Background
Oxygenation-sensitive (OS) Cardiovascular Magnetic Resonance (CMR) is a promising tool to assess myocardial oxygenation changes [
1]. In contrast to other imaging modalities like coronary angiography or cardiac CT, it is not an anatomy-derived surrogate parameter for myocardial perfusion and oxygen supply, as OS-CMR detects net changes in myocardial oxygenation incorporating the contributing factors like blood oxygenation, blood flow, myocardial workload and oxygen extraction but also collateralization or coronary steal [
2]. OS-CMR is based on the so-called Blood Oxygen Level-Dependent (BOLD) effect, which takes advantage of the paramagnetic deoxygenated hemoglobin as an inherent contrast. An increase in deoxyhemoglobin would result in a decrease in signal intensity in T2*-weighted sequences. Hence, a decrease in tissue oxygenation corresponds to a decrease in signal intensity. This technique was first proposed by Ogawa et al. and has since been used in functional MRI (fMRI) studies [
3], which detects the activation of brain areas in response to distinct stimuli or triggered by tasks with an increase in local blood flow [
4]. In recent cardiac studies, this technique has been used together with coronary vasodilators such as adenosine to measure the oxygenation reserve, by comparing the images under adenosine to a set of baseline images. Myocardial segments that fail to show an increase have a high probability of being subtended to a relevant coronary artery stenosis, being already maximally dilated after a fixed obstruction and thus failing to respond to adenosine [
5‐
8]. With newer sequences, the temporal and spatial resolution is sufficient to use OS-CMR to detect rapid and dynamic myocardial oxygenation changes using breathing maneuvers. Breathing maneuvers such as hyperventilation and apnea result in systemic changes in blood carbon dioxide levels, which act as a strong modifier of coronary vascular tone [
9,
10]. The sequence of hyperventilation with subsequent apnea has shown to yield changes in myocardial oxygenation at least as strong as adenosine infusion in healthy volunteers [
11]. Such a protocol would not only be free of radiation and pharmacologic contrast agents, but would also require neither potentially dangerous vasodilators nor uncomfortable carbon dioxide levels adjusted by complicated and expensive breathing circuits [
12]. OS-CMR is being utilized increasingly to investigate cardiac pathologies and offers an innovative and potentially clinically feasible diagnostic protocol together with breathing maneuvers [
1,
2,
11,
13‐
15].
However, many factors influencing the technical aspects of the underlying T2*-sensitive sequences remain confounders. The hydration status of an individual may be such a confounder [
16]. Lin et al. (1998) could show that hemodilution has an effect on baseline signal intensity in fMRI brain scans that use the same sequence principle [
17,
18]. More importantly, hemodilution has been shown to influence the relative signal response during the execution of identical tasks in fMRI scans [
19,
20].
The aim of this study was to assess the differences in oxygenation-sensitive signal intensity free of vasoactive effects in an in-vitro setup and during vasoactive breathing maneuvers in healthy subjects. We hypothesized that hemodilution results in an increase in baseline signal and leads to an attenuation in signal intensity changes after hemodilution during dynamic breathing maneuvers.
Methods
In-vitro study
Venous and fully oxygenated arterial blood samples from five pigs ventilated with 100 % oxygen were used to assess hemoglobin-dependent changes in oxygenation-sensitive (OS) signal intensity (SI) independent of vasomotor effects. The blood was obtained immediately after euthanasia of five animals used for another experiment [
21]. The study was conducted in accordance with the Guide to the Care and Use of Experimental Animals by the Canadian Council on Animal Care and approved by the local Animal Care and Use Board and followed internationally accepted guidelines. The arterial (A100) and venous (V100) blood was diluted with Lactated Ringer’s solution in 10 % steps down to 50 % of the original sample (A50 and V50, respectively), and stored in 10 ml syringes. Before the scan, the specimens were inverted, a 1 ml sample was withdrawn for a blood gas analysis (Radiometer-ABL780, Radiometer, Brønshøj, Denmark), and then the samples were placed in the MRI scanner. A heart rate of 60/min was simulated on the identical 3 T scanner with the same OS-sequence as described in the imaging protocol for the in-vivo part of this study.
