Cerebrovascular Responses to O₂-CO₂ Exchange Ratio under Brief Breath-Hold Challenge in Patients with Chronic Mild Traumatic Brain Injury

Abstract

Breath-by-breath oxygen-carbon dioxide (O₂-CO₂₎ exchange ratio (bER) is a respiratory gas exchange (RGE) metric, which is the ratio of the changes in the partial pressure of O₂ (ΔPO₂) to CO₂ (ΔPCO₂) between end-inspiration and end-expiration, has been demonstrated to characterize the cerebrovascular responses to breath-hold challenge in healthy individuals. We aimed to explore whether bER could characterize cerebrovascular responses in patients with chronic mild traumatic brain injury (mTBI) under breath-hold challenge. We also investigated the correlation between bER and the severity of post-concussion symptoms. Blood-oxygenation-level-dependent (BOLD) images were acquired using functional magnetic resonance imaging (fMRI) on 10 patients with chronic mTBI and 10 controls without brain injury history when performing a breath-hold task. Time series of RGE metrics of ΔPO₂, ΔPCO₂, and bER were computed, and their cross-correlation with regional change in BOLD (ΔBOLD) was calculated. Symptom burden was assessed using the Rivermead Post Concussion Questionnaire (RPQ), and its correlation with RGE changes was also measured. Compared with controls, a diffuse decrease in the correlation between regional ΔBOLD and bER was found in the brain of patients with mTBI (p_fdr < 0.05). No significant difference was found between patients and controls for the correlation of regional ΔBOLD with ΔPO₂ and ΔPCO₂. Symptom severity indicated by RPQ scores increased with a decrease in the averaged changes of bER (ρ = 0.79, p = 0.01) and ΔPO₂ (ρ = 0.70, p = 0.03) in breath-hold epochs. Our imaging and symptom severity findings suggest that bER can be used to characterize cerebrovascular responses to breath hold in patients with mTBI. The RGE may contribute to the post-concussive symptom severity.

Introduction

Traumatic brain injury (TBI) is a global public health problem.¹ A spectrum of post-concussion symptoms varying from headaches, fatigue, sleep, and cognitive disturbances may persist for months or years after TBI.² Perturbations in microvascular flow are believed to precede neuronal dysfunction after TBI.^3–5

Exogenous carbon dioxide (CO₂) challenge in conjunction with functional magnetic resonance imaging (fMRI) measurement of blood-oxygenation-level-dependent (BOLD) signal in the brain has been used to show reduced/delayed cerebrovascular reactivity (CVR) in patients with chronic TBI.^6–8 The CO₂ administration, however, involves a complex gas delivery setup. An alternative method for CVR assessment is a breath-hold challenge.^9–14 Increase in cerebral blood flow (CBF) is well established during breath holding. While authors in the recent neuroimaging literature believe that CBF increase during breath holding is because of hypercapnia only,^15–17 our team attributed CBF increase in breath holding to the synergistic effect of both mild hypoxemia and mild hypercapnia because of the continual oxygen (O₂) uptake and CO₂ release in the closed breathing circuit during breath holding.¹⁸

An increase in O₂ uptake had been demonstrated repeatedly to be larger than an increase in CO₂ release during breath holding since 1946, resulting in a reduction in respiratory exchange ratio (RER), a ratio of CO₂ to O₂ at a steady state.^19–24 The RER results in these studies are consistent with the findings in our recent breath-hold study, where the dynamic breath-by-breath O₂-CO₂ exchange ratio (bER), a ratio of breath-by-breath O₂ uptake to CO₂ release, was increased during breath holding.¹⁸

The bER is defined as the ratio of the changes in the partial pressure of oxygen (ΔPO₂) to the change in the partial pressure of carbon dioxide (ΔPCO₂) between end-inspiration and end-expiration.¹⁸ The time-averaged value of bER is related to the reciprocal of RER. The bER was found to demonstrate a strong positive correlation with dynamic global and regional cerebral hemodynamic responses under breath-hold challenge¹⁸; an increase in bER corresponds to an increase in CBF. The bER changes were also superior to ΔPCO₂ changes in coupling with global and regional cerebrovascular responses to breath holding.¹⁸

The regional CVR in response to external CO₂ challenge resembled that in response to bER, more than that in response to ΔPCO₂ under breath-hold challenge.¹⁸ Notably, ΔPCO₂ is interchangeable with the end-tidal partial pressure of CO₂ (P_ETCO₂) given the low ambient CO₂ level. All these findings lend motivation to investigate whether bER can characterize the cerebrovascular responses to breath-hold challenge in patients better than ΔPCO₂ (P_ETCO₂).

