Detailed assessment of the hemodynamic response to psychosocial stress using real-time MRI


  • Alexander Jones PhD,

    Corresponding author
    1. Centre for Cardiovascular Imaging, UCL Institute of Child Health & Great Ormond Street Hospital for Children, London, United Kingdom
    2. Department of Vascular Physiology, UCL Institute of Child Health, London, United Kingdom
    • Cardiovascular Unit, Great Ormond Street Hospital, Great Ormond Street, London WC1N 3JH, UK
    Search for more papers by this author
  • Jennifer A. Steeden MEng,

    1. Centre for Cardiovascular Imaging, UCL Institute of Child Health & Great Ormond Street Hospital for Children, London, United Kingdom
    Search for more papers by this author
  • Jens C. Pruessner PhD,

    1. Douglas Institute, Department of Psychiatry, McGill University, Montreal, Quebec, Canada
    Search for more papers by this author
  • John E. Deanfield MB,

    1. Department of Vascular Physiology, UCL Institute of Child Health, London, United Kingdom
    Search for more papers by this author
  • Andrew M. Taylor MD,

    1. Centre for Cardiovascular Imaging, UCL Institute of Child Health & Great Ormond Street Hospital for Children, London, United Kingdom
    Search for more papers by this author
  • Vivek Muthurangu MD

    1. Centre for Cardiovascular Imaging, UCL Institute of Child Health & Great Ormond Street Hospital for Children, London, United Kingdom
    Search for more papers by this author



To demonstrate that combining the Montreal Imaging Stress Task (MIST) with real-time cardiac magnetic resonance imaging (MRI) allows detailed assessment of the cardiovascular mental stress response.

Materials and Methods

22 healthy volunteers (1:1 M:F, 26–64 years) underwent MRI during rest and the MIST. Real-time spiral phase contrast MR, accelerated with sensitivity encoding (SENSE) was used to assess stroke volume (SV), and radial k-t SENSE was used to assess ventricular volumes. Simultaneous heart rate (HR) and blood pressure (BP) measures allowed calculation of cardiac output (CO), systemic vascular resistance (SVR), and arterial compliance (TAC). Endocrine responses were assessed using salivary cortisol.


In response to stress, BP increased due to increased CO and reduced TAC but not increased SVR, which fell. HR, not SV, determined CO increases. Greater BP responses occurred in men due to greater CO increases and relatively higher SVR. Older participants had greater BP responses due to greater falls in TAC. Greater cortisol response was correlated with greater falls in TAC but resting cortisol and TAC were not related.


This new approach allows detailed, accurate assessment of stress physiology. Preliminary findings suggest stress exposes relationships, not seen at rest, of cardiovascular function with age, sex, and endocrine function. J. Magn. Reson. Imaging 2011;33:448–454. © 2011 Wiley-Liss, Inc.

MENTAL STRESS is a potent stimulator of the cardiovascular system and has been linked to a number of cardiovascular diseases (1–5). Characterization of the cardiovascular response to mental stress is crucial for understanding the underlying mechanisms. Traditionally, heart rate (HR) and blood pressure (BP) are the main parameters of investigation, although they do not fully describe hemodynamic responses. A comprehensive approach would include measurement of cardiac output (CO), systemic vascular resistance (SVR), total arterial compliance (TAC), and ventricular function. Very different modulation of these parameters may result in identical BP responses but with dissimilar health implications.

Cardiac magnetic resonance imaging (MRI) is the reference standard method of measuring CO and ventricular function and MR data can be combined with BP to calculate SVR and TAC (6–8). Unfortunately, standard cardiac-gated sequences have limitations that make them unsuitable for mental stress studies. These include: requirement for multiple breatholds, long acquisition times, and intolerance to nonperiodic motion. Thus, MR has not been used to assess cardiovascular responses to mental stress. However, development of high temporal resolution real-time MR sequences makes this possible. Previous studies have shown the suitability of this approach by measuring the cardiovascular response to exercise (9).

Another necessary component for this approach is a stress task suitable for an MR environment. The Montreal Imaging Stress Task (MIST) was developed for neuroimaging functional MRI studies (10). Since the MR environment is confined and allows limited interaction with participants, the MIST was designed to overcome these impediments and incorporate crucial elements required to evoke a neuroendocrine stress response (11). These are: ego-involvement (participants must engage with the task and care about the result); social evaluative threat (participants must feel that their performance will be judged by others, and that this judgment will be negative if they perform poorly); and uncontrollability (participants must feel that they have limited authority to alter the sequence of events during the task).