In-vivo study
This study protocol was approved by the local research ethics board of the Institut de Cardiologie de Montréal (ICM #13-1447). In order to examine the relevance of hemodilution to clinical studies, 21 healthy volunteers were recruited to participate in this study by public advertisement. Inclusion criteria were age >18 years and ability to give informed consent. Participants with any history of cardiovascular, respiratory, cerebral or renal disease, pregnancy, smoking within the last 6 months or any contraindication for MRI scans were excluded. The subjects were required to refrain from the consumption of caffeinated beverages 12 h prior to the scan.
Experimental protocol
An 18-Gauge i.v. line was placed in a cubital or ante-cubital vein for the initial measurement of the baseline hematocrit and hemoglobin concentration. After acquisition of baseline images, 1 L of Ringer’s Lactate solution was rapidly infused over 5–8 min with a pressure bag (300 mmHg). Immediately after rapid i.v. hydration, a second blood sample was acquired before a second set of images was acquired (hyper-hydration) using the identical protocol as that used for baseline scans. Non-invasive blood pressure was measured before and after hemodilution, heart rate was continuously recorded during the scans and rate-pressure product was calculated by multiplying the systolic blood pressure with the heart rate. A vasoactive breathing maneuver according to Fischer et al. was employed as an activation-dependent stimulus analogue to fMRI scans at baseline and after hemodilution [
11].
Imaging protocol
Images were acquired in a clinical 3 T MRI system (Magnetom Skyra, Siemens Healthcare, Erlangen, Germany) using a 30-channel coil setup with body and spine coils. For the in-vitro study, the samples were scanned with a simulated 60/min heart rate immediately after localization, after being placed in the scanner to minimize sedimentation. For healthy volunteers, all images were acquired during a breath-hold at end-expiration. Left-ventricular (LV) function was assessed using a standard ECG-gated balanced steady-state free-precession (bSSFP) sequence in six LV-centered radial long-axis slices (slice thickness: 8 mm; TR: 3.26; TE 1.43 ms).
One mid-ventricular slice was acquired during all oxygenation-sensitive scans using a retrospective ECG-gated bSSFP prototype sequence (slice thickness 10.0 mm, TR 3.49 ms, TE 1.57 ms, flip angle 35°). After acquisition of baseline images for both the normo- and hypervolemic state during a short breath-hold, the subjects performed 60 s hyperventilation (35 breaths/min) with constant cine acquisition during a maximal consecutive breath-hold as a dynamic vasoactive maneuver at both levels. A metronome was used to ensure consistency of the respiratory rate. Volunteers were instructed to use the alarm bell to signal the break-point of the voluntary apnea and to immediately recommence breathing.
Image analysis
Image analysis was performed using certified CMR evaluation software (cvi42, Circle CVI, Calgary, Alberta, Canada). In the in-vitro samples, SI was measured after defining a contour in the long-axis (LAX) orientation of the syringes. For the assessment of left-ventricular function, parameters in the volunteers endo- and epicardial contours were traced in end-systole and end-diastole.
In OS-CMR images, manual endo- and epicardial contour tracing was completed in end-systolic frames. Mean myocardial SI was automatically generated by the software following contour tracing. All SI values were expressed as percent change (ΔSI[%]) between two images.
$$ \varDelta SI\ \left(\%\right) = \frac{SI(maneuver)-SI(baseline)}{SI(baseline)}*100 $$
The change in signal intensity between the baseline and hyper-hydration images was calculated, and the relative SI changes during the vasoactive breathing maneuvers were compared between the two hydration states. Specifically, differences between end-hyperventilation (HV), peak (SImax), as well as after 30 s apnea (SI30s) and end-breath-hold SI (SIend) changes were examined. A second reader, blinded to the hydration status, the maneuver and the subjects’ demographics, read 39 out of 220 images (18 %) to assess inter-observer reproducibility. For the in-vitro samples, the same equation was used to calculate the %change in SI from the undiluted blood samples (100 %) to the LRS-diluted samples (90, 80, 70, 60 and 50 %), respectively.
Statistical analysis
Statistical analysis was performed using Prism 6 (GraphPad Software Inc., California, USA). All variables were checked for normal distribution using the D’Agostino-Pearson omnibus normality test. Student’s t tests were performed for those values with normal distribution, otherwise Mann-Whitney or Wilcoxon tests were used. To test correlation between quantitative variables, Pearson correlation coefficients or Spearman rank test were determined. Differences in blood levels to baseline in the in-vitro data was analyzed with a one-way ANOVA (Friedman test) with a Dunn post-hoc test for multiple comparisons. Variables are presented as means ± standard deviation. Images of nine participants (39 readings) were replicated by a second independent reader, and inter‑observer reliability was assessed using a two-way intraclass correlation (ICC) test with SPSS version 23 (SPSS IBM, New York, USA). P < 0.05 was deemed significant.