In the current study, we aimed to characterize the cerebrovascular responses to RGE metrics of ΔPO₂, ΔPCO_2, and bER in patients with mild traumatic brain injury (mTBI) under breath-hold challenge. We hypothesized that the cerebrovascular responses to breath-hold challenge measured by BOLD signal changes (ΔBOLD) would have a stronger correlation with bER than with ΔPCO₂ in patients with mTBI. The ΔBOLD would also be less correlated with bER in compromised brain regions of patients than in normal brain regions of age-matched controls. Temporal and frequency features of RGE metrics in their coherence with ΔBOLD were also examined. The potential clinical impact was evaluated by correlating RGE metrics with symptom severity indicated by the Rivermead Post Concussion Questionnaire (RPQ) scores.

Methods

Study design

We performed a prospective breath-hold fMRI study on a cohort of patients with chronic mTBI and age-matched controls without a history of brain injuries. All the experimental procedures were explained to the subjects. Signed informed consent was obtained before participating in the study. All procedures were performed in compliance with the Declaration of Helsinki and approved by the Massachusetts General Hospital (MGH) Human Research Committee.

Participants

Patients with a diagnosis of mTBI for at least three months since the head impact were recruited at the MGH Neurology Clinic. The patients did not have a loss of consciousness or had a loss of consciousness of <30 min duration, and the Glasgow Coma Scale (GCS) score was ≥13.^25,26 No abnormality was reported in the structural brain images. Control subjects without a history of brain injuries were also recruited.

MRI acquisition

Three-dimensional T1-MEMPRAGE (TR/TE1/TE2/TE3/TE4 = 2530/1.64/3.5/5.36/7.22 msec, matrix = 256 × 256 × 176, resolution = 1 × 1 × 1mm³) and T2-SPACE FLAIR (TR/TE/TI = 5000/355/1800 msec, matrix = 192 × 192 × 176, resolution = 1 × 1 × 1 mm³) images were acquired on a 3-Tesla scanner (Siemens Medical, Erlangen, Germany). The BOLD-fMRI images were acquired with gradient-echo EPI sequence (TR/TE = 1450/30 msec, matrix = 64 × 64, resolution = 3.4 × 3.4 × 5 mm³), while the subject was performing the breath-hold task.

The timing of the breath-hold task was presented visually to the subject by a computer using the software Eprime Professional 2.0 (Psychology Software Tools, Inc., Pittsburgh, PA) (Fig. 1). A breath-hold rehearsal was given before fMRI. No special instruction on the breath holding after inhalation or exhalation was delivered to the subjects because BOLD signal amplitude has been suggested to depend on the depth of inspiration immediately preceding the breath-hold period,²⁷ and end-expiration breath hold may induce significant head motion.²⁸

FIG. 1.

The paradigm of breath-hold challenge. Subjects were instructed via visual cues to perform six epochs of 30-sec breath hold interleaved with 60–90 sec of free breathing. They were asked to have free breathing when white crosshair was projected on the screen with a black background and hold their breaths when the white crosshair changed to red. Yellow shades represent the breath-hold epochs.

The partial pressure of O₂ (PO₂) and CO₂ (PCO₂) and respiration were measured simultaneously with MRI acquisition. A small nasal tubing was placed at the subject's nostril to sample PCO₂ and PO₂ via gas analyzers (Capstar-100, Oxystar-100, CWE, Inc., Ardmore, PA) after calibrating to the barometric pressure immediately before the scan and correcting for vapor pressure. Respiration was measured using a respiratory bellow connected to a spirometer (FE141, ADInstruments, Inc., Colorado Springs, CO). All the physiological measurements were synchronized using trigger signals from the MRI scanner.

