Image Based Calculation of Perfusion and Oxyhemoglobin Saturation in Skeletal Muscle during Submaximal Isometric Contractions

Abstract

The relative oxygen saturation of hemoglobin (%HbO₂) and the rate of perfusion (θ̇) are important physiological quantities, particularly in organs such as skeletal muscle in which oxygen delivery and use are tightly coupled. The purpose of this study was to demonstrate the image-based calculation of %HbO₂ and quantification of perfusion in skeletal muscle during isometric contractions. This was accomplished by establishing an empirical relationship between the rate of RF-reversible dephasing (R₂′) and near infrared spectroscopy (NIRS)-observed oxyhemoglobin saturation (%HbO₂) under conditions of arterial occlusion and constant blood volume. A calibration curve was generated and used to calculate %HbO₂ from R₂′ changes measured during contraction. Twelve young healthy subjects underwent 300 seconds of arterial occlusion and performed isometric contractions of the dorsiflexors at 30% of maximal contraction for 120s. Muscle perfusion was quantified during contraction by arterial spin labeling and measures of muscle T₁. Comparisons between the %HbO₂ values predicted from R₂′ and that measured by NIRS revealed no differences between methods (p = 0.760). Muscle perfusion reached a value of 34.7 mL 100g⁻¹ min⁻¹ during contraction. These measurements hold future promise in measuring muscle oxygen consumption in healthy and diseased skeletal muscle.

INTRODUCTION

The relative oxygen saturation of hemoglobin (%HbO₂) and the rate of perfusion (θ̇) are important quantities for physiological characterization of organs. This is particularly true in organs such as skeletal muscle, in which oxygen use is tightly coupled to oxygen delivery. Measured at the muscle, these measurements reflect the integrated function of the respiratory, cardiovascular, and muscular systems. It would be desirable to characterize these quantities non-invasively.

The non-invasive possibilities for measuring %HbO₂ include ¹⁵O positron emission tomography (PET) and near-infrared spectroscopy (NIRS). Compared to NIRS, MRI offers greater tissue coverage; the absence of ionizing radiation is an advantage of MRI over PET. MRI produces image contrast based on spatial and/or temporal variations in the physiology of organs and blood. It may be possible to take advantage of these properties to produce image based measurements of %HbO₂ in skeletal muscle. MRI also allows several other potential physiological measurements, including arterial spin labeling (ASL) methods to quantify perfusion without the need to inject contrast agents.

Methods of measuring %HbO₂ based on the blood-oxygen level dependent (BOLD) effect have been developed for brain and applied under conditions of hypoxia (1–7); hypo- and hypercapnia (3,8,9), hemodilution (10), and ischemia (11,12). These methods are limited to quantification relative to a control condition. An et al. (13) presented absolute quantification of %HbO₂ in human brain using a multi-echo gradient-echo and spin-echo technique to estimate the rates of permanent and RF-reversible dephasing (R₂ and R₂′, respectively). A similar approach may be effective in skeletal muscle.

A BOLD effect also occurs in skeletal muscle (14), which is most robustly observed as a T₂* change (15). During exercise, however, intracellular water (16,17) and hydrogen ion (17,18) accumulation cause the T₂ to increase. Because changes in T₂ also affect T₂*, any use of BOLD data to estimate %HbO₂ must also account for these metabolically driven T₂ increases. One approach to calculating %HbO₂, as mentioned above, involves measuring R₂′. The relationship between R₂′ and the quantity (100-%HbO₂) has been modeled in the brain as a parabolic relationship under conditions where hematocrit (Hct), capillary orientation, and relative blood volume remained constant (19). It may be possible to observe the relationship between R₂′ and %HbO₂ in skeletal muscle conditions of arterial occlusion and constant blood volume and then use the relationship to predict %HbO₂ from R₂′ changes under isometric contraction conditions. Therefore, the purpose of this study was to test the hypothesis that an MRI-based calculation of %HbO₂ during isometric contractions, based on an empirical calibration, would produce similar %HbO₂ estimates to directly observed NIRS measurements. Because fully characterizing aerobic metabolism in skeletal muscle also requires quantification of oxygen delivery, we also used an ASL method to measure perfusion during muscle contraction.

METHODS

Subjects

These studies were approved by the Vanderbilt University Institutional Review Board and were in accordance with the Declaration of Helsinki. Twelve subjects (6 male, 6 female) with no known personal history of cardiovascular, endocrine, metabolic, neurological, or neuromuscular disorders participated in the study. The subjects had a mean ± standard deviation (SD) age of 22.8 ± 2.1 years, height of 174.5 ± 14.9 cm, and mass of 71.8 ± 14.9 kg. After receiving an explanation of the potential risks and benefits of participation in the study, the subjects provided written informed consent to participate.

