2.1.

Advantages

DCS offers several advantages over other methods that measure CBF. Unlike traditional clinical CBF measurement modalities, such as xenon-133 computed tomography (xenon-CT) and positron emission tomography, DCS does not expose the patient to ionizing radiation. Furthermore, unlike safer new methods such as ASL-MRI, DCS is relatively inexpensive, portable, and the DCS measurement can easily be carried out multiple times during a patient’s hospital stay. Lastly, in contrast to transcranial Doppler ultrasound, which probes blood flow in major vessels, DCS can be used at any location on the patient’s head to assess regional cortical microvascular flow.

While these advantages of DCS are important for patients of all ages, they are most significant for infants, a population for whom access to MRI is limited. Indeed, DCS holds the potential to fill a critical niche for the assessment of neonatal brain health, brain development, autoregulation, responses to physiological manipulation or therapeutic intervention, and functional responses to sensory and cognitive stimuli. Because measurements can be performed in a rapid, noninvasive, and low-risk fashion using a portable bedside instrument without anesthesia or sedation, DCS provides distinct advantages over other perfusion modalities used in infants, such as ASL-MRI, which requires transport from clinical units to the radiology department, sedated or sleeping patients to limit motion, and which suffers from noise when CBF is low (as is typically seen in infants).

DCS instrumentation is easy to assemble and requires only one wavelength that can be chosen based on laser availability and cost (e.g., by contrast, NIRS requires multiple wavelengths within the NIR tissue absorption window). In addition, unlike the NIRS technique, DCS measurements do not require reference phantoms for quantification. Finally, as noted above, when DCS is integrated with NIRS, the combined instrumentation derives a more robust picture of brain health, i.e., via measurement of tissue oxygenation, cerebral blood flow, and oxygen metabolism.

In addition to the direct advantages of the diffuse optical technology over the existing methods, DCS theoretical concepts can be adapted to improve other blood flow imaging techniques. For example, the elements and concepts from DCS have recently found use in more surface-sensitive CBF imaging methods, such as laser speckle contrast imaging (LSCI);¹⁵16.17.18.19.^–20 further, the combination of DCS theory, spatial frequency-domain imaging, and LSCI has led to quantitative measurements of the physical properties of liquids²¹ and depth resolved blood flow measurements.²² Recent experiments have also increased the depth sensitivity of blood flow imaging using DCS concepts.²³

2.2.

Limitations

Numerous validation studies in both adults and children have shown that the relative changes in ${\mathrm{CBF}}_{\mathrm{i}}$ over time (i.e., rCBF) measured by DCS agree quite well with rCBF measured by other modalities (see Fig. 2 and Ref. 2). However, the quantification of absolute ${\mathrm{CBF}}_{\mathrm{i}}$ remains to be determined across a wide range of subjects and tissue types. To date, the correlations between ${\mathrm{CBF}}_{\mathrm{i}}$ and absolute CBF (measured by other modalities) have only been observed in young piglets and in human neonates, i.e., population cohorts in whom the ratio of extracerebral thickness to the source-detector separation is approximately 0.25 or smaller. In older pediatric populations and in adults, the contribution of the thick scalp and skull layers to ${\mathrm{CBF}}_{\mathrm{i}}$ cannot be neglected; in these cases, the absolute quantification of ${\mathrm{CBF}}_{\mathrm{i}}$ will most likely require the use of multilayered modeling.²⁴ Moreover, because of the relevant scalp contribution in adults, it is desirable to account for probe pressure effects,²⁵ especially when the source-detector separation is less than 2.5 cm. In neonates and in small animals, these factors play a relatively small role, most likely because the ratio of extracerebral thickness to source-detector separation is small, and it is therefore more likely that the detected photons will travel through the cortex.

