Diffuse correlation spectroscopy measurements of blood flow using 1064 nm light

Abstract

Significance: Diffuse correlation spectroscopy (DCS) is an established optical modality that enables noninvasive measurements of blood flow in deep tissue by quantifying the temporal light intensity fluctuations generated by dynamic scattering of moving red blood cells. Compared with near-infrared spectroscopy, DCS is hampered by a limited signal-to-noise ratio (SNR) due to the need to use small detection apertures to preserve speckle contrast. However, DCS is a dynamic light scattering technique and does not rely on hemoglobin contrast; thus, there are significant SNR advantages to using longer wavelengths (>1000 nm) for the DCS measurement due to a variety of biophysical and regulatory factors.

Aim: We offer a quantitative assessment of the benefits and challenges of operating DCS at 1064 nm versus the typical 765 to 850 nm wavelength through simulations and experimental demonstrations.

Approach: We evaluate the photon budget, depth sensitivity, and SNR for detecting blood flow changes using numerical simulations. We discuss continuous wave (CW) and time-domain (TD) DCS hardware considerations for 1064 nm operation. We report proof-of-concept measurements in tissue-like phantoms and healthy adult volunteers.

Results: DCS at 1064 nm offers higher intrinsic sensitivity to deep tissue compared with DCS measurements at the typically used wavelength range (765 to 850 nm) due to increased photon counts and a slower autocorrelation decay. These advantages are explored using simulations and are demonstrated using phantom and in vivo measurements. We show the first high-speed (cardiac pulsation-resolved), high-SNR measurements at large source-detector separation (3 cm) for CW-DCS and late temporal gates (1 ns) for TD-DCS.

Conclusions: DCS at 1064 nm offers a leap forward in the ability to monitor deep tissue blood flow and could be especially useful in increasing the reliability of cerebral blood flow monitoring in adults.

Keywords: Monte Carlo simulations; blood flow; diffuse correlation spectroscopy; near-infrared; short-wave infrared.

Fig. 1

(a) Absorption spectra of main chromophores in brain tissue between 600 and 1200 nm assuming $\mathrm{HbT}=80\mu \mathrm{M}$, ${\mathrm{SO}}_2=62.5\%$, 20% volume fraction of fat and 75% volume fraction of water: oxy-hemoglobin (HbO), red squares; deoxy-hemoglobin (HbR), magenta stars; water, blue diamonds; and fat, green crosses. The thick black line is the resulting total absorption. Absorption of melanin, cytochrome oxidase, and other possible chromophores is below ${10}^{-3}{\mathrm{cm}}^{-1}$ in most biological tissues. For shorter wavelengths, hemoglobin absorption increases; for longer wavelengths, water absorption further increases. (b) Reduced scattering coefficient assuming a power law wavelength dependence with $a=20$ and $b=1.5$ for ${\lambda }_0=500\mathrm{nm}$, thick black line, and changing $a$ and $b$ within physiological ranges, colored lines with symbols. (c) Resulting effective attenuation coefficient: thick black line attenuation is derived from the total absorption of (a) and reduced scattering coefficient with $a=20$ and $b=1.5$ for ${\lambda }_0=500\mathrm{nm}$; for red squares and magenta stars, we assumed $b=1$ and = 2, respectively; blue diamonds for low hemoglobin, assuming $\mathrm{HbT}=40\mu \mathrm{M}$ and ${\mathrm{SO}}_2=50\%$; green crosses for high hemoglobin, assuming $\mathrm{HbT}=150\mu \mathrm{M}$ and ${\mathrm{SO}}_2=75\%$.

Fig. 2

Histograms of absorption, reduced scattering, and effective attenuation coefficients at 765, 785, 850, and 1064 nm, respectively, randomly sampled ($N=\mathrm{150,000}$) for $40\mu \mathrm{M}<\mathrm{HbT}<120\mu \mathrm{M}$, $40\%<{\mathrm{SO}}_2<85\%$, $0.55<f_{\mathrm{H}2\mathrm{O}}<0.95$, $0<f_{\text{fat}}<0.4$, $8<a<25$, $0.5<b<2.4$ to cover the vast majority of possible circumstances for tissue measurements. Vertical bars indicate average values. In pairwise comparisons (across the set of randomly sampled parameters), ${\mu }_{\mathrm{eff},1064}$ was lower than ${\mu }_{\mathrm{eff},765}$, ${\mu }_{\mathrm{eff},785}$, and ${\mu }_{\mathrm{eff},850}$ in 96%, 93%, and 99% of the time, respectively, indicating 1064 nm nearly always offers lower attenuation than the shorter wavelengths commonly used for DCS measurements.