Physiological data processing

The physiological data were analyzed using Matlab R2014a (Mathworks, Inc., Natick, MA). Technical delay of PCO₂ and PO₂ was corrected by cross-correlating the time series of PCO₂ and PO₂ with respiratory phases determined from the respiratory bellow. End-inspiration (I) and end-expiration (E) were defined on the time series of PO₂ and PCO₂, and were verified by the inspiratory and expiratory phases on the respiration time series. Changes in the gas parameters in the breath-hold periods were interpolated by the values measured immediately before and after the breath-hold periods.

The bER, the ratio of the change in PO₂ (ΔPO₂ = inspired PO₂–expired PO₂) to the change in PCO₂ (ΔPCO₂ = expired PCO₂–inspired PCO₂), is measured between end-inspiration and end-expiration in each respiratory cycle. The bER, ΔPO₂ and ΔPCO₂ measured in the baseline periods—i.e., the period before the first breath-hold epoch—were calculated. Because the breath-hold duration might vary from one epoch to another, the changes of bER, ΔPO₂, and ΔPCO₂ in each breath-hold epoch were normalized to a 30-sec breath-hold epoch by dividing the changes to the actual breath-hold duration in seconds and multiplying by 30.²¹ The normalized changes of each parameter over the six breath-hold epochs were then averaged.

Image processing

All the BOLD-fMRI data were analyzed using software AFNI.²⁹ The first 12 volumes of each functional dataset collected before equilibrium magnetization was reached were discarded. Each functional dataset was corrected for slice timing, motion-corrected, and coregistered to the first image of the functional dataset. The time series of each voxel was detrended with the fifth order of polynomials to remove the low drift frequency, and the components of motion were removed using orthogonal projection.

Head motions were quantified and compared between patients and controls (Supplementary Fig. S1). The clean time series was then normalized to its mean intensity value. Voxels located within the ventricles and outside the brain were excluded from the following analyses of functional images. Individual subject brain volumes with time series of percent BOLD signal changes (ΔBOLD) were derived.

Hilbert Transform analysis³⁰ was used to measure the cross-correlation between the ΔBOLD and the reference time series of RGE metrics (bER, ΔPO₂, and ΔPCO₂) in each voxel. Fisher Z-transformation was used to convert the cross-correlation coefficients in the individual subject maps to Fisher's z scores. Individual subject brain volumes with Fisher's z scores were then registered onto the subjects' anatomical scans and transformed to the standardized space of Talairach and Tournoux³¹ for group analysis.

Wavelet Transform coherence³² was used to measure the dynamic changes in coupling between the time series of RGE metrics and ΔBOLD extracted from brain regions in the time-frequency domain. The analysis procedures were the same as those described in our other breath-hold study.¹⁸ The brain regions in gray and white matter were segmented using the software FreeSurfer.^33,34

An example of squared wavelet coherence between bER and ΔBOLD in gray matter from a representative patient under breath-hold challenge is shown in Supplementary Fig. S2. The percentage of significant coherence (sCoh%) defined as the percentage of time during which the magnitude of wavelet transform coherence was significantly different from the red noise outside the cone of influence was used to quantify the dynamic coupling between the time series of each RGE metric and ΔBOLD (Supplementary Fig. 2C).

Statistical analysis

Continuous variables (i.e., age and RGE metrics) were summarized as median and interquartile range (IQR), and group (mTBI and control) differences were assessed using the Kruskal-Wallis tests. Categorical variable (i.e., gender) was summarized as frequencies; group difference was assessed using the Chi-square test. Spearman correlation analyses were used to correlate in pairs the RPQ scores, time after injury, age, and breath-hold duration, and normalized changes in RGE metrics (bER, ΔPO₂ and ΔPCO₂) averaged over the six breath-hold epochs. Significant differences or correlations were considered at p < 0.05.

