In vivo imaging is necessary in a wide range of biomedical applications. In the brain, neurovascular coupling has been widely studied due to its importance in developing a better understanding of both the normal and diseased brain. The hemodynamic component to the neurovascular response consists of the changes in cerebral blood flow (CBF), blood volume, and hemoglobin concentration. To fully understand the neurovascular coupling, it is desirable to image all of the components of the hemodynamic response simultaneously. Optical methods have been widely used to image hemodynamic changes in the cortex, however, most of the optical methods have been limited to imaging only the changes in blood volume or hemoglobin concentration. The primary limitation of full-field imaging of CBF changes has been the lack of a method with the necessary temporal and spatial resolutions. Only recently has full-field imaging of all hemodynamic components been demonstrated in studies of neurovascular coupling.1, 2 Therefore, it is essential to develop improved techniques for imaging blood flow. A light-based apparatus can be minimally to noninvasive, and easy to design. Two main classes of techniques for detecting CBF have emerged: laser Doppler velocimetry (LDV) and speckle contrast analysis (SCA). Both rely on the interferometric properties of coherent light. SCA techniques are based on the measurement of the local spatial blur rate of the intensity of speckle pattern present on the images. Based on a time-domain measurement, LDV techniques reveal the frequency distribution of Doppler-broadened light. It was shown that LDV and SCA were two distinct approaches to assess the same underlying phenomenon: dynamic light scattering.3, 4, 5, 6
Stern designed the landmark laser Doppler instrument for bioflow assessment.7 The first attempt to use the spatial properties of dynamic speckle patterns for flow visualization were reported by Fercher and Briers,8 whose work led to the development of laser SCA imaging techniques.9, 10, 11 Both techniques have shown a great potential in CBF assessment in vivo. LDV has become popular12, 13, 14 for focal measurements of dynamic changes of CBF, while SCA is rapidly becoming a common tool for imaging purposes.15, 16 The advantage of SCA is a high temporal resolution in conjunction with a good spatial resolution. Its major weakness lies in the fact that the speckle contrast values are not linearly related to blood flow velocity,15 although relative CBF images can be calculated from speckle contrast images by assuming a particular form for the underlying velocity distribution. On the other hand, LDV yields quantitative information (spectra), but its drawback is the limited performance of scanning systems in terms of spatial and temporal resolution. Postprocessing of the laser Doppler spectra17, 18 can enhance the detection capabilities and resolving power of LDV instruments, in particular, for low flows and vascular differentiation, but cannot improve temporal resolution.
The laser Doppler flowmetry signal was reported19 to be correlated to actual blood flow in vivo, but with significant variability from point-to-point measurements within an area of a tissue, which clearly pointed out the necessity for laser Doppler imaging (LDI) techniques. Standard scanning laser Doppler instruments,20 based on time-domain measurements, provide spatial resolution at the expense of temporal resolution. A time-domain laser Doppler measurement technique with a fast CMOS camera was introduced21, 22 as well as new methods to take advantage of the time-domain integration of cameras,23, 24 providing alternatives to speckle contrast analysis to perform wide-field imaging of bioflow.
We reported a CBF assessment in mice with a wide-field LDI instrument,26 relying on a frequency-domain detection of scattered light,25 which is an alternative to either time-domain or spatial-domain detection techniques.6 In this scheme, scattered light is analyzed by a detuned local oscillator (LO), which consists of a fraction of the main laser beam shifted in frequency, and detected by a CCD camera, acting as a narrow spectral filter, enabling one to scan the spectral content of the object field. The main strength of this tool lies in the spatial resolution of the Doppler maps, which is suitable for functional brain imaging since the lack of spatial resolution has been noted to be a limitation of conventional LDI. Here, we report the results of experiments designed to determine whether the instrument would enable detection of subtle changes of CBF in mice.