The principal aim of this study was to demonstrate that a combination of the MIST with real-time cardiac MR was feasible to enable a detailed assessment of the cardiovascular response to mental stress. Secondary aims were to assess age and sex differences in the responses and associations with endocrine stress responses. To demonstrate broad applicability, this novel methodology was tested in a wide age range of healthy adult men and women.


Study Population

Twenty-two healthy nonsmoking volunteers (11 men and 11 women) with a median age of 36.2 (range 25.9–63.7) years were recruited. Participants with cardiovascular disease, endocrine disorders of the hypothalamic–pituitary axes, or contraindications for MR, such as pregnancy or MR-incompatible implants, were excluded. The study was approved by the local research ethics committee and informed, written consent was obtained from all participants.

Montreal Imaging Stress Task

Mental arithmetic tasks were displayed on a computer screen (Fig. 1), visible to the participants by way of a mirror. Answers were selected from a circularly arranged set of numbers (0–9) using a fiberoptic mouse. A loud and unpleasant noise was played in response to an incorrect answer or a timeout. Pointers on a multicolored “performance” bar continuously indicated both participant performance and that of an “average performance” of a “peer group.” The three colored regions of this bar indicated poor (red), mediocre (yellow), and good (green) performance. Participants were asked to reach at least “average performance” to ensure useful data collection. They were also told that they would be scored and judged by the investigators. To increase performance pressure, a high-tempo musical track was played through headphones throughout the task.

Figure 1.

A screenshot of the MIST, showing (A) an example mathematical problem, (B) a blue timer bar advancing from left to right, (C) a rotary dial for selecting answers (0–9), and (D) a feedback box stating that the answer was “Incorrect.” The “peer” indicator (E) is above the colored bar (F) and in the green zone, suggesting a good performance and the participant's own indicator (G) is below the bar in the red zone, suggesting a poor performance. Provision of a proxy of peer performance is designed to enhance social evaluative threat.

Unknown to participants, the computer responded to correct answers by increasing the difficulty and reducing the time available for the questions, and the “average performance” was not of a peer group but was computer-generated. Thus, regardless of competence, participant performance was negatively biased such that they appeared to perform poorly in comparison to their “peer group.” During salivary cortisol collection between the two stress periods, their attention was directed to the performance indicators by the investigator and it was restated that a poor performance would not provide useful data. This ensured participant perception of their poor performance and of the possibility of negative judgment, stimulating a stress response. Importantly, upon completion of the protocol all participants were informed that their poor performance was artificial and that their data was useful.


All imaging was performed on a 1.5 T MR scanner (Avanto, Siemens Medical Solutions, Erlangen, Germany) using two spine coils and one body-matrix coil (giving a total of 12 coil elements). This is the standard coil configuration for cardiac MRI at our institution.

The mental stress protocol was: 1) a 1-minute practice session, with no scoring, time limits, or evaluation; 2) two consecutive 5-minute stress periods with scoring, time limits, and evaluation; 3) a recovery period beginning 25 minutes after the onset of the first stress period. Participants were told to rest after the termination of the stress protocol. Oscillometric BP was measured in the nondominant arm at 1-minute intervals. MR data were acquired on six occasions: at rest, at 1 and 3 minutes into the two 5-five minute stress periods, and finally during the recovery period. Each MR data acquisition lasted ≈30 seconds.

Real-Time Volume Assessment

Ventricular volumes were assessed using a real-time radial k-t SENSE sequence (field of view [FOV]: ≈380 mm, matrix: 128 × 128, voxel size: ≈3.0 × ≈3.0 × 10 mm, TE/TR: ≈1.14/≈2.3 msec, flip angle: 38°, pixel bandwidth [BW]: 1500 Hz/pixel, radial spokes: 128, k-t SENSE acceleration factor: 8, scan time: ≈1.5 seconds per slice, temporal resolution: ≈35.5 msec). Eleven to 13 contiguous slices were acquired in the short axis to ensure coverage of the ventricle. Image acquisition was performed during free breathing.

Real-Time Flow Assessment

Real-time flow quantification was performed through-plane in a cross-section of the ascending aorta as it passes the bifurcation of the pulmonary arteries. A uniform density spiral PC sequence with eight interleaves, undersampled by a factor of four was used (matrix: 128 × 128, voxel size: 3.9 × 3.9 × 6 mm, TE/TR: 1.93/9.63 msec, flip angle: 25°, pixel BW: 1860 Hz/pixel, VENC: 180 cm/s, temporal resolution: 38.5 msec) (9). For each scan, 76 consecutive frames were acquired over 3 seconds. The sampling pattern was rotated for each frame so that four consecutive frames comprised a fully sampled k-space with eight interleaves. These undersampled data were reconstructed offline using an iterative SENSE algorithm. In order to reconstruct undersampled spiral data using iterative SENSE, the FOV must be larger than the object. In this study of adults, the FOV was set at 500 mm due to double oblique slice angulations. All coil sensitivity and regularization information required for the reconstruction process was calculated from the sum-of-squares of the coil data over all time frames.

