)	17.1 (17.7)	7.3(7.1)	0.005*	0.81
Median distance to the center of the RR interval return map (rrHRV)	4.2(2.0)	4.2 (2.6)	3.4(1.3)	0.095	0.47
Low to high-frequency ratio (LF/HF)	2.1 (1.4)	1.9 (1.3)	2.2 (1.3)	0.762	−0.08

Conclusions

In this study, putative biomarkers of PTSD were identified using EEG and ECG data acquired during resting state, and two cognitive tasks.

Our findings reveal a trauma-related overall reduction in individualized relative alpha power, expressed as HC > MLS > PTSD. Additionally, we identified negative correlations between alpha power and anxiety symptoms within the PTSD group, and across all healthy subjects comprising HC and MLS groups. A significant reduction of alpha in eyes open condition is consistent with heightened processing of sensory input and hypervigilance. ⁵⁷ Negative correlations between alpha power and depression severity have been documented in patients with MDD^58–60. Taken together, these results suggest transdiagnostic links between alpha power and anxiety.

The aperiodic slope is considered an indicator of the overall balance between excitation and inhibition (EI balance). ⁶¹ Our findings demonstrate that individuals with PTSD exhibit flatter slopes compared to controls, consistent with previous reports.^62,63 Moreover, flatter aperiodic slope was associated with more severe depression symptoms among PTSD patients. Recent research has found evidence linking depression to a reduction in cortical inhibitory processes.^64,65 One study has found that increased excitability in the medial prefrontal cortex correlates with the suppression of natural reward-motivated behaviors and can predict the severity of anhedonia. ⁶⁶ Another investigation has associated an increase in the aperiodic slope with positive therapeutic effects in depression through deep brain stimulation. ⁶⁷ Our results suggest disrupted EI balance in PTSD that is likely due to a reduction in inhibitory processes.^29,68 We also document a reduction in heart rate variability associated with PTSD, consistent with previous reports. ⁶²

Similar to our findings regarding alpha power and aperiodic slope, a study of medication-free combat veterans found decreased Alpha1 power (8-10 Hz) and increased Beta power in veterans with PTSD compared to those without PTSD. ⁶⁹ Likewise, authors in ²⁷ found a decrease in Alpha1, decrease in Theta, and increase in Beta power. Both studies used fixed band frequency ranges and found an effect only in the lower alpha range. Increased IAF in PTSD was previously reported, ⁶² however no study to our knowledge incorporated this difference in defining alpha range, which could explain the lack of effect in the higher alpha range.

Overall, our resting state results replicate and extend earlier findings, however, we introduce a succinct set of personalized resting state biomarkers of PTSD, clearly distinguishing between global power spectrum alterations and oscillatory frequency-specific changes. Our results underscore the significance of aperiodic slope, and periodic alpha and beta components as robust biomarkers of PTSD. Moreover, we report novel correlations between these biomarkers and depression and anxiety severity, suggesting potential broader utility in various diagnostic contexts.

ERP alterations in the PTSD group compared to the HC group were found in both cognitive tasks. In the memory task, P2 latency delay of 16 ms was most prominently expressed in the frontal region. In the attention task, N1 peak exhibited a similar delay and a more negative N1 amplitude within the left posterior region. In both tasks, the late positive component in response to Target stimuli was reduced by ∼3 µV. These results are in line with previous reports of delayed and exaggerated early components in visual tasks of attention and working memory.^17,29 Such alterations in ERPs have been linked to deficits in performance in PTSD, primarily in relation to emotion or trauma-related stimuli. Our analysis did not indicate any correlations between ERP measures and symptom severity.

The analysis of event-related synchronization and desynchronization identified a cluster that differentiated patients from healthy controls during the attention task. PTSD patients showed increased posterior alpha synchronization around the N1 peak and reduced alpha/beta desynchronization during response selection. While alpha and beta desynchronization is typically linked to motor preparation, ⁷⁰ it is also sensitive to cognitive disruptions, including neurodegeneration and psychiatric conditions. ⁷¹ To our knowledge, this is the first report of PTSD-related disruptions in a sustained attention task.

Behavioral measures did not reveal any performance deficits. This finding is in line with an earlier study reporting trauma-related performance deficits in executive function and speed of information processing, yet no deficits in attention and working memory. ⁷² However, the absence of performance deficits in the sustained attention task contradicts previous findings.^27,73 This discrepancy may be due to an insufficient attentional challenge caused by the shortened task duration. The absence of behavioral deficits in the memory task is perplexing, given the general consensus on the negative impact of stress and PTSD. One possible explanation is that non-emotional, trauma-neutral stimuli are less impacted by stress-related mechanisms. For example, Swick and colleagues used neutrals stimuli in a working memory task with interference and have found only partial deficits. ⁷⁴ Additionally, the lack of distractors and delays between the encoding and testing phases in our paradigm may reduce its sensitivity to behavioral deficits specific to PTSD. For instance, Veltmeyer and colleagues found that PTSD patients showed improved performance in immediate verbal memory but exhibited impaired performance in verbal memory retention/access and vigilance. ²⁷

EEG screening for psychiatric conditions is emerging, and this work supports a multi-modal EEG testbed as a step forward. One of the limitations of the study is the lack of information regarding patients’ medication status. Future research should validate biomarkers and develop classifiers using larger samples, including individuals with anxiety, depression, and comorbidities.