Simultaneous in vivo measurements of the total hemoglobin, oxygen saturation, and tissue blood flow via hybrid near-infrared diffuse optical techniques

Stroke is one of the most common causes of death and disability and results in enormous societal costs associated with rehabilitation, long-term care, and workforce loss.¹ Approximately 90% of all strokes occur due to cerebral ischemia, with the remainder due to cerebral hemorrhage.² The only approved treatment for ischemic stroke is recanalization of occluded arteries via thrombolysis with alteplase within the first hours of symptom onset.^3–5 However, implementation of recanalizing therapy within this narrow therapeutic window is difficult in clinical practice because neurological examination and imaging are needed to exclude hemorrhagic stroke. In clinical reality, this treatment can be given to only a small minority (2%–5%) of stroke patients.^6,7 The establishment of mobile stroke units based on x-ray computed tomography scanning in Germany and the United States has made earlier identification and treatment of ischemic stroke a reality and has the potential to substantially improve outcomes.^8,9 However, compact, less expensive devices are needed for stroke diagnosis at emergency sites in order to increase the number of patients who benefit from effective treatment.

Near-infrared light systems (NIRS) may be the ideal method of overcoming the aforementioned clinical dilemma. They are sensitive to hemodynamic parameters, safe, noninvasive, cheap, and portable. There are two basic requirements for the near-infrared light techniques that can achieve the required goals. First, the techniques should output quantitative parameters that help to differentiate ischemic and hemorrhagic tissues. One common application of NIRS is measuring brain functional activity.¹⁰ However, these systems measure relative hemodynamic changes using the modified Beer-Lambert Law method, which is not suitable for diagnosing stroke. Commercial cerebral oximeters that use spatially resolved spectroscopy (SRS)^11,12 can measure the absolute tissue oxygen saturation (StO₂). However, their ability to measure hemodynamic parameters is limited. The frequency and time-domain methods^13,14 are ideal ways of obtaining the absolute total hemoglobin concentration (HbT) and tissue oxygen saturation, which are useful in stroke diagnosis. However, these devices are complex. Second, the technique used should measure as many parameters as possible in order to provide full hemodynamic stroke information. Diffuse correlation spectroscopy (DCS) has recently been used to measure blood flow.¹⁵ DCS is a dynamic technique that analyzes rapid temporal fluctuations in light scattered by moving red blood cells and uses a diffuse correlation approach to derive local microvascular tissue blood flow. A research group from the University of Pennsylvania has combined near-infrared diffuse optical spectroscopy (NIRS-DOS) with DCS to measure tissue oxygen saturation and blood flow.^16,17 However, they used separate NIRS-DOS and DCS systems with separate sources and detectors. Yu Shang et al. have proposed adding a second long coherent continuous wave laser diode to measure tissue oxygen saturation and blood flow simultaneously.¹⁸ D. Tamborini et al. have proposed using multi-color, multi-distance DCS to fit the absolute optical properties and blood flow index (BFI) using diffuse optical theory and diffuse correlation theory and then validate it using a tissue-like phantom.¹⁹ 

In this study, we fabricated a hybrid diffuse optical device that combines NIR-DOS and DCS. The device used hybrid light sources consisting of three incoherent laser diode (LD) sources (690 nm, 808 nm, and 830 nm) and one long coherent LD source (785 nm). Single-photon-counting avalanche photodiodes (APDs) detected light reemitted from the tissue. Optical switches were used to switch the four laser sources. A software correlator was utilized to calculate the optical intensity and the normalized correlation value of the optical intensity simultaneously for the four wavelengths.²⁰ The absolute HbT and StO₂ were calculated using the decays of the optical intensities of the four wavelengths based on a multi-color SRS method.^11,12 The absolute BFI was derived from the normalized autocorrelation value of the optical intensity of a long, coherent laser source. Thus, this system allowed the HbT, StO₂, and BFI to be obtained. The entire system was compact, cheap, and easy to manufacture. Arm cuff experiments were conducted using a healthy subject to validate the device capabilities. Changes in hemodynamic parameters including the HbT, StO₂, and BFI of the healthy subject were measured continuously. The results met expectations set by a previous study.^20,21 Since this device was able to measure the HbT, StO₂, and BFI, we believe that it may have potential clinical applications for stroke diagnosis at emergency sites.

