NeuroImage 161 (2017) 232–240

Contents lists available at ScienceDirect

NeuroImage

journal homepage: www.elsevier .com/locate/neuroimage

Hybrid photoacoustic and electrophysiological recording of neurovascularcommunications in freely-moving rats

Lei Xi a,b,*, Tian Jin a,b, Junli Zhou d,1, Paul Carney c,d,e, f,1, Hubei Jiang a,b,c,**,2

a School of Physical Electronics, University of Electronic Science and Technology of China, Chengdu, Sichuan, Chinab Center for Information in Medicine, University of Electronic Science and Technology of China, Chengdu, 610054, Chinac Department of Biomedical Engineering, University of Florida, Gainesville, FL, USAd Department of Pediatrics, Neurology and Neuroscience, University of Florida, Gainesville, FL, USAe Wilder Center of Excellence for Epilepsy Research, University of Florida, Gainesville, FL, USAf Mcknight Brain Institute, University of Florida, Gainesville, FL, USA

A R T I C L E I N F O

Keywords:Photoacoustic sensingEEGNeurovascular communicationFreely-moving

* Corresponding author. School of Physical Electronics,** Corresponding author School of Physical Electronics,

E-mail addresses: xilei1985@uestc.edu.cn (L. Xi), hjian1 Present address: Department of Neurology, University2 Present address: Department of Medical Engineering,

http://dx.doi.org/10.1016/j.neuroimage.2017.08.037Received 26 February 2017; Received in revised form 6 JAvailable online 14 August 20171053-8119/© 2017 Elsevier Inc. All rights reserved.

A B S T R A C T

Uncovering the relationships between neural activities and capillary-level hemodynamics such as blood flow andconcentration of hemoglobin in the brain plays an important role in the study of animal behaviors and braindisorders. Here, we developed a miniature probe integrating a photoacoustic sensor and micro-electrodes tosimultaneously record the dynamics of blood flow and total hemoglobin inside a single capillary and the activitiesof surrounding neurons with high spatiotemporal resolution in freely-moving rats. In the somatosensory cortex ofrats, we observed: 1) early hemodynamic response prior to the changes in local field potential duringpentylenetetrazol-induced localized and generalized seizure onsets in both freely-moving and anesthetized rats;and 2) different hemodynamic and neural responses to generalized seizure onsets between freely-moving andanaesthetized rats. These findings suggest that this high-resolution hybrid technique will enable a wide range ofnew studies of behaviors and brain disorders in small animals.

1. Introduction

Although magnetic resonance imaging (MRI) in conjunction with theblood-oxygen-level-dependent (BOLD)method has been used extensivelyin neuroscience research, drawing conclusions from the data has beenoften problematic. Specifically, interpretation has been impeded,because the neural dynamics responsible for BOLD changes remain un-known, entire populations of active neurons can change their activitywithout a significant change in firing rate or field potential activity, andthere is a substantial temporal delay between the onset of neuronal actionpotentials and the rise of the BOLD signal (Logothetis et al., 2001; Log-othetis, 2003; Fox and Raichle, 2007; Buzsaki, 2006). However, evidenceto date has shown that there is a significant, short-time scale correlationbetween the rise of the BOLD response and the increase in local fieldpotential (LFP) activity (Logothetis, 2002; Cauli et al., 2004; Huchzer-meyer et al., 2008), suggesting that a substantial increase in neural ac-tivity tenably sends a signal to local vasculature, inducing vasodilation/

University of Electronic Science andUniversity of Electronic Science andg1@usf.edu (H. Jiang).of North Carolina, Chapel Hill, USAUniversity of South Florida, USA.