Unpaired t tests (3dttest++) in AFNI were used to compare the cerebrovascular responses to breath-hold challenge (i.e., Fisher's z scores from Hilbert Transform analysis) between patients and controls. Multiple comparisons were corrected using the equitable thresholding and clustering (ETAC) method.³⁵ The group results were painted onto the averaged brain surfaces reconstructed by FreeSurfer.^33,34

The sCoh% at the phase lags of 0±π/2 and ±π/2 from the individual subject wavelet transform analysis were separately plotted for mTBI and control groups to explore the Fourier periods/frequency bandwidths that oscillations of ΔBOLD in gray and white matter were in synchrony with each physiological time series of bER, ΔPO₂, and ΔPCO₂ when the subjects were performing breath-hold tasks.

The sCoh% of RGE metrics with ΔBOLD in the six frequency bandwidths (0.008–0.015 Hz, 0.016–0.03 Hz, 0.031–0.062 Hz, 0.063–0.124 Hz, 0.125–0.249 Hz, and 0.25–0.5 Hz) extracted in the brain regions showing significant differences in Fisher's z scores between mTBI and control groups was calculated for each subject. The distribution of sCoh% in these brain regions of the mTBI and control groups was then compared using the Kolmogorov-Smirnov test. Significant differences in the distribution of the sCoh% were considered at p < 0.05.

Results

Twenty subjects aged from 20 to 57 years were included. Ten of them were patients with a diagnosis of mTBI (median = 38 years, IQR = 25.75–46.75 years; 5M and 5F). The selected clinical characteristics of the patients are shown in Table 1. The remaining 10 were control subjects without brain injury history (median = 28 years, IQR = 25.25–34.25 years; 6M and 4F). No significant differences in age (p = 0.34) and sex (p = 0.65) were found between patients with mTBI and control subjects.

Table 1.

Selected Clinical Characteristics of the Patients with Mild Traumatic Brain Injury

Subjects	Sex	Age	Time after injury (days)	Cause of injury	RPQ scores
m1	F	44	846	Motor vehicle accident	34
m2	F	47	1704	Motor vehicle accident	27
m3	M	38	109	Sports injury (basketball)	6
m4	M	37	1614	Motor vehicle accident	29
m5	M	49	1579	Motor vehicle accident	45
m6	M	57	237	Sports injury (skiing)	2
m7	F	22	630	Fall	41
m8	F	20	1000	Fall	0
m9	F	21	2588	Sports injury (soccer)	16
m10	F	46	481	Motor vehicle accident	24

RPQ, Rivermead Post Concussion Questionnaire; M, male; F, female.

The bER, ΔPO₂, and ΔPCO₂ measured in the baseline periods were summarized in Table 2. The averaged breath-hold duration, the averaged changes of bER, ΔPO₂, and ΔPCO₂ over the six breath-hold epochs were also included. The averaged changes of ΔPO₂ during breath holding were almost three to four folds larger than those of ΔPCO₂. No significant difference in the RGE metrics was found between mTBI and control groups in the baseline periods and breath-hold epochs (p > 0.05). No structural abnormality was identified in the T1-weighted and T2 FLAIR images of all subjects.

Table 2.

Baseline Respiratory Gas Exchange Values before the First Breath-Hold Epoch and the Averaged Changes of Respiratory Gas Exchange Metrics during Breath Holding

Baseline mean values (standard deviation [SD]) of P_ETCO₂, P_ETO₂, ΔPCO₂, ΔPO₂, and bER for all subjects before the first breath-hold epoch (left). The changes (SD) of breath-hold duration, ΔPCO₂, ΔPO₂, and bER from the onset to the end of the breath-hold epoch averaged more than six breath-hold epochs (right). Medians and interquartile ranges [IQR] were used to summarize respiratory gas exchange metrics in groups. No significant differences was found in the respiratory gas exchange metrics between patients with mild traumatic brain injury and controls (p > 0.05) (bottom row).

Correlation among RPQ scores, time after injury, age, and changes in RGE metrics

The correlation results among RPQ scores, time after injury, age, breath-hold durations, and RGE metrics changes are summarized in Fig. 2. During breath holding, a significant correlation was found between RPQ scores and the averaged changes in bER (Spearman rho = -0.79, p = 0.01), between RPQ scores and the averaged changes in ΔPO₂ (Spearman rho = -0.70, p = 0.03), and between the averaged changes in bER and ΔPO₂ (Spearman rho = 0.95, p < 0.001). No significant correlation was found in the remaining correlation pairs (p > 0.05).