Experimental Protocol

Subjects reported to the laboratory for an orientation/NIRS testing session and two MRI testing sessions. During the orientation/NIRS session, informed consent was obtained, health and MRI screening were performed, maximal voluntary contraction (MVC) force was measured, the contraction protocol was rehearsed, and NIRS data were acquired during contraction and arterial occlusion conditions. The MVC force was measured as the greatest force recorded during two or more 3 s isometric dorsiflexion contractions, as previously described (20). The subject rehearsed the submaximal contraction protocol (see below). Contractions were performed for 120 s at 30 % of MVC. This contraction intensity was chosen to allow for a flow permissive condition (21). Finally, NIRS data were acquired during 60 s of rest, 120 s of submaximal contraction, 46 minutes of rest, and five minutes of arterial occlusion (see below).

The protocol for the testing sessions is shown in Figure 1. The two MRI sessions were performed in random order. Session A included three warm-up contractions at MVC followed by a 120s, 30% MVC contraction while image data were acquired to calculate T₂*. After 45 minutes of rest, the warm-ups and contraction were repeated while image data were acquired to calculate T₂. After a 40 minute rest, subjects underwent five minutes of proximal arterial occlusion while image data were acquired to calculate T₂*. Session B included three warm-up contractions at MVC followed by a 120 s, 30% MVC contraction while image data were acquired to calculate perfusion. After a 40 minute rest, subjects underwent five minutes of proximal arterial occlusion while image data were acquired to calculate T₂. The subjects were instructed not to consume caffeine or use tobacco during the 6 hours prior to test sessions, and not to use alcohol or perform moderate or heavy physical activity for at least 24 hours prior to each session.

Figure 1. Experimental Design.

The experiment consisted of 3 days of testing. Following the initial session, the two MRI sessions (Sessions A and B) were performed in random order. Text in boxes refers to the parameters measured, the intensity of the contraction, and performance of either contraction or occlusion where appropriate. Forty five minutes of rest were provided between contraction T₂* and T₂ acquisitions during Session A; Forty minutes of rest were provided between contraction and occlusion measurements on each day. See text for further detail.

Data Acquisition Procedures

Isometric Contractions

The subject lay supine with his/her foot strapped into an in-house-built isometric exercise device. For the out-of-magnet studies the exercise device was bolted to a grid plate attached to an exercise table and for the in-magnet studies the device was bolted to a similar grid plate on the patient bed of the imager. The ankle angle was 90° and the knee was supported by a bolster and flexed to ~7.5°. The foot was firmly strapped to the exercise device by 3.8-cm-wide nylon straps placed across the foot proximal to the base of the fifth digit. To eliminate hydrostatic effects on the circulation, the heart was at approximately the same level as the tibialis anterior (TA) muscle, the primary dorsiflexor.

Force was measured as described previously (20). Briefly, the isometric dorsiflexion device included a load cell, the signals from which were amplified and connected to an analog-to-digital conversion card fit into a laptop computer. Custom data-acquisition software written in LabVIEW 7.1 (National Instruments, Austin TX) was used to sample force data at 1 kHz and provide real-time, 20-Hz visual feedback using a panel of simulated LED’s. During submaximal contractions, the subjects standardized the force onset kinetics by matching their force output to a sequence of LED’s programmed to illuminate over the initial 6 s of contraction. The force ramp occurred non-linearly, such that force initially increased rapidly and then more slowly as it approached the target level.

The force data were filtered with a 10^th order low-pass Butterworth filter with a cutoff frequency of 25 Hz. The resulting data were zero phase forward and reverse digitally filtered. The force-time integral was computed during the portion of the contraction at which force exceeded 10% of the target force. To determine the variability of force produced during each measurement trial, a coefficient of variation was calculated for each subject.

Arterial Occlusion

The subject lay supine on the exercise table or patient bed with his/her foot strapped into the exercise device. A pneumatic cuff was placed around the thigh and connected to a Hokanson E20 rapid cuff inflator and AG101 air source (Hokanson, Bellevue, WA, USA). The heart, foot, and knee were in the same position used for the contraction studies. Following 10 min of rest in this position, the cuff was rapidly inflated to 248 mmHg. The arterial occlusion continued for a period of 5 minutes, after which the cuff was rapidly released.

NIRS

Tissue oxygenation data were collected using a frequency domain, multidistance NIRS oximeter (Model 96208; ISS, Inc., Champaign IL) and its accompanying software, as described previously (20). Briefly, before each testing session, the oximeter was calibrated using a block with known absorption and scattering coefficients, and the performance was verified using a second block with different optical properties. The absorption and scattering coefficients were measured during contraction or occlusion and used to calculate %HbO₂ and the total hemoglobin concentration ([THb]) according to the manufacturer’s algorithms.