Data Processing

All images were processed using in-house plug-ins for the open-source software OsiriX (OsiriX Foundation, Geneva, Switzerland). Measurement of ventricular volumes was performed as previously described (12, 13). Flow images were manually segmented (using the modulus images) and SV and CO were measured. SVR (measured in mmHg.L−1.min−1, also known as WU) was calculated by dividing the mean blood pressure by CO. Compliance was calculated by optimization of a two-element windkessel model, as previously described (7, 8). Briefly, CO, mean arterial pressure, and SVR were used as inputs to a windkessel model. Pulse pressure was calculated for a series of modeled pressure curves generated using a range of compliance values from 0.1 to 5.0 mL.mmHg−1 in increments of 0.01. The compliance value that gave the smallest error between the modeled pulse pressure and the true pulse pressure was taken to be the true compliance.

Salivary Cortisol

Using established protocols, six saliva samples were obtained during the course of the experiment (Salivette Cortisol, Sarstedt, Nümbrecht, Germany): just after the first rest period, and then at median delays of 5, 12, 24, and 34 minutes following the onset of the first stress period (Fig. 2). Concentration of salivary free cortisol was measured using a commercially available chemiluminescence-immuno-assay (CLIA; IBL, Hamburg, Germany).

Figure 2.

Geometric mean ±95% confidence interval (CI) salivary cortisol at rest and in response to two 5-minute periods of stress (vertical gray bars). *P < 0.01, **P < 0.002 for pair-wise 2-sided t-test comparisons of mean values at each timepoint with initial values. P-value thresholds were Bonferroni-corrected to account for five comparisons.

Statistical Analysis

For all statistical tests, P-values below 0.05 were taken to be significant. Where possible, right-skewed variables were log-transformed prior to parametric testing. If transformation to normality could not be achieved, nonparametric comparisons were made using Wilcoxon rank-sum tests. Where multiple comparisons were made for time-series data, significance thresholds were Bonferroni-adjusted. For the purpose of calculating cardiovascular stress responses, mean values across both stress periods were used. Pearson partial correlation was used to examine relationships of cardiovascular parameters with age and cortisol, adjusting for age and sex as necessary. Cortisol stress response (CSR) was defined as the difference between the greater of the cortisol concentrations at timepoints 5 and 6 and the lowest value at timepoints 1–3 (see Fig. 2). To account for decline throughout the day as part of diurnal rhythm, salivary cortisol measures were adjusted by removing the inverse linear relationship between time of day and log-transformed cortisol concentration (r = −0.31, P = 0.0002). End-diastolic volume (EDVi), stroke volume (SVi), cardiac output (COi), systemic vascular resistance (SVRi), and total arterial compliance (TACi) were indexed to body surface area (BSA).


Image quality at rest and during stress was very similar and all data was suitable for analysis. Figure 3 shows representative images from the real-time sequences.

Figure 3.

Examples of the (a) real-time radial k-t SENSE sequence at end diastole and end systole, and of the (b) uniform density spiral PC sequence show no appreciable evidence of an effect of stress on image quality.

Cardiovascular Response to MIST

Arterial BP (systolic [SAP], mean [MAP], and diastolic [DAP] arterial pressures), COi, and HR all increased in response to mental stress (all P<.00025). SVRi and TACi both fell significantly during stress (Fig. 4). Other cardiovascular parameters did not change significantly. TACi and DAP remained significantly reduced during recovery, while other parameters returned to baseline. The second stress period provoked a significant incremental response of HR, COi, SVRi, and TACi in comparison to the first (Fig. 4).

Figure 4.

Mean ±95% CI values (geometric values for a,d,f, and systolic arterial pressure in b) at rest and in response to two 5-minute periods of stress (vertical gray bars) for (a) heart rate; (b) systolic, mean, and diastolic arterial pressure; (c) stroke volume indexed to body surface area (BSA); (d) cardiac output indexed to BSA; (e) systemic vascular resistance indexed to BSA and (f) total arterial compliance indexed to BSA. *P < 0.0125, **P < 0.0025, ***P < 0.00025 for pair-wise 2-sided t-test comparisons of mean values during each stress period, or the recovery period, with those at rest. ++P < 0.0025 for pair-wise 2-sided t-test comparisons of values during first and second stress periods. P-value thresholds were Bonferroni-corrected to account for four comparisons per variable.

