1.

Introduction

Diffuse correlation spectroscopy (DCS) is a non-invasive, optical technique that has been used to relate rapid fluctuations in light intensity of a single/few detected speckles to the motion of red blood cells (RBCs) in the underlying tissue.^1,2 The mathematical units for DCS estimates of tissue perfusion (typically called a “blood flow index” or BFI) are ${\mathrm{cm}}^2·{\mathrm{s}}^{-1}$, although it is more common to report changes in tissue blood flow as a function of time, i.e., %- or fold-change from a resting baseline, by normalizing BFI to a resting baseline period.^3,4 Importantly, while changes in DCS BFI are thought to be driven by changes in RBC velocity, in silico studies by Boas et al. demonstrated that changes in hematocrit or blood vessel diameter can also affect the relative change in DCS BFI,⁵ which Sathialingam et al. recently corroborated in vitro using a two-dimensional (2D), single-plane phantom model.⁶ Moreover, Boas et al. also found that DCS BFI can be affected by shear-diffusion, which describes differences in RBC velocity at different points within a microcapillary (i.e., faster RBC velocity in the lumen isocenter and slower velocities near the capillary walls).

As such, changes in DCS BFI during physiological perturbation will match changes in tissue perfusion ($\mathrm{mL}/100\mathrm{g}$ tissue/min) only if there are no changes in: (1) blood vessel diameter, (2) shear-diffusion dynamics, and (3) hematocrit. However, the recent 2D phantom work by Sathialingam et al. showed that photons can indeed pass through blood vessel diameters as large as 85 to $100\mu \mathrm{m}$, a size that would certainly be susceptible to vasodilation in human muscle in vivo.⁶ This suggests that the assumptions of constant blood vessel diameter, shear-diffusion dynamics, and microvessel hematocrit may not be physiologically valid in vivo. Indeed, the relative change in DCS BFI in muscle is consistently lower than the relative change in either Doppler-derived conduit artery blood flow³ or microvascular perfusion measured by arterial spin labeling (ASL) MRI.⁷

Here, we report several recent observations that highlight an uncoupling between DCS BFI from changes in skeletal muscle perfusion in vivo. We hypothesize that DCS BFI in skeletal muscle is confounded by changes in vasodilation/constriction that alter the microvascular flow area (${\mathrm{MV}}_{\mathrm{A}}$), which would allow net tissue perfusion to increase or decrease independent of changes in RBC velocity. Subsequently, we provide preliminary evidence for a mathematical adjustment that appears to overcome these confounding limitations in vivo, by adjusting BFI for changes in frequency domain near-infrared spectroscopy (NIRS)-derived total-heme content (HbTot) normalized to the magnitude of the NIRS-derived Hb-nadir.

2.

Methods

All study procedures were approved by the University of Texas at Arlington and University of Texas Southwestern Medical Center Institutional Review Boards. All subjects gave written informed consent to participate prior to being enrolled.

3.

Results

As shown in Fig. 1, we observed only minor increases in quadriceps BFI above baseline following high-intensity knee extension (top) and cycling (bottom) exercises, despite marked decreases in muscle oxygenation. This finding is inconsistent with numerous investigations measuring femoral artery blood flow during knee extension¹³ or cycling exercises¹⁴ and suggests that increases in quadriceps perfusion are primarily driven by increases in vasodilation, thereby uncoupling the relationship between muscle perfusion and DCS BFI.

Fig. 1

DCS BFI and muscle oxygenation during maximal-effort knee-extension exercise (top) and whole-body cycling exercise (bottom). (a), (c) Changes in DCS BFI (fold-change above baseline; FC) from a representative subject recorded throughout 2 min of resting baseline, high-intensity exercise, and 10 min of passive recovery. (b), (e) Group changes in DCS BFI (fold-changes for both) during exercise and throughout recovery. (c), (f) The decreases in muscle oxygenation that were observed during exercise. Note that in both cases, DCS showed little-no change above baseline despite substantial decreases in muscle oxygenation, indicating that the muscle mass beneath the DCS probe was indeed active. Group data are presented as mean ± SD.

Accordingly, we investigated whether adjusting BFI for changes in microvascular flow area [Eqs. (2)–(4)] would reduce this discrepancy by leveraging an existing dataset that simultaneously measured changes in forearm BFI, NIRS-derived HbTot, and BABF during handgrip exercise.⁸ As shown in Fig. 2, the fold-change in BABF with exercise was markedly higher than the fold-change in the unadjusted BFI signal. However, using the increase in HbTot to adjust BFI for estimated changes in microvascular flow area (i.e., $\mathrm{\Delta }{\mathrm{MV}}_{\mathrm{A}}$) reduced this discrepancy, such that the $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\mathrm{LIN}}$ and $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\pi r2}$ were not different from BABF at the end of handgrip exercise ($p=0.073$ and $p=0.743$, respectfully). Moreover, the $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\pi r2}$ values were almost superimposed upon the BABF measures temporally and in magnitude. The line of identity between end-exercise BABF and $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\pi r2}$ was also close to 1.0, where as the unadjusted BFI and $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\mathrm{LIN}}$ calculations were much lower [Figs. 2(b)–2(d)].