MRI Data Acquisition

Data were obtained on a 3T Philips Intera Achieva MR imager/spectrometer. As the subject lay supine with the dominant foot in the exercise device, the maximum cross-sectional area of the TA muscle was identified during a light dorsiflexion contraction. A 16-cm, eight-channel sensitivity encoding (SENSE) RF knee coil was positioned such that the coil’s center was aligned with this location and the subject was advanced into the magnet. After acquiring three-plane GRE scout images to verify the location of the TA’s maximum girth, a volume of interest was specified in the portion of the leg to be imaged and used for second-order shimming. A T₁-weighted anatomical image was obtained using a fast spin-echo (FSE) sequence and slice thickness (ST)=10 mm, field of view (FOV)=18×18 cm, acquired matrix=256×128 (reconstructed matrix = 512×512), repetition time (TR)=500 ms, echo time (TE)=20 ms, echo-train length (ETL)=4, and number of excitations (N_EX)=2.

During the contraction and occlusion protocols, data to calculate T₂* were acquired with a dual gradient-echo (GE) echo-planar imaging (EPI) sequence. These images were obtained using the same geometric parameters as the anatomical image, acquired matrix=64×64 (reconstructed matrix=128×128), fat suppression using a spectral selection attenuated inversion recovery (SPAIR) pulse applied 202ms before the excitation pulse and frequency bandwidth of 250Hz, N_EX=1, 50% k-space sampling, and TR/TE=2500 ms/6, 46 ms. Data to calculate T₂ were obtained using a spin-echo (SE) EPI sequence for 60 s before, during, and 600 s after the contraction, using the same geometric parameters as for the anatomical image, acquired matrix= 64×64 (reconstructed matrix=128×128), SPAIR fat suppression, N_EX=1, 100% k-space sampling, and TR/TE=2500 ms/42.5, 85 ms). For both contractions and occlusion, MRI data were acquired during 60 s of rest, 300 s of occlusion or contraction, and 600 s of recovery.

To calculate perfusion during contraction, tissue T₁ was measured and ASL data were acquired. To measure T₁, a 3D GE multi-flip angle imaging sequence using ST=18 mm, FOV= 18×18 cm, acquired matrix=192×192 (reconstructed matrix=256×256), TR/TE=8.345/4.6 ms, and N_EX=1. Data were acquired in 10 dynamic scans with linearly decreasing flip angles from 18 to 0°. Next, a flow-sensitive alternating inversion recovery (FAIR) sequence employing a hyperbolic secant inversion pulse was used to acquire data during a 30% MVC contraction, using ST=10 mm, FOV=18×18 cm, acquired matrix=64×64 (reconstructed matrix=64×64), TR/TE= 5000/23.29 ms, inversion time (TI)=1000 ms, SPAIR fat suppression, and N_EX=1. The TI was set on the basis of pilot studies showing that this TI resulted in a maximum signal difference between the selective and non-selective inversion recovery images. These data were collected over 78 dynamic scans corresponding to 60 s rest, 120 s contraction, and 600 s recovery.

Data Analysis Procedures

General

Image analysis was conducted using in house-written routines for Matlab v. 7.5.0.342 (The Mathworks, Natick, MA, USA). Functional images were registered to the anatomical using a rigid registration algorithm based on the optimization of a mutual information metric (22). A region of interest (ROI) was specified along the border of the TA muscle in the anatomical image, carefully excluding any resolved vessels and resized to match the matrices of the functional images; placements within functional images were inspected to ensure exclusion of any flow artifacts arising from large vessels.

Relaxation rates

For each subject, the mean signal (S) from the ROI was calculated for each echo of the GE and SE images. Leg motion during contraction abruptly altered the image signal intensity at the beginning and end of contraction in the GE images; but in the SE images, this abrupt signal change only occurred at the start of contraction. While such a signal change at the start of exercise is expected (due to unsaturated spins entering the slice place as the muscle shortens (23), the negative signal shift at the end of exercise is not. This suggests a true baseline shift that occurred in the GE images but not the SE images. Because the signal decay model we used could not have included a baseline term (signals at only two echo times were sampled), a change in baseline signal would affect the R₂* calculations. To avoid this problem, a correction of GE image signal time courses, based on the assumption that the end-exercise and initial post exercise signals should be continuous, was applied. To do so, we calculated the difference between these signals and added it to all contraction values. Figure 2 shows an example of the artifact and its correction. The perfusion time courses also had large motion artifacts at the end of contraction only and were corrected with linear interpolation of the adjacent time points. Relaxation time constants were calculated from multi-echo GE and SE images, respectively, using the following equations:

$$T_2^{\ast }=\frac{\mathrm{\Delta }TE}{ln(\raisebox{1ex}{$S_6$}\!\left/ \!\raisebox{-1ex}{$S_{46}$}\right.)}$$ [1]

$$T_2=\frac{\mathrm{\Delta }TE}{ln(\raisebox{1ex}{$S_{42.5}$}\!\left/ \!\raisebox{-1ex}{$S_{85}$}\right.)}$$ [2]

where the subscripts refer to the TEs. Relaxation rate constants (R₂* and R₂) were calculated as the inverse of their respective time constants. R₂′ was computed as (R₂* – R₂).

Figure 2. Artifact Correction.

Initial data analysis revealed that motion associated with the start and end of contraction altered the baseline of the GE-measured signals. Above is a representative example of this artifact (A) and its correction (B). The artifact only occurred in gradient echo signals; a spin echo signal is also presented for reference (C). Correction consisted of calculating the difference between post (t=122.5) and pre (t=117.5) artifact signals, then adding this difference to all contraction values (t=0–120). This correction was applied to all subjects.

Perfusion Calculation

The multi-flip angle data were used to calculate tissue T₁ maps as described in (24). R_1tissue was calculated from T₁ maps. The R₁ of arterial blood was calculated from the data of Lu et al. (25), assuming arterial hematocrit = 0.46 for males and 0.42 for females (26). Perfusion was calculated according to the method proposed by Kim et al. (27):

$$\stackrel{.}{Q}=\frac{\lambda }{-2\alpha }\times \frac{\mathrm{\Delta }S}{S_{\mathit{sel}}}\times \frac{R_{1\mathit{blood}}-R_{1\mathit{tissue}}}{e^{-\raisebox{1ex}{$TI$}\!\left/ \!\raisebox{-1ex}{$T_{1\mathit{tissue}}$}\right.}-e^{-\raisebox{1ex}{$TI$}\!\left/ \!\raisebox{-1ex}{$T_{1\mathit{blood}}$}\right.}}$$ [3]

where λ = the ratio of the blood water fraction to the tissue water fraction, α = cos(π) (assuming perfect inversion), ΔS is the difference in non-selective and selective signals, and S_sel is the signal from the selective inversion image.

%HbO₂ Calibration

For contraction and arterial occlusion conditions, R₂′ values were expressed as ΔR₂′ by subtracting the mean pre-contraction rate from the rates of the full mean time course. ΔR₂′ was used (rather than R₂′) in order to account for inter-individual and inter-day variations in the goodness of shimming, which would add additional variability to the measurement and complicate the formation of a whole-group calibration curve. The relationship between %HbO₂ and ΔR₂′ was established using data from the portion of the mean arterial occlusion time course over which [THb] remained effectively constant (t=5 to t=295; mean 80.2 (0.3), CV=1.3%). From the initial 12 subjects, two subjects were excluded in whom an increase in blood volume was observed during cuff occlusion, as this trend is opposite to that observed during contraction (vide infra). Data from the remaining ten subjects were averaged and plotted with 100-%HbO₂ as a function of ΔR₂′. The data were fitted to a second order polynomial using MATLAB’s built-in Nelder-Meade algorithm. %HbO₂ was calculated during contraction from the ΔR₂′ data obtained during contraction and the above described calibration.

Statistics

Data are expressed as mean ± standard error (SE). Statistical analysis was performed in SPSS 15.0. Force-time integrals for each measurement trial, %HbO₂, [THb], R₂*, R₂, and R₂′ from pre-contraction or occlusion (t=0) and at selected time points during either contraction (t=10, 30, 60, 90, 120) or occlusion (t= 60, 120, 180, 240, 300) were compared using a one-way ANOVA. Significant main effects were followed by pairwise comparisons between specific time points using with Bonferroni correction for multiple comparisons. Because the validity of applying the %HbO₂-ΔR₂′ calibration obtained during arterial occlusion to contraction depends on the similarity of [THb] responses to the two protocols, [THb] from contraction and arterial occlusion were compared with a two-way ANOVA (Time, Procedure). The agreement between the %HbO₂ measured by NIRS and that predicted from ΔR₂′ was assessed using a two-way ANOVA (Time, Method). The time factor had levels of t = 10, 30, 60, 90, 120, and the method factor encompassed the NIRS and MRI methods. Significance for all tests was set at α = 0.05.