FIG. 2.

Correlation among Rivermead Post Concussion Questionnaire (RPQ) scores, time after injury, age, breath-hold duration, and respiratory gas exchange (RGE) metric changes from the onset to the end of the breath-hold periods averaged over six breath-hold epochs. A significant correlation was found between RPQ scores and the averaged changes in breath-by-breath O₂-CO₂ exchange ratio (bER), between RPQ scores and the averaged changes in partial pressure of oxygen (ΔPO₂₎, and between the averaged changes in bER and ΔPO₂. No significant correlation was found between breath-hold duration and RGE metrics.

Comparison of the cross-correlation of ΔBOLD with RGE metrics between patients with mTBI and controls

Time series of ΔBOLD in left gray matter and the corresponding RGE metrics in a representative patient with mTBI and a control subject under breath-hold challenge are shown in Fig. 3. The time series of bER followed more closely to ΔBOLD than ΔPO₂ and ΔPCO₂ in both representative patient and control subject.

FIG. 3.

Time series of regional change in blood-oxygenation-level-dependent (ΔBOLD) signals in left gray matter and the corresponding breath-by-breath O₂-CO₂ exchange ratio (bER), change in partial pressure of oxygen (ΔPO₂), change in partial pressure of carbon dioxide (ΔPCO₂), end tidal partial pressure of oxygen (P_ETO₂), and end tidal partial pressure of carbon dioxide (P_ETCO₂) in a representative patient with mild traumatic brain injury (mTBI) (A) and a representative control subject without brain injury history (B) under breath-hold challenge. Shaded areas represent breath-hold periods. Thin green lines represent PO₂, while the thin blue lines represent PCO₂. Thick green lines and thick blue lines represent P_ETO₂ and P_ETCO_2, respectively. In the representative control subject, the time series of bER followed closely to the ΔBOLD. A significant increase in ΔPO₂ was observed in all breath-hold epochs, while an increase in ΔPCO₂ was shown in only three of six breath-hold epochs. In the representative patient, an increase in ΔBOLD but decreases in both ΔPO₂ and ΔPCO₂ were observed from the second to the sixth breath-hold epochs. The bER was found to increase in the second and the fifth breath-hold epochs.

Compared with control subjects, a significant decrease in the cross-correlation between ΔBOLD and bER was mapped by fMRI in patients with mTBI at superior and middle frontal, precentral, and postcentral, insular, superior parietal, middle and inferior temporal, paracentral, precuneus, posterior cingulate, cuneus, lingual, hippocampal and parahippocampal areas (p_fdr < 0.05) (Fig. 4). No significant difference was found in the cross-correlation of ΔBOLD with ΔPO₂ and ΔPCO₂ between patients and controls.

FIG. 4.

The comparison of the cross-correlation of regional change in blood-oxygenation-level-dependent (ΔBOLD) signals with breath-by-breath O₂-CO₂ exchange ratio (bER) between patients with mild traumatic brain injury (mTBI) and controls. Warmer colors represent a stronger correlation in patients, while cooler colors represent a weaker correlation in patients when compared with controls. Diffuse decrease in correlation between ΔBOLD and bER was found in patients with mTBI. Multiple comparisons were corrected at a false discovery rate p_fdr< 0.05. SF, superior frontal; MF, middle frontal; PrC, precentral; PoC, postcentral; Ins, insular; SP, superior parietal; MT, middle temporal; IT, inferior temporal; ParaC, paracentral; PreC, precuneus; PC, posterior cingulate; Cun, cuneus; Ling, lingual; Parahipp, parahippocampal.