uly 2017; Accepted 12 August 2017

constriction (Paulson et al., 2010). To closely examine the interplay be-tween vasculature and neural dynamics, alternative approaches areneeded. For example, different optical absorption-sensitive methods incombination with electrophysiological recordings such as near-infraredspectroscopy (NIRS) (Devor et al., 2012; Jobsis, 1977), optical intrinsicsignal imaging (OISI) (Grinvald et al., 1986) and photoacoustic imaging(PAI) (Wang and Hu, 2012; Beard, 2011; Xi et al., 2012) have beendeveloped. Optical scattering-sensitive methods have also been studied,including optical coherence tomography (OCT) (Huang et al., 1991; Chenet al., 1997), laser speckle contrast imaging (LSCI) (Briers, 2011) anddiffuse correlation spectroscopy (DCS) (KimEmail et al., 2010). In addi-tion, exogenous fluorescence-sensitive techniques are extensivelyexplored, including two-photon microscopy (TPM) (Kuhn et al., 2008),wide-field imaging (Sakad�zi�c et al., 2009) and voltage imaging (Ake-mann et al., 2010). Of all these optical techniques, photoacoustic imaging(PAI), utilizing the absorption of ultra-short light pulses in tissue toinduce wideband acoustic waves, owns greater depth penetration and

Technology of China, Chengdu, Sichuan, China.Technology of China, Chengdu, Sichuan, China.

.

L. Xi et al. NeuroImage 161 (2017) 232–240

higher spatial resolution compared with other optical absorption-basedtechniques (Wang and Hu, 2012).

Advances in PAI include transducer array-based photoacoustic to-mography (PAT) systems for epilepsy studies of anesthetized and awakerats in two- or three-dimensions (Xiang et al., 2013a, 2013b; Tang et al.,2015, 2016a, 2016b), high-speed function optical resolution photo-acoustic microscopy (ORPAM) systems for the study of brain activity inanesthetized and awake mice (Yao et al., 2015; Cao et al., 2017), and anintegrated method combining PAT with ORPAM to fluorescently visu-alize glucose response in mouse brains (Wang et al., 2013; Yao et al.,2013). In addition, functional acoustic-resolution photoacoustic micro-scopy (fARPAM) was used to image hemodynamic changes in rat brainsduring forepaw electrical stimulations (Liao et al., 2010). Efforts havealso been made to develop various algorithms for extraction of bloodflow through photoacoustic Doppler Effect (Cao et al., 2017; Fang et al.,2007; Yao and Wang, 2010; Yao et al., 2010).

Despite these advances, current photoacoustic imaging/flowmetrysystems are bulky to be used in freely-moving rodents, and the relativelylow spatial resolution of PAT and the limited temporal resolution of PAMprevent photoacoustic imaging from broad studies of brain. Furthermore,simultaneous recording of photoacoustic signal and the associated elec-trocorticography (ECoG) signal at the same location remains a majorchallenge for current photoacoustic techniques. To overcome these lim-itations and to explore how neural hemodynamics relates to the under-lying neural activity, here we describe an approach that integrates anultra-high numerical aperture single-mode fiber (UHNA-SMF) basedphotoacoustic sensing probe with micro-electrodes. By integrating pho-toacoustic (PA) sensor and microelectrodes into a miniature device, itwill be possible to study the relationship between blood flow, total he-moglobin and network dynamics in awake, freely-behaving rats, therebyproviding an analogous link between human fMRI/BOLD research androdent in vivo physiology.

2. Materials and methods

2.1. Hybrid miniature sensing probe

The 532 nm ultra-short illumination pulses from a Nd:YAG pulsedlaser (Surelite I-20, Continuum, CA) with a 20 Hz repetition rate and6 nanosecond (ns) duration was attenuated by a group of neutral densityfilters (ND, Thorlabs Inc, NJ), split through passing a beam splitter(BS025, Thorlabs Inc, NJ), spatially filtered using a high-power precisionpinhole (P25C, Thorlabs Inc, NJ) and coupled into a UHNA-SMF(UHNA3, Thorlabs Inc, NJ) via two concave lenses (LB1761, ThorlabsInc, NJ) (Fig. 1a). A fast responsive photodiode (SIR-FC, Thorlabs Inc, NJ)was used to record the fluctuations of the laser pulses to calibrate the raw