RESULTS

Arterial Occlusion Condition: NIRS and MRI Observations

NIRS Data

Mean [THb] and %HbO₂ data for the arterial occlusion condition are presented in Figure 3. The pre-occlusion mean [THb] was 82.9 ± 8.4 μM. [THb] was 81.4 ± 9.3 after 60 s, 81.0 ± 9.4 after 120 s, 79.9 ± 9.0 after 180 s, 79.4 ± 9.1 after 240 s and 85.0 ± 10.8 after 300 s. There was no main effect of time (p = 0.23, N = 10). The pre-occlusion mean %HbO₂ was 69.1 ± 1.1 %. Occlusion decreased %HbO₂ relative to pre occlusion by 11.0% after 60s, 21.0% after 120s, 29.9% after 180s, and 33.0% after 240s. There was a main effect for time (p < 0.001, N = 10); comparisons between time points revealed that all occlusion time points were significantly less than the pre-occlusion value (p < 0.01, N=10).

Figure 3. NIRS Data from Arterial Occlusion Condition.

Mean (SE) [THb] (gray line) and %HbO₂ (black line) before and during 300 s of arterial occlusion. Statistics performed using pre-occlusion mean and means from t = 60, 120, 180, 240, 300 time points. *represents a difference between pre-occlusion and the indicated time point; # represents differences between the time points indicated by the brackets.

MRI Relaxation Data

Mean R₂*, R₂, and R₂′ data for the occlusion protocol are presented in Figure 4 and Table 1. The pre-occlusion mean R₂* was 47.3 ± 2.3 s⁻¹; R₂* increased during occlusion (time main effect p<0.016, N=10), but there were no significant differences between individual time points. The pre-occlusion R₂ was 40.9 ± 1.5 s⁻¹; R₂ did not change during occlusion (p=0.718, N=10). The pre-occlusion R₂′ was 6.7 ± 2.4 s⁻¹; R₂′ increased during arterial occlusion (time main effect p=0.032, N = 10), but there were no significant differences between individual time points.

Figure 4. Relaxometry Data from Arterial Occlusion Condition.

Mean (SE) R₂* (dotted line) R₂ (gray line), and R₂′ (solid line) before and during 5 min of arterial occlusion. Statistics performed using pre-occlusion mean and means from t = 60, 120, 180, 240, 300 time points. There were main effects of time, but no differences between time points.

Table 1. Relaxation Rate Parameter Data.

Mean (SE) relaxation rate measurements for selected time points from contraction and occlusion conditions. For a presentation of statistical differences between measurements, see Figure 4 for occlusion, and Figure 7 for contraction.

Parameter	Pre-occlusion	10s	30s	60s	90s
R2*	47.3 (2.3)	51.1 (2.5)	51.7 (2.5)	51.7 (2.4)	51.9 (2.4)
R2	40.9 (1.5)	40.9 (1.3)	41.2 (1.3)	41.3 (1.4)	41.2 (1.4)
R2′	6.7 (2.4)	10.2 (1.9)	10.5 (1.9)	10.5 (1.7)	10.7 (1.8)

Parameter	Pre-contraction	60s	120s	180s	240s
R2*	45.8 (1.8)	48.1 (1.9)	49.6 (1.7)	48.2 (1.5)	47.0 (1.3)
R2	40.3 (0.8)	40.2 (0.9)	40.2 (1.0)	39.4 (1.0)	38.2 (0.8)
R2′	5.4 (1.4)	7.9 (2.0)	9.4 (1.8)	8.8 (1.5)	8.8 (1.4)

The %HbO₂, ΔR₂′ relation

The relationship between %HbO₂ and ΔR₂′ is presented in Figure 5, with the open circles reflecting data acquired during occlusion and the filled circles reflecting data acquiring following occlusion. The solution for the best fit line (with 95% confidence intervals of the parameter estimates in brackets) was:

$$100-\%{\mathit{HbO}}_2=1.22[0.32,2.12]{(\mathrm{\Delta }{R}_2^{\prime })}^2+8.66[6.05,11.27](\mathrm{\Delta }{R}_2^{\prime })+28.22[26.46,29.98]$$ [6]

Figure 5. Fitting ΔR₂′, HbO₂ Relation.

Mean HbO₂ and ΔR₂′ data and taken from arterial occlusion t=5 to t=295. Fit to a second order polynomial. Open circles represent data from occlusion and closed circles represent data from post occlusion, demonstrating the dependence of the relationship on changes in blood volume. Diamonds represent residuals for the polynomial fit of occlusion data only. The equation for the fit line with 95% confidence intervals for the coefficients in brackets was:

$$100-\%{\mathit{HbO}}_2=1.22[0.32,2.12]{(\mathrm{\Delta }{R}_2^{\prime })}^2+8.66[6.05,11.27](\mathrm{\Delta }{R}_2^{\prime })+28.22[26.46,29.98]100-\%{\mathit{HbO}}_2$$

R² = 0.93.