Sex and Age Differences

Men were taller (median: 182 vs. 162 cm; P = 0.0002) and heavier (median: 78 vs. 61 kg; P = 0.002) than women, but there were no sex differences in age, BMI (median: 23.6 [range: 20.2–30] in men vs. 23 [19.2–27.3] in women; P = 0.41) or proportion of overweight participants. At rest, men had a trend towards lower HR, significantly lower COi, and greater SVRi than women, but no difference in BP (Table 1). During mental stress and recovery, men had greater BP (SAP, MAP, and DAP), with no significant difference in COi and SVRi during stress. They had higher SVRi during recovery. However, the increase in COi in men was significantly greater than that in women when adjusted for SVRi during stress (P = 0.02). The other parameters did not differ by sex during rest or stress.

Table 1. Mean (SD) Cardiovascular Measures During Rest, Stress, and Recovery
 Women (n = 11)Men (n = 11)P-valuesa
  • a

    P-values refer to 2-sided t-test comparisons between sexes.

  • b

    Mean of all four measures during stress.

  • c

    Geometric means and SDs.

  • *

    P < 0.05.

  • **

    P < 0.01.

  • ***

    P < 0.001 for within-sex 2-sided t-test comparisons with mean value at rest.

HR (bpm)c70.3 (1.28)90.4 (1.24)**70.0 (1.23)59.7 (1.17)80.6 (1.18)***62.4 (1.18)*0.0810.180.16
SAP (mmHg)c104.3 (1.10)118.2 (1.10)***104.4 (1.06)109.2 (1.10)129.8 (1.12)***112.8 (1.10)**0.250.0450.034
MAP (mmHg)80.3 (7.4)89.0 (7.6)***79.7 (5.0)83.1 (6.2)96.9 (8.2)***85.0 (6.1)0.340.0300.035
DAP (mmHg)62.1 (8.8)65.9 (6.5)*56.2 (9.1)**65.7 (5.3)75.2 (8.5)***64.0 (8.0)0.260.00910.047
COi (L.min−1.m−2)c3.87 (1.20)4.90 (1.28)**3.75 (1.21)3.27 (1.16)4.56 (1.22)***3.32 (1.21)0.0290.470.15
SVRi (WU.m2)21.0 (3.9)18.7 (4.5)*21.6 (4.4)25.7 (4.5)21.6 (3.9)**25.7 (4.1)0.0170.120.033
SVi (mL.m−2)55.4 (6.9)54.6 (6.7)54.0 (7.7)55.3 (8.3)56.9 (7.2)53.6 (7.4)0.970.450.91
EDVi (mL.m−2)69.6 (9.0)72.0 (8.9)73.0 (7.3)*74.7 (12.9)79.5 (14.7)*73.8 (12.1)0.300.160.85
EF (%)80.2 (10.2)76.4 (7.4)74.1 (8.1)**74.6 (7.3)72.8 (8.6)73.5 (9.4)0.150.310.87
TACi (mL.mmHg−1.m−2)c0.61 (1.20)0.46 (1.20)**0.53 (1.20)*0.72 (1.30)0.53 (1.28)***0.59 (1.22)**

During stress, greater age was associated with higher SAP (r = 0.47, P = 0.03), a trend towards higher MAP (r = 0.4, P = 0.07), and lower TACi (r = −0.45, P = 0.04). During rest and recovery, these parameters did not correlate with age. Age was also correlated with DAP during recovery (r = 0.55, P = 0.01) but not significantly so during rest or stress. EDVi was inversely related to age at rest (r = −0.56, P = 0.008) and recovery (r = −0.55, P = 0.009) but not significantly so during stress (r = −0.39, P = 0.08). Ejection fraction (EF) was positively correlated with age at all times (rest: r = 0.63, P = 0.002; stress: r = 0.5, P = 0.02; recovery: r = 0.75, P = 0.0001).

Endocrine Response to MIST

The MIST provoked a CSR in all cases. Geometric mean salivary cortisol levels rose by 48%, peaked 24 minutes after stress onset, and were still raised significantly at 34 minutes (Fig. 2). Response magnitude varied widely from 0.2 to 24.1 nmol.L−1. There were no age or sex differences in baseline cortisol or CSR.