A schematic of the hybrid diffuse optical device is shown in Fig. 1. The device combined DCS and NIRS-DOS. The DCS portion consisted of a mode-hope free, long coherence-length, continuous-wave laser (785 nm, 90 mW, DL785-100-3O, CrystaLaser Inc., USA) and four single-photon avalanche photodiode detectors (SPCM-AQRH-W4, Excelitas, Canada) whose outputs were fed to a custom-built software autocorrelator.²⁰ The software correlator sampled the photon count and calculated both the optical intensity and its normalized autocorrelation value. The continuous-wave NIRS-DOS module consisted of three continuous-wave lasers (Wavespectrum Laser, China) at 690 nm, 808 nm, and 830 nm with optical power levels of 50 mW, 100 mW, and 100 mW, respectively. The BFI was derived from the normalized autocorrelation value at 785 nm. The normalized autocorrelation values of the other wavelengths were abandoned. The HbT and StO₂ were derived from the decays of the optical intensities of all four wavelengths. Optical switches were used to switch the four measurement wavelengths (one for DCS and three for NIRS-DOS). The data acquisition frequency was 20Hz and the optical switch frequency was 0.2 Hz. Each measurement cycle required 20 s.

The only hardware in the software correlator was a set of counters (PCIe-6612, NI, USA). A shielded I/O connector block (SCB-68, NI, USA) was used to connect the PCIe-6612 board to the outputs of the single-photon-counting APD. The four counters on the PCIe-6612 board were synchronized by the same sampling clock (1/Δt) at up to 5 MHz. The lag times were $∑M=09∑N=078+N⋅2M⋅Δt$ and $∑M=00∑N=07N⋅2M⋅Δt$⁠. Since the correlation bin width was always the same as the sampling clock and only the lag time increased in line with a multi-tau scheme,²² the algorithm was quite simple. The optical intensities and normalized autocorrelation values of the four channels were calculated in real-time using a computer with an integration time of 0.05 s (20 Hz).

Optical probes consisting of a multi-mode source fiber (62.5 μM/0.27 NA, Corning, USA) and three single-mode detection fibers (4.4 μM/0.13 NA, 780-HP, Nufern, USA) with source-detector separations of 1.08 cm, 1.95 cm, and 3 cm were custom built.

Arm cuff experiments were conducted to test device performance. The subject was asked to sit comfortably in a chair. Their forearm was put on their leg. A probe with embedded sources and detection fibers was secured on the forearm of the subject. An arm cuff was placed on the bicep on the same side of the probe. Black cloth was used to protect the probe and forearm from external light. After 3 min of baseline condition measurements, the blood in the arm was blocked for 3 min by inflating the blood pressure cuff to 200 mmHg. The experiment also collected post-occlusion data for 3 min. The in vivo experiment was approved by the Institutional Review Board of the Arm Medical University.

$zb=2/μs′1+Reff/1−Reff$⁠, R_eff is the effective reflection coefficient that accounts for the index mismatch between tissue and air, k₀ is the wave number of light in the medium, λ is the wavelength of light, τ is the lag time, ρ is the source-detector separation, BFI is the quantitative blood flow index, and μ_a and μ_s’ are, respectively, the absorption and reduced scattering coefficients.

The arm cuff experiment optical intensity changes at 1.08 cm, 1.95 cm, and 3 cm for the four wavelengths are shown in Fig. 2. The vertical black lines in the figure indicate when the arm cuff is inflated and deflated. The optical intensity is averaged over 1 s. For the baseline period, the optical intensities are stable and are indicated in the figure. The optical intensities all decrease and then recover during the arm occlusion period. The optical intensities stabilize after arm cuff deflation.