Another important parameter needed for quantification of ${\mathrm{CBF}}_{\mathrm{i}}$ is the tissue reduced scattering coefficient, ${\mu }_{\mathrm{s}}$’, i.e., the reciprocal of the photon random walk step length. Although ${\mathrm{CBF}}_{\mathrm{i}}$ is influenced by both tissue absorption and scattering coefficients, ${\mu }_{\mathrm{a}}$ and ${\mu }_{\mathrm{s}}$’, respectively, uncertainties in ${\mu }_{\mathrm{a}}$ lead to relatively small errors in ${\mathrm{CBF}}_{\mathrm{i}}$ ($<20\%$), whereas uncertainties in ${\mu }_{\mathrm{s}}$’ can translate to substantial errors in ${\mathrm{CBF}}_{\mathrm{i}}$.²⁶ Thus, when quantifying an absolute ${\mathrm{CBF}}_{\mathrm{i}}$, it is desirable to combine DCS with a method to determine absolute ${\mu }_{\mathrm{a}}$ and ${\mu }_{\mathrm{s}}$’ for the tissue of interest, e.g., frequency- or time-domain NIRS. We note, however, that in neonates, Jain et al. observed a strong correlation between ${\mathrm{CBF}}_{\mathrm{i}}$ and CBF in the sagittal sinus despite assuming a fixed ${\mu }_{\mathrm{s}}$’ across the entire cohort [see Fig. 2(b)].¹² This agreement, at least in part, may have been a result of the homogeneity of the patient cohort, i.e., all were $<1$ week neonates with congenital heart defects. Clearly, in future studies it will be better to measure ${\mu }_{\mathrm{s}}$’ in all subjects whenever possible. (Note: the effects of uncertainties in ${\mu }_{\mathrm{a}}$ and ${\mu }_{\mathrm{s}}$’ on ${\mathrm{CBF}}_{\mathrm{i}}$ are diminished when monitoring relative changes in ${\mathrm{CBF}}_{\mathrm{i}}$ over time, since these tissue parameters change little except in cases of extreme interval blood loss/gain, edema, etc).

As is the case with other optical modalities, the penetration depth of DCS, while larger than wide-field optical methods, such as LSCI, is limited compared to traditional clinical modalities, such as MRI and CT. Like NIRS, a common rule-of-thumb for DCS is that it samples a banana-shaped region underneath the optical probe that is roughly one-third to one-half of the source-detector separation;¹ thus, typical DCS experimental setups with source-detector separations of 2 to 3 cm are sensitive to superficial cortical regions. CBF in deep brain structures has not been quantified by DCS. To increase the penetration depth, larger source-detector separations are required. The use of larger source-detector separation configurations, however, is challenging since DCS optimally requires single-mode detection fibers for measurement of the intensity autocorrelation function. The throughput of single-mode fibers is relatively low and leads to a low signal-to-noise ratio (SNR), especially in adults where large source-detector separations ($>2\mathrm{cm}$) are needed to detect cerebral signals. This low SNR becomes even more problematic in the presence of hair and/or dark skin; it can increase acquisition times and reduce temporal resolution. To overcome the issue of low SNR, most groups employ multiple detector fibers bundled together in the same position, and then the autocorrelation curves from all detectors are averaged to derive a single curve. Unfortunately, this strategy is a relatively expensive option that requires many photon counting detectors. Alternative strategies to improve SNR include increased light collection through improved collection optics, such as those presented by Dietsche et al.,²⁷ or the use of few-mode (rather than single mode) detection fibers.²⁸ Additionally, improved SNR can be achieved by increasing the amount of light delivered to the tissue;²⁹ however, the ANSI standards for skin exposure must be considered, and a large diameter source fiber and/or diffuser should be used to keep light exposure levels within safe limits.

Finally, a significant limitation of DCS is its susceptibility to motion artifacts. Of course, most neuroimaging techniques are sensitive to motion artifacts, hence this issue is not unique to DCS. To understand this problem more deeply, recall that blood flow contrast in DCS primarily derives from intensity fluctuations due to moving particles in the sampled volume. Therefore, the motion of the probe when it is in contact with the patient can cause additional fluctuations due to static tissue (and other) structures that can add noise (random and systematic) to estimates of ${\mathrm{CBF}}_{\mathrm{i}}$. In practice, securing the optical probes as firmly as possible to the subject’s head has reduced motion artifacts. Looking forward, DCS can employ the same analytical approaches used in functional NIRS to account and filter for motion (principle component analysis, wavelet analysis, spline interpolation, etc.^30,31) and thereby eliminate artifacts via improvements in optical probes.³² On the experimental side, Yucel et al. have had excellent recent success in reducing motion artifacts in NIRS data by gluing optodes to the tissue surface with collodion, a glue used with electroencephalogram (EEG) electrodes.³² This approach may also be feasible for DCS, although it may be more practical for utilization in critically ill patients when the potential benefits from cerebral blood flow monitoring outweigh the annoyances of gluing fibers. Additionally, recent approaches with DCS in muscle studies during exercise have successfully employed a dynamometer to gate DCS acquisition and partially remove the effects of motion;^33,34 variants of this approach might be useful for the brain.