The coefficient of determination was R² = 0.93.

Contraction Condition: Force, NIRS, and MRI Observations

Submaximal Force Production and Muscle Cross-Sectional Area

The mean MVC force was 303.0 ± 22.7 N. Subjects reproduced 30% of MVC target force both within and between different measurement trials. The force time integrals for the NIRS, FAIR, T₂ and T₂* measurement trials were 17055 ± 1274, 17141 ± 1275, 16939 ± 1241, and 17075 ± 1241 Ns. There were no differences between measurement trials. The average coefficient of variation over contraction trials for the force time integral was 1.4 ± 0.32 %.

NIRS Data

Mean [THb] and %HbO₂ data for the contraction protocol are presented in Figure 6. There was a main effect of time (p < 0.001, N = 10; Fig 5) such that the pre-contraction mean [THb] (72.4 ± 8.2 μM) decreased by 7.0% after 10s, but only by 6.4% after 30s, 5.5% by 60s, 4.9% by 90s of contraction. Time point differences were only detected between 120s and other points, excluding the pre contraction value. Contraction decreased %HbO₂ relative to pre contraction by 7.8 % after 10s, 28.8% after 30s, 32.3% after 60s, 32.3% after 90s, and 23.8% after 120s of contraction. There was a main effect for time (p < 0.001, N = 10; Fig 5).

Figure 6. NIRS Data from Contraction Condition.

Mean (SE) total hemoglobin (gray line) and hemoglobin saturation (solid line) before and during 2 min of contraction at 30% of MVC. Statistics performed using pre-contraction mean and means from t = 10, 30, 60, 90, 120 time points. *represents a difference between pre-occlusion and the indicated time point; # represents differences between the time points indicated by the brackets.

MRI Relaxation Data

Mean R₂*, R₂, and R₂′ data for the contraction protocol are presented in Figure 7 and Table 1. The pre-contraction mean R₂* was 45.8 ± 1.2 s⁻¹; R₂* increased during contraction (time main effect of p < 0.001, N=11). The pre-contraction mean R₂ was 40.3 ± 0.8 s⁻¹ and did not change significantly during contraction (time main effect p = 0.18, N=11). The pre-contraction R₂′ was 5.4 ± 1.4 s⁻¹; R₂′ increased during contraction (time main effect p < 0.001, N=11).

Figure 7. Relaxometry Data from Contraction Condition.

Mean (SE) R₂* (dotted line) R₂ (gray line), and R₂′ (solid line) before and during 2 min of contraction at 30% of MVC. Statistics performed using pre-contraction mean and means from t = 10, 30, 60, 90, 120 time points. *represents a difference between pre-occlusion and the indicated time point; # represents differences between the time points indicated by the brackets.

Calibration of %HbO₂

Comparison of the [THb] during the first 120 s of contraction and occlusion using a two-way ANOVA revealed no main effect of the procedure (contraction or occlusion; p = 0.080; N = 10), and no main effect of time (p = 0.952; N = 10). The absence of significant interaction indicated that [THb] during both procedures behaved similarly over time (p = 0.083; N = 10). Application of Equation 6 to the ΔR₂′ data from contraction resulted in the %HbO₂ time course shown in Figure 8. Also shown in Figure 8 is the NIRS-observed %HbO₂ time course from contraction; good agreement exists between the two methods. There was no main effect for method (p = 0.760; N=10), but there was a main effect for time (p < 0.008; N=10). The absence of significant interaction indicated that %HbO₂ from both methods behaved similarly over time (p = 0.856; N=10).

Figure 8. Calibrated HbO₂ for Contraction Condition.

Mean (SE) %HbO₂ measured (gray line) and predicted (solid line) during contraction. There is good agreement between measured and predicted %HbO₂. # represents differences between the time points indicated by the brackets; no differences existed between the methods.

Muscle Perfusion

Mean perfusion data during the contraction protocol are presented in Figure 9. The pre-contraction perfusion was 23.3 ± 2.8 mL 100g⁻¹ min⁻¹. Contraction increased perfusion relative to pre contraction by 18.2% after 10s, 15.4% after 30s, 23.4% after 60s and 36.7% after 90s of contraction. There was a main effect for time (p = 0.001, N=12).

Figure 9. Perfusion Data.