We used the optical intensity decay over different separations to obtain the μ_eff, as shown in Fig. 3. The change in μ_eff at 690 nm was much bigger than those of the other wavelengths, since the μ_eff at 690 nm is more sensitive to the Hb. Note, changes in μ_eff and changes in the optical intensities are not consistent with each other. For example the optical intensity at 830 nm is lower during the occlusion period than during the baseline period. This indicates that attenuation increases but the actual μ_eff at 830 nm is lower during the occlusion period than the baseline period. Since μ_eff removes the same fiber coupling influence at all distances and is robust against movement artifacts, we believe that it should be used instead of the change in optical intensity to calculate the hemodynamic parameters. This is because the single-mode detection fiber used here is susceptible to movement artifacts due to its small diameter of only 4.4 μm.

Equation (4) is non-linear equation, so we use the optimization method to solve it. The fraction of water is fixed at 14% and does not change between experiments and subjects. The initial HbT, StO₂, a, and b are assumed to be 0.126 mmol/L, 67%, 0.576, and 20 respectively.²⁴ We determine the absolute HbT and StO₂ using the ‘fminsearch’ function in Matlab. For this subject, the HbT and StO₂ are 0.08 mmol/L and 63%, respectively. We assume that scattering is constant and solve the linear equation using constant a and b values to identify the relative changes in HbT and StO₂, as shown in Fig. 4. The HbT is nearly constant and the StO₂ decreases linearly from 63% to 51% during the 3 min arm occlusion period. After occlusion, overshot increases in the HbT and StO₂ occur. This is reasonable because of the overshot perfusion of blood.

Figure 5(a) shows the normalized autocorrelation value of the optical intensity at 785 nm. The low value of β at 1.08 cm occurs because the high detected optical intensity saturates the detector.²² Obviously, the correlation curve delay is much slower during occlusion than under normal conditions. Fig. 5(b) shows the BFI as determined from the correlation values based on diffuse correlation theory with calculated μ_a (0.15cm^-1) and μ_s’ (4.93cm^-1) values. At the three distances noted, the BFI decreases by 24%, 15%, and 10% immediately upon occlusion of the arm. This represents a mean 16% decrease. This is reasonable since blood cell movement slows immediately. After the cuff is released, the BFI experiences an overshot increase.

Unlike a traditional CW-NIR or NIR-DOS device, our device can measure blood flow directly by adding a long, coherent near-infrared laser source. The relative BFI has already been shown to indicate relative changes in blood flow.^15,25 We have developed a hybrid diffuse optical device that combines NIRS-DOS with DCS to diagnose strokes at an emergency site. The HbT and StO₂ are derived using the multi-color SRS method. The BFI is determined from the normalized autocorrelation value of the optical intensity of a long coherent laser source. The single-photon-counting APDs detect light reemitted from the tissue, which simplifies detector calibration and enhances detection sensitivity. The resulting system is compact, cheap, and easy to manufacture.

The arm cuff experiments showed that the device can measure the HbT, StO₂, and BFI simultaneously. The relative changes in the StO₂, HbT, and BFI agree well with previous studies.^20,21 The quantitatively derived StO₂, HbT, and BFI parameters are reasonable.^12,24 The results also show that quantitative hybrid diffuse optical measurements can distinguish ischemic tissue, and therefore that the proposed technique can be used to diagnose a stroke at an emergency site. However, further improvements are needed. First, shorter wavelengths that are sensitive to deoxy-hemoglobin should be added to improve the fitting accuracy since the μ_eff values at 780 nm, 808 nm, and 830 nm are similar. Second, a quantitative brain reconstruction model should be developed. There is software available that may help to achieve this goal.^26–28 

In conclusion, we have built a hybrid diffuse optical device that combines NIR-DOS and DCS to noninvasively and simultaneously measure the total hemoglobin, oxygen saturation, and tissue blood flow. The device was demonstrated by measuring hemodynamic parameters during an arm cuff experiment. Since strokes are related to abnormal hemodynamic changes in cerebral vasculature and our device is sensitive to such changes, portable, and potentially easy to apply without extended technical expertise, we believe that it has potential clinical applications in stroke diagnosis at emergency sites.

This work is funded by the National Key Technology R&D Program of China (2017YFC1200400, 2014BA104B01, 2014BA104B05, 2015BA101B01) and the National Natural Science Foundation of China (11604316). We thank AIP Author Services for its linguistic assistance during the preparation of this manuscript.