Description of the Instrument
The basic principles of the instrument were described previously.26 However, we provide a summary of the experimental setup here for clarity. The apparatus is designed to analyze scattered light in the frequency domain. The object is illuminated by the main laser beam, whose field undergoes dynamic scattering. The scattered (or object) field carries the signature of the local dynamic state of the object within its spectrum. Early studies27 of Doppler broadening of light by flows showed that the Doppler shift for one scattering event isis the incident wave vector, is the scattered wave vector, and is the instantaneous particle velocity. Further work showed that the distribution of scatterer velocities is linked to the frequency distribution of probing light.4 To measure a spatially resolved spectrum onto the object, is mixed with a detuned LO field on an array detector.26
In the experimental setup, the object (a mouse cranium) was illuminated over the region to be imaged with an average incidence angle . The scattered field was mixed with a frequency-shifted LO field , split off from the main laser beam. To achieve sensitive detection, a spatiotemporal heterodyning of the scattered field was performed. The optical configuration is based either on a classic off-axis lensless Fourier holography scheme28 [sketched in Fig. 1b ] or a slightly modified one, into which a microscope objective (Olympus, magnification) took place in the object arm [Fig. 1a]. The choice of the presence of a microscope objective in the apparatus is dictated by spatial resolution demand. A single-mode cw Mitsubishi ML120G21 diode provided the main laser beam at (angular frequency ). Two Bragg cells (acousto-optic modulators, AOMs), were used to shift the LO frequency by the tunable difference of their (coherent) driving frequencies: . The LO beam was focused by a -focal-length lens, to place the virtual LO focal point in the object plane (Fig. 1). A beamsplitter was used to combine and with a small off-axis angle . The distance between the detection plane and the object plane (or virtual object plane) was . A PCO PixelFly camera [frame rate , square pixels, pixel size ] was used to record a sequence of images , where . For each detuning frequency of interest, the camera sampled a modulated interference pattern. This modulation is a consequence of optical mixing of the object field with the detuned LO field.
A digital hologram was recorded according to the frequency-shifting method introduced in the heterodyne holography technique, used in several imaging and detection schemes29, 30, 31 and that turns out to be sensitive enough to discriminate Doppler-shifted photons among heavily scattered light29, 31 according to their frequency. The frequency-shifting method was chosen for its accuracy.32
The LO frequency was set to , where , to sample a map of the angular frequency component [corresponding to the frequency, in hertz] of the scattered field impinging on the camera.25 The choice of meant that four images were used to measure both quadratures of the scattered field. A sequence of images was recorded at each detuning frequency for averaging purposes to improve the signal-to-noise ratio (SNR). Each spatial map at a given temporal frequency point was measured in and averaged over .
Signal Demodulationis a constant, and . Two-dimensional maps of were realized for each arbitrary frequency point , one at a time, set by the detuning frequency of the LO.
Two similar optical configurations were used. One is a modified lensless Fourier holographic setup, into which a magnification microscope objective was placed in the object arm [as sketched in Fig. 1a]. The other is a regular lensless Fourier setup [Fig. 1b]. In both cases, the LO focal point was positioned in the relevant object plane to cause the and average curvatures to match in the detection plane. The presence of the microscope objective in the setup was guided by the desired spatial extent of the image. The field of view was without and with the objective.
The lensless Fourier setup was used to avoid the necessity for a general Fresnel reconstruction algorithm33 to assess the spatial distribution of the scattered field in the object plane. In this case, the reconstruction algorithm consisted of one spatial fast Fourier transform34 (FFT).
A logarithmic-scale quantity was formed to display spectral mapsis the quantity assessed in a region of the reconstructed hologram where the object light contribution is null. It was reported to be shot-noise dominated.31 between 0 and (the spacing interval is ).
To increase the SNR, a sequence of images was taken to build each map at one given frequency point, setting , where . Note that was averaged over the -image sequence. Each frequency point of the spatial distribution was measured in . The measurement of a whole cube of data (space, frequency) took and ( exposure time).
The results of three different experiments are presented: postmortem residual flow mapping (detection of reduced blood flow), minimally invasive CBF detection in vivo, and monitoring of drug-induced CBF changes through the intact cranium.