Dynamic coupling between RGE metrics and ΔBOLD

In both the patients with mTBI and the control subjects, the dynamic coupling indicated by the sCoh% between bER and ΔBOLD at phase lag of 0±π/2 was the highest in the frequency range of 0.008–0.03 Hz while that between ΔPCO₂ and ΔBOLD was the lowest in the same frequency range (Fig. 5). The dynamic coupling between RGE metrics and ΔBOLD was weak at the phase lag of π±π/2, and the sCoh% stayed around 20% or below (Supplementary Fig. S3). Comparing with control subjects, the dynamic coupling of bER and ΔPO₂ with ΔBOLD in the frequency range of 0.008–0.03 Hz and at phase lag of 0±π/2 was reduced in patients (Fig. 5).

FIG. 5.

The percentage of significant coherence between time series of respiratory gas exchange (RGE) metrics (breath-by-breath O₂-CO₂ exchange ratio [bER], change in partial pressure of oxygen [ΔPO₂], change in partial pressure of carbon dioxide [ΔPCO₂]) and regional change in blood-oxygenation-level-dependent (ΔBOLD) signals at the phase lag 0±π/2 in patients with mild traumatic brain injury (mTBI) and controls. The mean percentages of significant coherence in the frequency bandwidths from 0.008 to 0.25 Hz were plotted (thick color lines). Color shaded areas represent standard error of the mean. Large variations in the percentages of significant coherence were found across the patients and the reduced coherence of bER and ΔPO2 with ΔBOLD compared with controls. LRGM, left and right gray matter; LRWM, left and right white matter.

In the brain regions showing significant differences in the Fisher's z scores between mTBI and control groups, the distribution of the sCoh% for the dynamic coupling of bER with ΔBOLD was significantly different between mTBI and control groups at the frequency bandwidths of 0.008–0.015 Hz and phase lag of 0±π/2 (p = 0.03) (Fig. 6). No significant difference in the distribution of sCoh% was found between patients and controls for the dynamic coupling of ΔPO₂ and ΔPCO₂ with ΔBOLD in all frequency bandwidths and phase lag of 0±π/2.

FIG. 6.

Comparison of the percentage of significant coherence of respiratory gas exchange (RGE) metrics (breath-by-breath O₂-CO₂ exchange ratio [bER], change in partial pressure of oxygen [ΔPO₂], change in partial pressure of carbon dioxide [ΔPCO₂]) and regional change in blood-oxygenation-level-dependent (ΔBOLD) signals in selected brain regions at the phase lag 0±π/2 between patients with mild traumatic brain injury and controls. Each grey circle represents the percentage of significant coherence for each subject in the frequency range shown on the x-axis. The thick horizontal black line, the box, and the vertical rod represent the mean, 95% confidence interval, and standard deviation of the group. The distribution of the percentage of significant coherence of bER with ΔBOLD in patients was significantly different from controls at 0.008–0.015 Hz (p = 0.03). LRGM, left and right gray matter; LRWM, left and right white matter.

Discussion

To our knowledge, this is the first study to characterize cerebrovascular responses with bER under brief breath-hold challenge in patients. We showed that the dynamic coupling of cerebrovascular responses with bER and ΔPO₂ was stronger than that with ΔPCO₂ in patients with chronic mTBI, similar to our previous breath-hold results on healthy subjects.¹⁸ In patients with mTBI, the bER-ΔBOLD correlation under brief breath-hold challenge was significantly reduced in multiple brain regions when compared with control subjects who did not have a brain injury history.

The averaged changes in bER and ΔPO₂ in the breath-hold epochs decreased with increased symptom severity indicated by RPQ scores, while no such changes were observed between ΔPCO₂ and RPQ scores. Patients with more severe symptoms showed a smaller increase in bER during breath holding. The correlation between the symptom severity and the changes in bER and ΔPO₂ during breath holding could not be attributed to the breath-hold duration, because the changes in RGE metrics (bER, ΔPO₂, and ΔPCO₂) were normalized to a 30-sec breath-hold epoch, and the breath-hold duration was not correlated with RPQ, age, or changes in RGE metrics.

Many research teams^19–24 including ours¹⁸ reported earlier a larger increase of O₂ uptake than CO₂ release under breath-hold challenge, resulting in reduced RER or increased bER for healthy subjects. Our current gas measurements on patients with mTBI again suggest the synergistic role of mild hypoxemia and mild hypercapnia during breath holding. A strong correlation of the changes in bER with ΔPO₂ but not with ΔPCO₂ in patients during breath holding (Fig. 2) indicates that bER, being the ratio of ΔPO₂ to ΔPCO₂, is predominantly affected by ΔPO₂.