Fig. 1. The schematic of the experimental system and hybrid recording probe. (A) The integrasystem are in green, black and orange color, respectively. (B) Cross-sectional close-up view of tscrews and dental acrylic, the ultrasound transducer and fiber-electrode probe were immobile uThe inserted photograph shows a typical rat brain mounted with the recording device in freely

233

photoacoustic signals. The micro-electrodes (Stablohm, California FineWire Company, CA) with a diameter of 30 μmwere positioned and gluedin parallel with the optical fiber. The integrated optical fiber and micro-electrodes was housed using a sterile needle (27G, Exel InternationalMedical Products). The electrical (SL 12C/SB, Plastics One Inc, VA) andoptical (MJP532-46-FC, Princetel Inc, NJ) rotatory joints were integratedto allow both optical fiber and micro-electrodes to rotate following themovements of freely-moving rats during the in vivo experiments. Acustomized ultrasound transducer (5.5 MHz central frequency and 6 mmin diameter, Blatek Inc, PA) was used to record the induced photo-acoustic signals which were then amplified by a pre-amplifier (5073 PR,Olympus, MA) with an amplification magnitude of 39 dB and digitalizedby a 12-bits data acquisition card (NI5124, National Instruments Cor-poration, TX) with a sampling rate of 50 MS/s. LFPs were recorded indifferential mode by a neural recording system (RZ5, Tucker-DavisTechnologies, Inc.) at a sampling rate of 50 kHz with a broad band-pass filter (0.5 Hz–6 kHz).

In order to secure the device, a custom-designed resin holder (Fig. 1b)was attached to the skull via two screws after the surgery which was usedto remove the scalp and drill a hole on the skull. The positions of theultrasound transducer and sterile-needle-housed probe were optimizedusing two high-resolution manipulators in the stereotaxic instrument(David Kopf Instruments, Tujunga, CA). Once optimally placed, the spaceinside the holder was filled with ultrasound gels permitting the trans-mission of high-frequency photoacoustic signals. A thin layer of remov-able epoxy was used to immobilize the transducer and sterile-needle-housed probe inside the holder, and a thick layer of dental acrylic wascarefully applied around the holder to anchor it rigidly on the skull of therat brain. After the experiment, the ultrasound transducer and sterile-needle-housed probe were carefully removed from the holder, cleanedand sterilized for the following experiments.

2.2. Derivation of fractional changes of blood flow and total hemoglobin

Blood flow measures the velocity of red blood cells (RBCs) inside thevessels. The contrast of photoacoustic effect inside capillary comes fromthe absorption of total hemoglobin inside the RBCs. The motion of theRBCs will lead to the shift in frequency of the acoustic waves due tophotoacoustic Doppler Effect described by Equation (1) below (Fanget al., 2007; Yao and Wang, 2010; Yao et al., 2010).

f ¼ cþ vrcþ vs

f0 (1)

where f is the observed frequency of acoustic waves, f0 is the emittedfrequency, c is the velocity of acoustic waves in the medium, vr is the

ted experimental setup. The light path, photoacoustic detection unit and ECoG recordinghe hybrid recording probe. The resin holder was tightly fixed on the mouse skull via twosing removable epoxy, and the rest space inside the holder was filled with ultrasound gels.moving condition.

Fig. 2. The schematic of the setups for phantom experiments. (A) The setup to estimatethe minimum detectable photon density to generate detectable photoacoustic signal forcurrent detection unit. (B) The setup to estimate the relationship between BW andblood flow.

L. Xi et al. NeuroImage 161 (2017) 232–240

speed of the acoustic detector and vs is the speed of the source relative tothe medium. Since the velocity of light is several orders larger than thatof the RBCs inside the capillary, the contribution to the frequency shiftdue to the motion of the light source is neglected. Consequently, thephotoacoustic Doppler shift is determined by Equation (2) (Fang et al.,2007; Yao and Wang, 2010; Yao et al., 2010).