Animal procedures were conducted in strict compliance with approved institutional protocols and in accordance with the provisions for animal care and use described in the European Communities Council directive of November 24, 1986 (86-16-09/EEC). In total, five C57/Bl6 mice (Charles River) aged P25 to P30 were used. Anesthesia was achieved by injecting a mixture of xylazine [ intraperitoneal (IP)] and ketamine ( IP). Cranial skin and subcutaneous tissues were excised linearly over the sagittal suture. Animals were positioned on a stereotaxic frame (World Precision Instruments) to ensure stability of the preparation. Cortical bones were regularly hydrated with a saline solution to prevent them from drying.
A first set of data was obtained from three control animals, which were used to determine the optimal setup configuration, to determine the sensitivity of our technique and to visualize intact superficial blood vessels and extract Doppler information. Throughout this procedure, a microscope objective was used in the object arm, as represented in Fig. 1a. The resulting imaging field of view is .
Detection of Reduced Blood Flow
The laser Doppler technique is a useful complement to other noninvasive methods, especially clinical assessment, in the determination of brain death in the newborn. Brain death is characterized by well-defined sequences in deterioration of the flow velocity waveform in blood vessels. This sequence was previously shown to consist of a loss of diastolic flow, appearance of retrograde flow during diastole, diminution in systolic flow in the anterior cerebral artery, and ultimately, lack of flow in the anterior cerebral artery.36 These modifications are the results of a progressive increase in cerebrovascular resistance and a progressive decrease in cerebral perfusion, compatible with the diffuse cerebral necrosis and oedema documented postmortem. Our first set of experiments aimed to determine the threshold of sensitivity of our apparatus. We sought to determine whether our apparatus would detect the Doppler signature of the remaining perfusion in a dead animal. The animal was deeply anesthetized using an overdose of the xylazine-ketamine mixture. A cranial window (excision of a small part of the skull) was performed and images of CBF were collected more than 10 min after cessation of visible heart-beating. Figures 2a, 2b, 2c, 2d represent maps at the Doppler shift , taken at intervals, and possibly suggest a residual capillary flow. Those images are consistent with the presence of a residual flow in the brain persisting soon after death, although the exact origin of this observed Doppler signal is not known.
Minimally Invasive Assessment of Cerebral Blood Flow
Images of the brains of two mice were done through their intact skulls (measured thickness of medial and lateral), without altering it. A sequence of maps at different frequency shifts is reported in Figs. 3a, 3b, 3c, 3d, 3e, 3f . Each of the maps represented in Fig. 3 was obtained in the same mouse in a measurement time of [from 32 images recorded at ]. The most striking result is the contrast reversal between the vessels and the parenchyma regions in the map at [Fig. 3a] and the map at [Fig. 3c]. A more subtle difference can be observed between Figs. 3c and 3d, corresponding to the 384- and Doppler shifts. The flow in the right arm of the central Y-shaped vessel (arrow) has a weak contribution at the , suggesting a lower CBF through it than through its left counterpart. An even more subtle effect can be seen in Fig. 3e: One can distinguish the presence of a transversal vessel (arrows), which cannot be perceived below , suggesting it has a high CBF. Those preliminary results attest to a sensitivity and dynamic range suitable for CBF studies. The typical response frequencies of the flow in the Y-shaped vessels in Figs. 3c and 3d are in the 400- to 500-Hz range, whereas in Fig. 2, we see this signal for a frequency shift, which demonstrates the ability of the setup to map the Doppler signature of very small flows. As an indication, if we make the assumption that the collected light undergoes one scattering event and only in-plane velocities participate in the Doppler shift of light, then, according to Eq. 1, the order of magnitude of the velocity corresponding to a Doppler signal at , mapped in Fig. 2, would be .