Fig. 2

Adjusting DCS BFI for changes in microvascular area (${\mathrm{MV}}_{\mathrm{A}}$). During handgrip exercise, the fold-change in BFI was either: (1) left unadjusted, (2) linearly adjusted for estimated changes in ${\mathrm{MV}}_A$ (${\mathrm{Adj}}_{\mathrm{LIN}}$), or (3) adjusted for changes in ${\mathrm{MV}}_{\mathrm{A}}$ squared (${\mathrm{Adj}}_{\pi r2}$). Whereas the unadjusted (b) calculations underestimated the fold-change in brachial artery blood flow [(a), (b) $p=0.002$; BABF] at the end of handgripping, the ${\mathrm{Adj}}_{\mathrm{LIN}}$ [(c) $p=0.073$] and ${\mathrm{Adj}}_{\pi r2}$ [(d) $p=0.743$] calculations were not different from BABF. The line of identity was best, however, when using the ${\mathrm{Adj}}_{\pi r2}$ calculation. $n=12$ for all panels.

As a final consideration, we compared DCS BFI calculations against ASL MRI (Fig. 3). Although changes in BFI were temporally aligned with changes in ASL during reactive hyperemia [Fig. 3(a)], the peak fold-change from the unadjusted BFI calculations were ∼½ the peak fold-change in ASL, which is consistent with data presented in Yu et al.⁷ In contrast, the peak fold-change from the $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\pi r2}$ calculation was very close to that of ASL. Similar observations were made following plantar flexion exercise, with the unadjusted BFI calculation underestimating ASL measures of muscle perfusion, but the $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\mathrm{LIN}}$ and $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\pi r2}$ calculations of muscle perfusion more closely matching those from ASL both temporally, and in magnitude [Fig. 3(b)].

Fig. 3

NIRS-DCS BFI versus ASL-MRI. We compared DCS BFI calculations of muscle perfusion against arterial spin-labeling MRI during (a) reactive hyperemia and (b) plantar flexion exercise. The unadjusted BFI calculation consistently underestimated ASL-measures of muscle perfusion, whereas the ${\mathrm{Adj}}_{\mathrm{LIN}}$ and $\mathrm{BFI}\text{-}{\mathrm{Adj}}_{\pi r2}$ calculations were much closer. (c), (d) Mean residuals (BFI – ASL) throughout the recovery periods.

Overall, the results presented here support in silico work by Boas et al.⁵ and in situ results from Sathialingam et al.,⁶ which demonstrated that changes in DCS BFI are susceptible to changes in ${\mathrm{MV}}_A$. Similarly, given that changes in ${\mathrm{MV}}_A$ will directly affect the shear-diffusion dynamics of RBCs, it is likely that changes in shear-diffusion also affect changes in DCS BFI within skeletal muscle. However, additional work is needed to further validate the BFI-adjustment calculations presented in this Letter [Eqs. (2)–(4)]. For example, as shown in Figs. 3(c) and 3(d), none of the DCS BFI calculation methods matched ASL-MRI throughout the recovery period. Although the $\mathrm{Adj}\text{-}\pi r^2$ calculation best matched ASL-MRI following plantar flexion exercise [Fig. 3(d)] and the peak of reactive hyperemia [Fig. 3(c)], it was the least-accurate method over the final $\sim 3\min$ of reactive hyperemia. Moreover, ASL-MRI measures of muscle perfusion recovered much faster than all 3 DCS BFI methods during reactive hyperemia.

The mechanism(s) underlying these discrepancies are not readily apparent but could be due to partial-volume effects. Notably, ASL-MRI measures of blood flow emanated exclusively from the medial gastrocnemius, whereas DCS BFI measures of “muscle” blood flow may have been affected by the overlying skin and adipose tissues that photons must first travel through before reaching the underlying muscle. Our recent work suggests that skin blood flow does not affect DCS BFI under thermoneutral conditions, but we cannot exclude the possibility that local heating may have stimulated increases in cutaneous perfusion that affected DCS BFI. Similarly, ASL-MRI measures of muscle perfusion were measured via the axial plane, whereas the DCS laser-detector orientation was longitudinal due to spatial/setup restrictions within the MRI-scanner. Future works will need to focus on better matching the tissue volumes measured by ASL-MRI and DCS BFI.

Our DCS signal intensity was above the estimated minimum threshold of $\sim \mathrm{15,000}$ photon counts/second required for our MetaOx setup and sampling frequency.¹⁵ However, because reactive hyperemia and exercise increase total heme-content, and therefore near-infrared photon absorption (${\mu }_a$), our results may have been affected by signal intensity. Additional work is needed to optimize NIRS-DCS devices and probes for use in skeletal muscle, such as utilizing longer wavelengths (e.g., 1064 nm) for DCS BFI,¹⁶ that permit long source-detector separations and maximize depth of penetration into skeletal muscle.

NIRS-based considerations may also have affected the results. Although hematocrit affects DCS BFI in silico and in vitro,^5,6 to the best of our knowledge, no one has investigated how changes in hematocrit affect DCS BFI in vivo. Accordingly, future works should aim to directly manipulate hematocrit (e.g., isovolumic hemodilution¹⁷) and examine its impact on DCS BFI. Similarly, our BFI adjustment calculations depend on calculations of the heme-nadir, which could be affected by myoglobin, as this metabolite is indistinguishable from hemoglobin within the near-infrared spectrum.¹⁸ Accordingly, more work is needed to assess how myoglobin affects calculation of the heme-nadir, as well as work that establishes standard methods for measuring $\mathrm{\Delta }{\mathrm{MV}}_A$ across individuals and populations.

4.

Conclusions

The results presented herein provide novel empirical evidence that relative changes in DCS BFI do not translate directly to relative changes in muscle perfusion, but that adjusting relative changes in BFI for estimated changes in microvessel flow area may help overcome this issue. However, additional work is needed to further validate the BFI-Adj calculations presented herein, which will likely require substantial interdisciplinary collaboration between physicists, engineers, and physiologists alike to optimize NIRS-DCS devices and probes for use in skeletal muscle.