Fig. 3

Simulated $g_2$ curves for measurements with 3-cm source–detector separation on a homogeneous sample with optical properties matching Table 1, assuming $\beta =0.5$, an integration time of 10 s, and a blood flow index ${\mathrm{BF}}_i=2\times {10}^{-8}{\mathrm{cm}}^2/\mathrm{s}$: (a) noise free; (b) with realistic noise added assuming a 7-kcps count rate at 765 nm, 8.05 kcps at 785 nm, 10.9 kcps at 850 nm, and 42.5 kcps 1064 nm on each of 4 co-located fibers.

Fig. 4

Probability of detecting a statistically significant change in deep-tissue blood flow from a pair of 10 s measurements as a function of the actual amount of change for noise-realistic simulated measurements at 765 nm (red diamonds), 850 nm (blue squares), and 1064 nm (green circles). We assumed $30\mu \mathrm{M}$ HbT, with a 66% ${\mathrm{SO}}_2$, water fraction of 0.6 (no fat), and ${\mathrm{BF}}_i={10}^{-8}{\mathrm{cm}}^2/\mathrm{s}$ for the superficial layer (1-cm thick), and $80\mu \mathrm{M}$ HbT with a 62.5% ${\mathrm{SO}}_2$, water fraction of 0.75 (also with no fat), and ${\mathrm{BF}}_i=6\times {10}^{-8}{\mathrm{cm}}^2/\mathrm{s}$ for the deep layer. The source–detector separation was set to 3 cm, and we assumed 7 kcps at 765 nm on four co-located fibers, scaling up to 10.9 and 42.5 kcps at 850 and 1064 nm, respectively, reflecting conservative estimates of the benefits of longer wavelengths.

Fig. 5

Intensity autocorrelation function ($g_2$) measurements at 765 nm (red diamonds) and 1064 nm (green circles) in (a) a silicone oil phantom and (b) an intralipid phantom.

Fig. 6

CW-DCS measurements of ${\mathrm{BF}}_i$ (in this case, the Brownian diffusion coefficient $D_b$) in a liquid phantom across wavelengths and phantom optical properties. (a) Optical absorption values, (b) corresponding optical scattering values, (c) corresponding attenuation coefficient values, and (d) recovered ${\mathrm{BF}}_i$ taking into account actual optical properties.

Fig. 7

Measurements of muscle blood flow before, during, and after an arm cuff occlusion maneuver in three different subjects. Measurements at 765 nm (red diamonds) and 1064 nm (green circles) were conducted one after the other. Despite the nonsimultaneous nature of the measurement, there is very good agreement between measurements taken in the same subject.

Fig. 8

$g_2$ measurements at 765 nm (red diamonds) and 1064 nm (green circles) on the forehead of a human subject. The 1064-nm measurements show a later decay and significantly higher SNR.

Fig. 9

CW-DCS ${\mathrm{BF}}_i$ time course measurements during pressure modulation maneuvers on the forehead of a human subject using a 10 Hz acquisition rate at (a) 785 nm (orange/red symbols) and (b) 1064 nm (light/dark green symbols), respectively. The short separation (5 mm: orange/light green open circles, respectively) ${\mathrm{BF}}_i$ decreases as expected during the pressure period and is no longer pulsatile; cardiac pulsation remains apparent in the long separation channel (30 mm: red/dark green filled circles, respectively) at both wavelengths; however, only 1064 nm offers sufficient SNR to resolve the pulsation at a 3-cm source–detector separation.

Fig. 10

TD-DCS comparison between 765 and 1064 nm. (a) Comparison of the ${\mathrm{BF}}_i$ CoV versus gate start time for TD-DCS measurements conducted at 765 nm (red diamonds) and 1064 nm (green circles) using a 1-cm source–detector separation and the maximum power allowed by regulatory standards. (b) Evidence of cardiac pulsation in ${\mathrm{BF}}_i$ at a 5-Hz acquisition rate for early (0 to 150 ps, light green filled circles) and late (1000 to 1150 ps, dark green open circles) gates.