Real-Time Assessment of Drug Effects
We tested whether our setup would enable detection of small and reversible variations of CBF under minimally invasive conditions. Therefore, we used a pharmacological paradigm well known to induce modification of blood flow. Effects of variation in monoamine levels (NE, 5-HT) on blood flow are well studied. We thought to monitor the modifications in CBF occurring after a single IP injection with pargyline ( , single injection). Pargyline is an inhibitor of monoamine oxidase (I-MAO) type A and type B (MAOA; MAOB), which are two membrane-bound mitochondrial flavoproteines that oxidatively deaminate a broad range of biogenic amines, including monoaminergic neurotransmitters serotonin and catecholamines37 (dopamine, norepinephrine, and adrenaline). In rodents, as in many other species, cerebral blood vessels display aminergic innervations and MAO activities.38, 39 I-MAOs are largely used and known to produce a fast increase in brain serotonin and norepinephrine levels within after a single intraperitoneal injection40, 41, 42 and are known to increase blood vessel pressure within minutes in vivo.42
In this set of experiments we chose to increase the field of view of the control procedure. To make the image area match the whole cranium of the mouse , the microscope objective is removed from the setup to benefit from the natural heterodyne image field.43 In our experiment, data were collected in the same animal from the instant of drug injection and every after the injection over a period of . An average rms Doppler shift map is shown in Fig. 4a , into which 16 regions of interest (ROIs) are defined. ROIs 1.x and 3.x correspond to vessels, ROIs 2.x to midline, and ROIs 4.x are in the parenchyma. We observed a decrease in CBF occurring between 10 and after the injection that mostly affects small-caliber blood vessels [10 to 20% decrease in rms Doppler shift see Figs. 4c, 4d, 4e, 4f]. At after pargyline injection, CBF had reached preinjection values. The most important relative CBF variation appears to occur in the parenchyma and the vessel ROIs. As an indication under the previously cited assumption, the order of magnitude of the rms velocity corresponding to a rms Doppler shift is .
Our study is consistent with previous ones showing that blood vessel contractions are in part induced by activation of catecholaminergic and serotoninergic receptors. In vivo and in vitro, catecholamines have been shown to produce responses that are consistent with vasoconstriction (contraction, isometric tension, increase in perfusion pressure, and decrease in flow).45, 46 Serotonin has more complex effects on the cardiovascular system but is generally contractile in its effects,47, 48 except for meningal blood vessels.49 Serotonin may act on S2-serotonergic receptors located on vascular smooth muscles and has also been described to act indirectly by amplifying the response to norepinephrine50, 51 and other agonists, of releasing constrictor substance(s) from the endothelium. In addition, the increase in heart rate elicited by monoamine elevation may partially account for an increase in blood pressure. The design of our study does not enable discrimination of whether the observed changes are the result of elevations of 5-HT and/or catecholamines, or whether changes in CBF are due to an increase in heart rate, but this was beyond the scope of our study.
The scope of this study was to report monitoring and mapping of drug-induced average blood flow changes in mice brains, in minimally invasive conditions. As an indicator of flow, we used the rms Doppler shift calculated with 80 linearly spaced spectral points between 0 and . Because the whole spectrum measurement lasted several minutes, this indicator is a time-averaged value of the Doppler signature of the actual flow.
Our setup was based on a camera detection used to measure a map of one spectral component of the scattered light out of a few recorded images. Both quadratures of the scattered field carrying the Doppler information where measured in the detector plane and propagated numerical to the object plane, to produce a tunable spectral image. The heterodyne off-axis configuration enables sensitivity suitable for in vivo Doppler mapping of CBF. Our study demonstrated that the frequency-domain wide-field laser Doppler imager is able to detect subtle changes in the CBF of a mouse. We report the suitability of the presented scheme to efficiently measure the spectral signature of a wide range of blood flows, beyond the detector bandwidth.
Our results show a significant spatial discrepancy of the signal, justifying the necessity of the proposed high-spatial-resolution functional imager. To enable a high temporal resolution, fewer spectral points might be recorded to increase the sampling rate, and indicators other than the rms Doppler shift used here to assess flow might be used with this instrument.
The authors are thankful to the reviewers of this paper who have contributed to making it clearer. We acknowledge support from Paris VI University (BQR grant).