Δf ¼ v cos θc

f0 (2)

where v is the velocity of the RBCs inside the capillary and θ is the anglebetween the direction of blood flow and the propagating direction ofacoustic waves. In our case, we used Doppler bandwidth BW instead ofcentral frequency shift to quantify Doppler Effect. BW is determined bythe two extreme photoacoustic rays of the moving RBCs which arereceived by the edges of the transducer. With known parameters of thetransducer, BW is given by Equation (3) (Fang et al., 2007; Yao andWang, 2010; Yao et al., 2010).

BW ¼ fb � fa ¼ 2f0vcsinðθÞsinðφÞ (3)

where fb and fa are Doppler shift contributed by the projection of trans-verse blood flow on two different lines from focal point to the outmostboundary of the transducer. φ is the aperture angle of the acoustictransducer. As the transducer and optical fiber are fixed onto the skull, θand φ are constants. The relative changes of the velocity of RBCs isproportional to the changes of the Doppler bandwidth that can be esti-mated by the autocorrelation function of sequential time-resolved pho-toacoustic signals along the depth direction generated by each individuallaser pulse. For both in vivo and phantom experiments, all photoacousticsignals were first filtered by a digital bandpass filter (central frequency:5 MHz; �3 dB bandwidth: 1 MHz). Every four sequential depth-resolvedA-line signals were chosen for autocorrelation in order to estimate thedepth-resolved standard deviation of the photoacoustic Doppler spec-trum, which described the relative changes of BW .

The amplitude of the raw photoacoustic signal is proportional to theoptical absorption which mainly comes from deoxyhemoglobin (Hb) andoxyhemoglobin (HbO) in the blood. At the wavelength of 532 nm, theoptical absorptions of Hb and HbO are almost the same. Hence, theamplitude of PA signal is proportional to the concentration of total he-moglobin (HbT) (i.e., the sum of Hb and HbO). In addition, blood vesselsincluding arteries, veins and capillaries will dilate/shrink during thehemodynamic changes such as the increase of HbT and blood flow (Yaoet al., 2015). Furthermore, the detected number of RBCs within a certaintime increases as the blood flow increases (Chen et al., 2010, Chen et al.,2011). Hence, we employed the change of photoacoustic pressure toestimate the relative change of total hemoglobin inside the detectedcapillaries.

The data processing was performed using custom software pro-grammed in MATLAB (Mathworks, Nantucket, MA). For the calculationof total hemoglobin change, the peak-to-peak value of each photo-acoustic signal was calculated and normalized. As each BW was derivedfrom 4 PA A-lines, to match the frequency of blood flow (5 Hz) andimprove the signal-to-noise ratio (SNR) for the hemoglobin measure-ment, the PA intensities were averaged over four sequential A-lines. Toshow the longitudinal change of BW representing the relative dynamicchange of blood flow, we back projected the calculated depth-resolvedsignals within the blood vessel sequentially to form a 2D image.

2.3. Phantom experiments

Fig. 2a shows the concept for phantom experiments to estimate theminimum photon density needed to generate detectable photoacousticsignal using the current detection unit. We simulated the blood vesselwall using a plastic tube with an inner diameter of 100 μm and filled itwith heparinized fresh blood extracted from rats. An objective lens

234

(RMS4�, Thorlabs Inc.) with a NA value of 0.1 in the air was used toconverge the collimated laser beam. The plastic tube was placed at thebottom of a water tank, and the ultrasound transducer and objective lenswere mounted on two high-resolution three-dimensional (3D) stages tooptimize the position. In front of the transducer, a piece of rat skull wasplaced to simulate a setting for an animal experiment. With a pre-selectedinput light energy, we moved the objective in axial direction to achievedifferent excitation photon density and recorded the generatedPA signals.

The experimental setup for estimating the relationship between BWand blood flow is given in Fig. 2b. A polyethylene tube (inter diameter:580 μm, Becton Dickinson and Company, Franklin Lakes, NJ) filled with0.1% solution of silver microspheres (10 μm) was used to mimic thecapillary and the RBCs. The flowwith speeds from 0mm/s to 8mm/s wascontrolled by a 1-mL syringe driven by a syringe pump (KDS 210, KDScientific, Holliston, MA). The optical fiber tip was positioned near theouter wall of the tube with a fixed angle.