In this study, we showed a significant decrease in the correlation between ΔBOLD and bER in response to breath-hold challenge in multiple brain regions of patients using Hilbert Transform analysis (Fig. 4). These brain regions are responsible for the brain functions for attention, memory, visual processing, auditory processing, executive and autonomic functions.³⁶ They also showed a general pattern that resembles the structurally atrophied regions of patients with mTBI reported by Warner and associates.³⁶ The intermittent cross-correlations between bER and ΔBOLD were then characterized as a function of both time and frequency in these brain regions using Wavelet Transform coherence analysis based on continuous wavelet transform.^32,37

Like our previous breath-hold study in healthy individuals,¹⁸ bER had the most robust dynamic coupling with cerebral ΔBOLD for both patients with mTBI and controls at the frequency range of 0.008–0.03 Hz; ΔPCO₂ had the weakest coupling while ΔPO₂ had an intermediate coupling (Fig. 5). Compared with controls, the coherence of bER with ΔBOLD was reduced in patients at the frequency range of 0.008–0.03 Hz. Concerns of using BOLD signal as a CBF surrogate had been addressed by another independent verification of the coupling between bER and CBF¹⁸ using CBF velocity measured by transcranial Doppler ultrasound at a sampling rate of 100 Hz.

The dynamic coupling between bER and ΔBOLD was then measured in the selected brain regions showing a decreased correlation in Hilbert Transform analysis and showed significant intersubject variations among patients at the frequency range of 0.008–0.015 Hz compared with controls (Fig. 6). Such intersubject variability may be due to the heterogeneity of brain injuries.

The bER, but not ΔPO₂ or ΔPCO₂, showed a significant difference in coupling with ΔBOLD between patients and controls at the frequency range of 0.008–0.015 Hz. Given that bER is predominantly affected by ΔPO₂, we attributed the somewhat surprising lack of statistical significance for ΔPO₂-ΔBOLD to the effect of ventilatory volume fluctuations common to both ΔPO₂ and ΔPCO₂. In comparison with ΔPO₂, the format of bER as a ratio of ΔPO₂ to ΔPCO₂ has the advantage of being less affected by the ventilatory volume fluctuations resulting from isolated deep breaths. Future studies with a larger sample size are warranted to clarify the role between bER and ΔPO₂.

Conclusion

The bER is a more useful metric than ΔPCO₂ (i.e., P_ETCO₂) to characterize the cerebrovascular responses to breath-hold challenge in patients with mTBI. The correlation between bER changes and symptom severity during breath holding also suggests a relationship between RGE and post-concussion symptoms. Future studies with a large cohort of subjects would help explore the underlying mechanisms between RGE, cerebral hemodynamics, and post-concussion symptoms in patients with mTBI.

Authors' Contributions

Suk-Tak Chan, design and conceptualized study; data collection and analysis; data interpretation; drafted the manuscript for intellectual content; Cora Ordway, data collection and analysis; Ronald J. Calvano, design and conceptualized study; data analysis; data interpretation; revision of manuscript; Ferdinando S. Buonanno, design and conceptualized study; data interpretation; revision of manuscript; Bruce R. Rosen, drafted the manuscript for intellectual content; revision of manuscript; Kenneth K. Kwong, design and conceptualized study; data collection and analysis; data interpretation; drafted the manuscript for intellectual content; revision of manuscript.

Funding Information

This research was performed in whole at the Athinoula A. Martinos Center for Biomedical Imaging at the Massachusetts General Hospital, using resources provided by the Center for Functional Neuroimaging Technologies, P41EB015896, a P41 Biotechnology Resource Grant supported by the National Institute of Biomedical Imaging and Bioengineering (NIBIB), National Institutes of Health, as well as the Shared Instrumentation Grant S10RR023043. This study was also supported, in part, by National Center for Complementary and Integrative Health (NCCIH) grant (R21AT010955).

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3