Discussion
Our results demonstrate for the first time that hemodilution may increase baseline signal intensity in oxygenation-sensitive images but also attenuate the observed signal response to vasoactive breathing maneuvers. The understanding of this relationship is important for the clinical applications of OS-CMR [
2]. Patients with changing fluid status may show altered signal intensity responses, which may confound results and their interpretation. During heart failure, for example, an attenuated signal intensity response may underestimate coronary vascular reactivity or even falsely suggest inducible oxygenation deficits.
In the in-vitro data, the concentration and fraction of deoxyHb were negatively correlated to the signal intensity of the sample and positively correlated to the hemoglobin saturation. Interestingly, both the hemodilution of arterial and venous blood resulted in an increase in signal intensity of the sample. Although there was no significant difference in the slope of the regressions, the less pronounced increase observed in venous blood can be explained by the signal-attenuating effects of the deoxyHb.
During the dilution of deoxyHb, the T2* effect based on the paramagnetic properties of deoxyHb is reduced, which augments the relative impact of T2 effects, leading to a higher signal intensity. Establishing this relationship in vitro is an important step to distinguish the SI effects of the simple hemodilution of deoxyhemoglobin from the more complex SI-changing effects by the vasoactive breathing maneuver in the participants during hemodilution.
The in-vitro data corresponds to the increase in baseline signal intensity after rapid infusion of crystalloids in healthy volunteers, leading to an attenuation of the T2* effect, which explains the observed rise in signal intensity. This effect was also observed in neurologic fMRI studies by Lin et al. [
17,
18]. The group assessed the effect of mild and moderate acute hemodilution. The group concluded that intra-voxel concentration of deoxyhemoglobin was the main contributing factor causing changes in signal intensity. It was shown that the increase in signal intensity (or reduction in R2* or inverse of T2*) was proportional to the decrease in hematocrit.
The linearity of the regression lines in our results suggests that the baseline signal intensity can be normalized for a known deviation of hemoglobin concentration. Interestingly, the change in hemoglobin concentration was only related to changes in signal intensity in male subjects.
Levin et al. (2001) performed repeated photic-stimulation functional MR experiments before and after hemodilution [
19] in male subjects. Corresponding to a 6 % decrease in hematocrit, the authors observed a statistically significant reduction of the inducible response of 8 %. These results are comparable to our data. Interestingly, in our study the extent in signal intensity attenuation was only correlated to the increase of hemodilution in males and not in females, which showed a preserved response, despite a similar level of hemodilution, similar body mass index values and the same infusion volume. From the present data, we cannot be sure about the physiologic reasons for the absence of such a relationship in females. Yang et al. found that the relationship between BOLD-SI changes in the brain and hematocrit levels are modulated by sex [
20]. In their study, the inter-hemispheric connectivity within the pars occularis extending to the precentral gyrus was found to be positively correlated with hematocrit in females and negatively correlated in males. However, the study could not clarify if these affects were acute or to neural adaptations to chronic hematocrit differences.
Yet, the observed sex differences in our study may be explained by a higher total vasodilatory capacity which had not been exhausted by the intravascular hypervolemia. A larger sample size might have detected a weaker yet still existing relationship in female participants.
The increase in baseline SI and the attenuation in SI response during vasoactive breathing maneuvers were already observed at a mean drop in hemoglobin concentration of 13.7 g/L (corresponding drop in HCT 3.8 %) in our study. Such a relatively small change is easily encountered in clinical settings. In some patients, the drop in hemoglobin concentration during volume substitution after blood loss can exceed such change by far. Conversely, diuretic therapy may significantly increase in hematocrit level. Thus, our data have clinical relevance for the future use of OS-CMR in non-invasive cardiac testing.