3.1.

Animals

Animal models have provided a rich environment for the validation of DCS. Early studies focused on measurement and validation of relative changes in CBF with time (rCBF): Zhou et al. used a closed head injury neonatal piglet model to demonstrate excellent one-to-one agreement of DCS changes in CBF against the changes in CBF measured with fluorescent microspheres [see Fig. 2(a)];¹⁴ Carp et al. validated rCBF measured by DCS in a rodent model against ASL-MRI.³⁵ Recent work has also shown agreement between absolute ${\mathrm{CBF}}_{\mathrm{i}}$ with absolute CBF measured with bolus tracking time-domain NIRS using indocyanine green dye in piglets. Diop et al. calibrated DCS ${\mathrm{CBF}}_{\mathrm{i}}$ to physiologic units ($\mathrm{ml}/\min /100\mathrm{g}$) in piglets,¹⁰ while Verdecchia et al. performed a similar experiment to calibrate ${\mathrm{CBF}}_{\mathrm{i}}$ for individual piglets during various physiological manipulations.¹³

In addition to these validation studies, the quantitative and noninvasive aspects of DCS CBF measurements are very attractive for a variety of animal studies. In rodents, DCS offers the possibility of deep tissue CBF quantification without the need for thinning or removal of the skull, thus leaving the brain unperturbed (i.e., as opposed to more traditional methods of CBF measurement in animals such as laser Doppler flowmetry and laser speckle contrast imaging). DCS has been used in animal models to study baseline and activated hemodynamics as well as related metabolic changes.³⁶ DCS has also been used to study diseases such as ischemic stroke,^37,38 phenomena such as cortical spreading depressions,³⁹ and CBF responses to different anesthetics and carbon dioxide.⁴⁰ Additionally, DCS has been used to extract depth-resolved spatial maps of blood flow.^37,39 In sum, these studies have demonstrated the potential of DCS in preclinical and neuroscience animal studies. Arguably, it appears to be a more robust and quantitative measure of CBF than laser Doppler (which has limited depth penetration and poor repeatability), and its validation has opened the possibility for future longitudinal studies due to the noninvasive nature of the technique.

3.2.

Adults

Although DCS has been largely restricted to measures of rCBF in adults due to difficulties in absolute baseline quantification (see Sec. 2.2), the above-mentioned advantages (see Sec. 2.1) have resulted in numerous preliminary studies demonstrating the clinical potential of DCS. DCS has been successfully applied to monitor rCBF during clinical interventions, such as postural manipulation or blood pressure regulation in patients with ischemic stroke,⁴¹ traumatic brain injury, and subarachnoid hemorrhage.^42,43 These works have demonstrated the need (and potential) for individualized monitoring of cerebral hemodynamics during therapeutic interventions. In essence, we are discovering that the standard treatment protocols designed to optimize cerebral perfusion in damaged tissue do not always yield the desired result; this limitation of the standard protocols is understandable in light of the often highly heterogeneous nature of these brain injuries, and it points to the need for more individualized management (see Fig. 3). In addition, DCS has been used to detect acute spinal cord ischemia in sheep models,⁴⁴ to assess CBF changes due to vocal stuttering,⁴⁵ and to elucidate the effects of normal aging on posture manipulations.⁴⁶ Per new therapeutics, DCS has been used to characterize the cerebral hemodynamic responses due to transcranial magnetic stimulation (TMS) in an effort to elucidate the mechanisms of TMS.⁴⁷ Finally, per functional activation studies, DCS has shown spatiotemporal sensitivity to finger tapping⁴⁸ and visual stimulation,^49,50 thereby demonstrating its utility for basic neuroscience/physiological studies in patients in a natural environment.