Mean (SE) perfusion before and during 2 min of contraction at 30% of MVC. Statistics performed using pre-contraction mean and means from t = 10, 30, 60, 90, 110 time points. * represents a difference between pre-occlusion and the indicated time point.

DISCUSSION

This study demonstrates the parabolic relationship between the %HbO₂ and ΔR₂′ in skeletal muscle during arterial occlusion, measured using NIRS, gradient-, and spin-echo MRI. This empirically obtained calibration was used to calculate %HbO₂ from ΔR₂′ during isometric contraction. Similar MRI methods have been employed in brain (13,28), but this is the first study to demonstrate this type of technique in skeletal muscle. In addition, the method accounts for changes in the T₂ of the tissue and provides a method for calibrating the relationship between %HbO₂ and tissue relaxation rates without using breathing of non-physiologic gas mixtures. The development of this technique first depended on the correction of motion artifacts in images and second on establishing agreement between measured and predicted %HbO₂. We also interpret the ASL data from contracting skeletal muscle.

Correction of Motion Artifacts

With any magnetic resonance study in vivo, and particularly with studies of muscle contraction, subject motion presents unique difficulties. Studies of the MR signal time course during isometric contraction (23,29,30) suggest a specific pattern of response characterized by a short lived increase when contraction initiates, an early decrease, and longer duration secondary increase. In addition, the observed signals include noise and any potential artifacts. Initial analysis of the MR data from contraction in the current study showed that the signal in the GE sequence increased abruptly with the initiation of contraction, and decreased abruptly at the end of contraction, in manners inconsistent with the expected signal pattern and consistent with a baseline shift (see Figure 2a). These changes in signal in the GE images were greater than those observed in the SE images, and the SE images did not exhibit signal shifts at the end of contraction. (Figure 2c). Examination of the images revealed that bulk limb motion in the anterior direction occurred at the initiation of contraction, position was maintained during contraction, and posterior motion occurred with relaxation. However, following registration correction, this unexpected signal intensity pattern persisted, even after verifying that the position of the ROI within the in-plane representation of the muscle had not changed. This pattern of bulk motion in both the GE and SE images, the constancy of the ROI within the muscle following registration correction, and the appearance of the artifact in the GE images only, may suggest that the observed changes in signal baseline resulted from an alteration of the quality of shimming. It is clear from the comparison of Figure 2b and c that the correction procedure was effective in restoring the expected shape of the GE signal time course, having the same characteristics as the SE signal time course that was unaffected by the motion artifact. Thus, we are confident in the further application of the signal time courses in the calculation of relaxation rates.

Agreement Between Measured and Calibrated %HbO₂

There is good agreement between the measured and calibrated %HbO₂ shown in Figure 8. The agreement arises from three general conditions. The first is that both NIRS and BOLD signals arise through interactions with the hemoglobin molecule. The second is that arterial occlusion provides an excellent model to observe the relationship between R₂′ and %HbO₂. The third is that contraction conditions approximate the conditions of arterial occlusion. We will now discuss these three conditions in greater detail.

Hemoglobin contains an iron moiety, the chemical and physical environments of which change with oxygenation state. NIRS measurements depend on the change in NIR peak absorbance wavelength between oxygenated and deoxygenated hemoglobin. BOLD measurements depend on the different effects on the local magnetic field between diamagnetic hemoglobin and paramagnetic deoxyhemoglobin. In addition, both methods detect the bulk of their signal from the venous circulation where the majority of the blood volume resides. In both cases, the signals arise primarily from the small vessels, which are unresolved and therefore included in ROI-based signal measurements in MRI and are small enough that light is incompletely absorbed (i.e., some light is scattered back to the detector) in NIRS.

The effects of paramagnetic deoxyhemoglobin on the local magnetic field lead to an increase in the relaxation rate of magnetization measured in GE images. These effects depend on Hct, B₀, capillary orientation, saturation, and relative blood volume (19,31). Data from our lab (32,33) as well as observations from other investigators (34) suggest that extravascular contributions to skeletal muscle BOLD effects at 3T are practically insignificant, removing the need to consider the effects of capillary orientation. [THb], and presumably the Hct and relative blood volume, remain essentially constant when tissue blood flow is occluded (see Fig. 2). Because B₀ remains unchanged during the experiment, the change in %HbO₂ remains as the sole contributor to the relationship with ΔR₂′. We note that the influence of myoglobin on the relationship is minimized by its relatively smaller concentration and its limited dynamic range of saturation with typical tissue PO₂ values (14,32). With regard to the form of the equation used to describe the relationship between ΔR₂′ and %HbO₂, we note that theoretical models of the R₂′-%HbO₂ relationship do not include the linear term in Eq. 6 (35). However, inclusion of this term does result in a more accurate empirical calibration of %HbO₂ values, which was our interest in the present study.