2.4. Animals handling

A total of 20 Male Sprague-Dawley rats (Harlan Labs, Indianapolis)weighing 120–250 g were used for the in vivo experiments. Animals werepair housed in a controlled environment (12:12-h light/dark cycle; foodand water ad lib). We performed surgeries on the rat brain before theexperiment. Rats were anesthetized prior to surgery with isoflurane inoxygen administered by inhalation using a precision vaporizer (4% iso-flurane was used for induction followed by 1–2.5% for the remainder ofthe surgery). After the induction with isoflurane, the animals receivedxylazine (10 mg/kg SQ) to smooth the course of the anesthesia andprovide pain relief. After being secured in the stereotactic frame (DavidKopf Instruments, Tujunga, CA), an incision was made along the skin toexpose the skull. A single hole was drilled on the skull (1 mm in diameter)above the motor cortex and the dura was carefully removed. A stainless-steel screw was driven into the bone above the cerebellum and served asground and recording reference. The entire PAI/electrophysiologyapparatus was carefully mounted on the brain following the procedureintroduced in the part of hybrid miniature sensing probe. For the dura-tion of the surgery, body temperature was maintained with an electric

L. Xi et al. NeuroImage 161 (2017) 232–240

heating pad. The surgeries were performed 12 h before the in vivo ex-periments to let the rats fully recover from the anesthetized state. Afterthe experiments, all the rats were sacrificed. All the procedures wereapproved by the University of Electronic Science and Technology ofChina (UESTC) and the Institutional Animal Care and Use Committee(IACUC) at the University of Florida.

2.5. Epilepsy model

Rats were induced to generalized tonic-clonic seizures by intraperi-toneal injection (i.p.) of Pentylenetetrazol (PTZ, Sigma-Aldrich, Inc.), orfocal seizures by focally applied high-concentration PTZ at the site ofoptical fiber and microelectrodes (Zhang et al., 2014).

All anesthetized rats received the same type of surgery as the freely-moving rats. Before the in vivo experiments, the rats were anesthetizedwith urethane (1.25 g/kg, i.p. Sigma-Aldrich, Inc.) instead of isofluranein the surgery. To induce the focal seizure in the detection area, weinjected 5 μL of 400 mg/mL PTZ via micro syringe pump. A total numberof 7 rats were used to conduct the focal seizure study under the anes-thetized state and additional 3 rats received injection of saline as control.In the experiments of evaluating the effect of urethane on the develop-ment of seizures, we applied 80 mg/kg PTZ to 5 rats in freely-movingstate and 160 mg/kg PTZ to other 5 rats in anesthetized state as thecomparison group via intraperitoneal injection.

In the freely-moving studies, the rats were kept in awake and freely-moving state within a cylindrical polymethyl methacrylate (PMMA)

Fig. 3. Results of phantom experiments. (A) Concept of high spatial resolution sensing using UHlight spot in the air versus the distance from the fiber surface, the calculated area of the light spdensity versus the distance from the fiber surface with input pulse energy of 50 nJ. (C) The SNRWhen input photon density decreases to 22 mJ/cm2, the SNR of the photoacoustic signal isphotoacoustic signals at different flow velocities from phantom experiments. BW had positive ans, but deviated thereafter.

235

container with a diameter of 40 cm.

2.6. Neural data recording

In epilepsy studies, synchronized LFPs/optical signals were recordedcontinuously for 30–60 minutes following each PTZ or saline (control)injection. Prior to each injection, 5–10 minutes of data was recorded in aresting state. Power spectral density was calculated using the function offast Fourier transform (FFT) in the platform of MATLAB (Mathworks,Nantucket, MA).