The changes in ventricular volumes and function data are explained by the increase in preload and consecutive improvement in contractility, based on the Frank-Starling mechanism. The increase in blood pressure leads to an elevated rate-pressure product, with an increased myocardial workload and thus higher oxygen extraction. While this may suggest an explanation for the decreased signal intensity response as well, the blood pressure remained within the auto-regulation range for myocardial blood flow and therefore, a decreased myocardial oxygen supply would result in a consecutive compensatory increase in blood flow to maintain a constant tissue oxygenation. Thus, an increase of the myocardial oxygen extraction less likely explains the observed hemoglobin-dependent changes in OS-CMR.
A study of Zheng and colleagues on brain functional connectivity in hemodialysis patient with end-stage renal disease [
22] showed that a weakening of the cortical and subcortical network connectivity in these patients was related to anemia. This observation could be explained by a decreased T2* effect in anemic patients. Another study observed a blunted myocardial oxygenation response to adenosine in patients with chronic kidney disease and renal transplant patients [
13]. The authors attributed this finding to the possible presence of microvessel disease, as these patients have a higher incidence of cardiovascular events. However, the authors do not state if the MRI scans were performed before or after dialysis in these patients. The differences in fluid state before and after dialysis could have had a profound effect on the measured OS-signal response. It is possible that the SI response may have been underestimated, which leads to the conclusion of the presence of coronary artery or microvascular coronary disease in these patients. Further studies are warranted to assess the relationship between altered fluid status, OS-signal response and heart disease in order to deduct correct conclusions.
Limitations
The precise execution of breathing maneuvers may have varied between participants and between breathing maneuvers within subjects. To minimize variations, we paced the respiratory rate for the subjects with a metronome during hyperventilation and visually monitored it by a camera installed in the scanner room. If breathing appeared insufficiently deep or too slow, we immediately intervened to correct that. As breath-hold duration during maximal tolerable apnea is highly variable, we used standard time points for comparison between fluid states. The 30 s time point was reached by every participant. Further, we used the peak signal intensity change during apnea and the signal intensity change at the end of the maneuver, when participants had the urge to breathe again for comparison. However, these time points vary in the time point of acquisition during the maneuver between participants and between runs. The primary reader of the exams was not blinded to the fluid state during which the images were acquired. However, the second reader, who analyzed 18 % of the images, was blinded to the fluid state, and there was an excellent agreement between the readers. Results may be very different in a patient cohort with a more chronic systemic fluid overload. Although our data opens up the potential for normalizing the measured signal intensity response to hemoglobin changes, there needs to be further research, especially in patients with coronary pathologies present in patients with heart failure that can be scanned before and after re-compensation. Larger sample sizes may allow for a more precise determination of the relationships and thus correction factors.
Abbreviations
∆, delta (difference); ∆OS-SImax, delta (difference) in maximal oxygenation-sensitive signal intensity; A, arterial blood sample; ANOVA, analysis of variance; BH, breath-hold; BMI, Body Mass Index; BOLD, Blood Oxygen Level-Dependent; bSSFP, balanced steady-state free-precession; CI, confidence interval; cm, centimeter(s); CMR, Cardiovascular Magnetic Resonance; Conc., concentration; CT, Computed Tomography; deoxyHb, deoxygenated hemoglobin; ECG, electrocardiogram; FHHb, fraction of deoxyhemoglobin of total hemoglobin; fMRI, Functional Magnetic Resonance Imaging; g, gram(s); h, hours; Hb, hemoglobin; HCT, hematocrit; HR, heart rate; HV, hyperventilation; i.v., intra-venous; ICC, intra-class correlation; ICM, Institut de Cardiologie de Montréal; kg, kilogram(s); L, Liter; LAX, long-axis; LRS, Lactated Ringer’s solution; LV, left-ventricular; m, meter(s); ml, milliliter(s); mmHg, Millimeter Mercury; MR, Magnetic Resonance; MRI, Magnetic Resonance Imaging; OS, oxygenation-sensitive; s, seconds; SAX, short-axis; SI, signal intensity; SI30s, signal intensity at 30 s into the breath-hold; SIend, signal intensity at the end of the breath-hold; SImax, maximal signal intensity during breath-hold; SO2, oxygen saturation of hemoglobin; SvO2, venous oxygen saturation of hemoglobin; T, Tesla; TE, Echo time; TR, Repetition time; V, venous blood sample; vs., versus
Acknowledgements
We would like to thank Camilo Molina for his assistance in the preliminary research which led to the conception of this study.