The validity of the calibration of %HbO₂ during contraction from a relationship observed during occlusion depends on replication of the conditions described above during submaximal contraction. Hct remains unaltered during isometric contraction within the venous circulation where most of the blood signal arises. The effect of changes in blood volume on the ΔR₂′-%HbO₂ relationship can be observed by comparison of open (arterial occlusion) and closed (occlusion release) data points in Figure 5. Therefore, to assure that the calibration obtained during occlusion would be applicable to contraction, we directly compared the [THb] during contraction to the first 120 s of occlusion. After expressing [THb] during contraction and occlusion procedures relative to their pre-procedure values, we found no differences between [THb] responses to these procedures.

Taking all of these considerations together, we conclude that the changes in calibration between ΔR₂′ during arterial occlusion reflect the effects of %HbO₂ changes only and that the resulting calibration between ΔR₂′ and %HbO₂ can be validly applied to the condition of isometric contraction. The good agreement between the two methods of %HbO₂ measurement follows directly from this validity.

Interpretation of Arterial Spin Labeling

The current study is one of the few to examine perfusion during contraction in humans using ASL (36–38), and the first to our knowledge to provide measurements in the TA during submaximal isometric contraction. The FAIR technique employed in the current study does not require additional coils or contrast agents to allow perfusion quantification. Despite these advantages, the technique is signal-to-noise ratio (SNR) limited and measurements should be interpreted carefully, especially in studies of resting muscle. The perfusion values typically reported for resting skeletal muscle range from 10–15 mL 100g⁻¹ min⁻¹. The values that we observed were greater than this (25 mL 100g⁻¹ min⁻¹) and may reflect prolonged effects of the warm-up contractions on muscle blood flow.

Muscle perfusion can increase to 246 mL 100g⁻¹ min⁻¹ under maximal dynamic exercise conditions (39), demonstrating a large dynamic range and sufficient SNR under maximal flow conditions. We did not observe large magnitudes of perfusion during contraction in the present study. This observation may be explained by the application of the Law of LaPlace to isometric contraction: when curved structures are placed under stress, a pressure gradient will develop across the structure. Because contracting muscle fibers develop curvature (40), the interstitial fluid pressure will increase during contraction and be directly proportional to muscle fiber tension. Measurements of intramuscular pressure during isometric contraction presented by Sejersted et al. (41) indicate that tissue pressures exceeding systolic blood pressure are attainable during maximal isometric contraction MVC’s of large muscles (41). While in the present study of 30% MVC contractions, the tissue pressure was apparently not great enough to have impeded arterial inflow completely (Q increased significantly), it may have been at least partially impeded in some parts of the muscle. This conclusion is supported by the moderate hyperemia observed following the contraction (see Figure 9).

The effective time resolution for the FAIR sequence is twice the TR of the individual acquisition (2×5s in the current study). This time resolution was used in the present study to ensure suitable SNR. In future studies, we will explore the limits of time resolution to further examine perfusion kinetics at the onset of contraction.

Future Applications to Muscle Oxygen Consumption

It has not escaped our notice that these image-based measures of %HbO₂ and perfusion can be combined through the Fick principle to calculate muscle oxygen consumption; making reasonable assumptions concerning the arterial oxygen content and scaling this by %HbO₂ to calculate venous oxygen content, a value of 28 mL kg⁻¹ min⁻¹ at end-exercise is suggested by the data in Figures 8 and 9. However, such measurements would require careful validation from invasive measures of gas exchange in conduit arteries supplying the muscles, or rigorous and complete modeling of the relationships between ATP synthesis and oxygen consumption for submaximal exercise in a single muscle for valid comparison to rates of ATP synthesis measured by ³¹P MRS. Invasive measurements fall outside the scope of this study and are not consistent with its purpose. In addition, the complete metabolic model of the system relating ATP synthesis rates to oxygen consumption during submaximal contraction in a specific muscle remains under investigation.

Summary and Conclusions

This study demonstrates the accurate calculation of %HbO₂ during contraction based on the empirical relationship between the %HbO₂ and ΔR₂′ established during arterial occlusion. We also demonstrate the implementation of FAIR in skeletal muscle and provide quantitative estimates of perfusion during contraction. To our knowledge, this represents the first use of MRI to calculate %HbO₂ in exercising skeletal muscle and one of only several implementations of ASL in skeletal muscle. These methods hold promise for the robust calculation of muscle oxygen consumption and quantification of the kinetics of perfusion and oxygen consumption during prolonged exercise and following single muscle contractions, in highly spatially and temporally resolved manners.