Fig. 3

Relative changes in cerebral blood flow measured with DCS during postural manipulation of head-of-bed (HOB) position from 30 to $-5°$ (inset) for (a) a typical healthy subject, (b) a stroke patient who demonstrated the typical response of impaired autoregulation in the injured hemisphere when HOB is lowered, and (c) a stroke patient who demonstrated a paradoxical response in the injured hemisphere during the lowering of HOB, seen in approximately 20% of stroke patients. (a) Left hemisphere in dark gray, right in light gray, (b,c) contralesional in dark gray, ipsalesional in light gray. Figure courtesy of Mesquita et al.³

By directly addressing some of the limitations noted above (Sec. 2.2), it should be possible to expand the use of DCS for clinical monitoring in adults. Improvements to increase SNR and temporal resolution will help the development of functional experiments in adults. Lower costs for sources and detectors should permit whole brain functional imaging experiments that will start to investigate multiregion evoked CBF and ${\mathrm{CMRO}}_2$ changes as well as global resting state CBF and ${\mathrm{CMRO}}_2$ for exploration of regional functional activity and functional connectivity in the adult brain.

3.3.

Infants

Infants and young children are ideal populations to study with DCS, because they have thinner extracerebral layers than adults. Thus, in this population the community has observed higher SNR, excellent reproducibility of ${\mathrm{CBF}}_{\mathrm{i}}$ (i.e., within 10% to 15%¹¹), and greater sensitivity to cortical tissue as compared to adults. Further, infants often have less hair than adults, a mundane but important factor that improves SNR.

To date, neonatal studies with DCS have branched in many directions. In contrast to adults, where the majority of clinical research has focused on changes over time caused by relatively brief interventions (i.e., interventions of order 1 h), neonatal experimental results encompass both episodic monitoring as well as periodic quantification of absolute ${\mathrm{CBF}}_{\mathrm{i}}$. Episodic monitoring has been used to quantify the response to hypercapnia,⁵¹ postural manipulation,⁵² drug delivery,⁵³ surgical intervention,⁵⁴ therapeutics,^53,55 and functional activation.⁵⁶ These studies have provided previously unknown and very valuable information about CBF and ${\mathrm{CMRO}}_2$ responses to intervention. For example, Buckley et al. observed significant and substantial dose-dependent increases in CBF following bolus administration of sodium bicarbonate to correct metabolic acidosis.⁵³ Recognizing these dose-dependent effects may be especially important in treating patient populations where the risks of rapid fluctuations in CBF may lead to subsequent brain injury and thus limit the benefits of correcting the underlying acidemia. Dehaes et al. have observed substantial decreases in ${\mathrm{CBF}}_{\mathrm{i}}$ and ${\mathrm{CMRO}}_{2\mathrm{i}}$ in hypoxic ischemic injured neonates under therapeutic hypothermia.⁵⁵ The ability to measure ${\mathrm{CMRO}}_2$ during therapeutic hypothermia and assess its dynamic may help to individually optimize and guide this therapy.

Further, Roche-Labarbe et al. recently demonstrated the feasibility of DCS functional studies in infants.⁵⁶ By combining NIRS with DCS, both CBF and ${\mathrm{CMRO}}_2$ changes are measured (see Fig. 4). Understanding metabolic and vascular functional measures, as well as their coupling relationships, holds tremendous potential for assessment of perceptual and cognitive development and for deriving early indications of neuropsychological and behavioral deficits in at-risk populations.

Fig. 4

Fractional changes of oxy- and deoxy-hemoglobin concentrations (rHbO and rHbR, respectively), as well as cerebral blood flow (rCBF) and cerebral oxygen metabolism (${\text{rCMRO}}_{2\text{i}}$) measured with NIRS and DCS overthe left somatosensory cortex in preterm infants during 5 seconds of gentle stimulation of the hand, indicated by the gray shaded region.⁵⁶

Following numerous validation studies,² the periodic measures of ${\mathrm{CBF}}_{\mathrm{i}}$ have been used to investigate changes in cerebral hemodynamics over the first month of development in term and preterm infants,^11,57 as well as regional and hemispheric differences in cortical flow.⁵⁸ These works elucidate the normal evolution of ${\mathrm{SO}}_2$, CBF, and ${\mathrm{CMRO}}_2$ in the developing neonate, which is critical to understanding brain in disease states. Furthermore, the works by Roche-Labarbe et al.⁵⁷ and Dehaes et al.,⁵⁵ in particular, highlight the relative insensitivity of ${\mathrm{SO}}_2$ alone, i.e., the standard parameter provided by commercially available stand-alone NIRS devices, as a measure of brain health. These works demonstrate the obvious advantage that DCS adds in providing a more complete picture of neonatal development through quantification of CBF and ${\mathrm{CMRO}}_2$.