There was a significant inverse association between CSR and TACi during stress (r = −0.59, P = 0.006) and recovery (r = −0.48, P = 0.03), but not rest (r = −0.32, P = 0.17). HR was strongly, positively correlated with CSR at all times (rest: r = 0.54, P = 0.01; stress: r = 0.65, P = 0.002; recovery: r = 0.60, P = 0.005). CSR was correlated inversely with EDVi at all times (rest: r = −0.59, P = 0.006; stress: r = −0.61, P = 0.004; recovery: r = −0.54, P = 0.01) and with SVi at rest (r = −0.50, P = 0.02).


The key findings in this study were: 1) It was possible to comprehensively assess the cardiovascular response to mental stress using a combination of the MIST and real-time cardiac MR; 2) The cardiovascular mental stress response differed by age and sex; 3) The MIST produces an endocrine stress response in addition to the cardiac response; 4) Endocrine and cardiovascular responses correlated.

Recent meta-analysis has shown that the weight of evidence supports an association between greater mental stress responses and poor cardiovascular outcomes (14) but many studies have failed to demonstrate this, leading to controversy. Such studies have often relied on limited cardiovascular parameters (BP and HR) and unreliable mental stress paradigms. This may explain their inability to demonstrate associations between mental stress and cardiovascular risk. To improve the quality of future studies, an approach like ours, which can assess the cardiovascular response to mental stress accurately and comprehensively, is required. Such assessment should include measures of ventricular function, CO, SVR, and TAC. These measures have previously been attempted using techniques such as echocardiography (with Doppler), pulse waveform analysis, and thoracic impedance cardiography. However, none of these can accurately assess both the cardiac and vascular response to mental stress with the same level of accuracy as MRI. A better approach is real-time cardiac MR, a proven method of assessing the cardiovascular system. The benefits of real-time over gated MR are that: 1) it is fast, allowing frequent measurements during stress; 2) breatholding, which alters physiology, is not required; 3) it is less susceptible to the participant motion that often occurs during mental stress. Its disadvantages are that it has reduced image quality, lower spatial resolution, and poorer temporal fidelity than gated cardiac MR. However, the two real-time sequences used in this study have been validated successfully against reference standard gated techniques at rest (9, 12). Importantly, they have also been shown to be accurate and reproducible during exercise, which produces a similar cardiovascular response to mental stress, associated with a significant increase in heart rate (9, 13). Thus, we believe that these techniques are suitable for assessment of ventricular function and aortic flow during mental stress.

This study is the first demonstration of real-time cardiac MR characterization of the mental stress response. We showed that the BP response was associated with increased CO and vascular stiffening. HR rather than SV drove the increased CO. Age and sex differences in the cardiovascular stress response were analyzed in order to demonstrate the potential utility of this approach in mechanistic studies. We showed that the greater BP responses in men were due to a greater increase in COi and higher SVRi. Similarly, correlation between age and SAP during stress was partly explained by lower TACi. Such differences were not apparent at rest and highlight the importance of a stressor for unmasking important physiological mechanisms. For this reason, we believe that MR assessment of the cardiovascular response to mental stress will be useful in future prognostic and mechanistic studies.

A vital part of our approach is the use of a suitable stress test. Although existing tasks such as serial subtraction or the Stroop color-word interference task are possible in the confines of an MR scanner, they do not stimulate the cardiovascular and endocrine systems reliably (11). The MIST was specifically designed for an MR environment and contains elements designed to enhance motivation, social evaluative threat, and uncontrollability. We have shown that the MIST produces a marked cardiovascular response with a significant endocrine response in most participants. This is important because the strongest epidemiological evidence linking life stress with cardiovascular disease comes from studying life stresses that have elevated cortisol as a typical feature (15–17). Thus, the MIST offers an opportunity to examine a stress response that is known to be associated with disease.

Interestingly, endocrine responsiveness was associated with increased arterial stiffening during stress. However, cortisol and TACi were unrelated at rest, implying that cortisol does not directly control arterial compliance, although it may potentiate or prolong the vascular response. These links between the endocrine and cardiovascular system warrant further examination and importantly may only be elucidated using a stress paradigm.

In conclusion, we present a novel method for measuring the cardiovascular response to stress in a comprehensive, reliable, and accurate way. Even in this small study, we have shown associations between the cardiovascular response and participant characteristics. Thus, our method shows promise as a tool to bring greater understanding of the mechanisms underlying the links between mental stress and the development of cardiovascular disease.


We thank Rod Jones, Wendy Norman, Jacob Salmon, Stuart Stagg, and Michael Hansen for valuable contributions to the study.