3. Results

3.1. Spatial and temporal resolution

A key advantage of our technique is the high spatiotemporal resolu-tion recording of the dynamics of both blood flow and total hemoglobin.Current optical resolution photoacoustic imaging/sensing techniquesemploy conventional objective lenses to achieve optical focus, which isbulky, heavy and most importantly limited by insufficient penetrationdepth (<1 mm) due to the high scattering and absorption properties ofthe brain tissue. Instead of conventional optical converging lenses, weutilized a UHNA-SMF and optimized the input laser energy to highlyconfine the illumination volume. As shown in Fig. 3a, the divergingvolume angle of the output laser beam from the surface of the fiber corein the air was 42� determined by its NA value (0.35). The core diameter

NA-SMF. (B) The curves indicate the theoretical relationships of the calculated area of theot in the brain tissue versus the distance from the fiber surface, and the calculated photonof the raw photoacoustic signal at different photon densities in the phantom experiment.close to 0 dB which is too weak to be detected. (D) Relative bandwidth broadening ofd nonlinear relationship with the flow speed when the flow speed was smaller than 2 mm/

L. Xi et al. NeuroImage 161 (2017) 232–240

of this optical fiber was 1.8 μm. Based on the given parameters of theUHNA-SMF, the output emission at 532 nm is a Gaussian beam, and wecan estimate the lateral resolution using the full width at half maximum(FWHM) of the light spot. In Fig. 3b, we plot the relationship between thecalculated area of the light spot and the distance from the fiber surface(blue line) in the air. However, for the in vivo experiments, the refractiveindex of the brain tissue is around 1.35 which is much larger than that ofair (1.0) (Sun et al., 2012; Binding et al., 2011). With the increasedrefractive index of the surrounding medium, the diverging volume angleof the laser beam decreases to 30�. The green line in Fig. 3b shows therevised curve with the consideration of the brain tissue. From thefundamental principle of photoacoustic, the intensity of photoacousticsignal is proportional to the illuminated photon intensity in general. Weconducted phantom experiments to simulate a real situation and foundthe minimum photon density needed to generate detectable photo-acoustic signals using the current detection unit. From the results shownin Fig. 3c, when the photon intensity decreased to 22 mJ/cm2, the SNRdecreased to 0 dB which is too weak to be detected. In the in vivo ex-periments of rat brains, considering that the output laser energy is 50 nJ/pulse, when the area of the light spot is equal to 0.232 � 10�5 cm2, thephoton density decreases to 22 mJ/cm2 which is not high enough togenerate detectable acoustic signals using the current detection unit (redline in Fig. 3b). In this scenario, the diameter of the light spot is 17.6 μmand it is 33 μm away from the fiber surface. Hence, the spatial resolutionis variable from 0.9 μm to 8.8 μm.

Different from the definition of optical imaging/sensing, the axialresolution of photoacoustic imaging/sensing is determined by both thebandwidth and central frequency of the transducer (Chen et al., 2010). Inthis device, we used an ultrasound transducer with a central frequency of

Fig. 4. Results of seizure prediction. (A) Changes of blood flow, total hemoglobin and LFP beforeveal the changes of total hemoglobin and BW is proportional to the changes of blood flow. Whecontrol case with the injection with saline, there is no obvious changes in PA signal and BW. (C)in Fig. 4A, and the power spectral density of the LFP data. t1 represents the beginning of chan

236

5 MHz and a -3 dB bandwidth of 4 MHz leading to an axial resolutionof 370 μm.

3.2. Blood flow simulation

According to the phantom results shown in Fig. 3d, we found that BWhad a positive relationship with the flow velocity inside the tube whenthe flow velocity was smaller than 2 mm/s, but deviated thereafter.Considering the basal flow speed in the somatosensory cortex (0.5 ± 0.44to 0.77 ± 0.51 mm/s) (Yao and Wang, 2010), and the resultant changesdue to PTZ-induced seizure onset, the maximum measurable value(2 mm/s) was still sufficient for the in vivo study. However, in some caseswith a blood flow greater than 2 mm/s, we may lose the changes over themaximum measurable velocity.

3.3. In vivo validations

As the monitoring of blood flow and total hemoglobin is often used asdiagnostic tool in order to evaluate the human brain in compromisedstates, we sought to determine the interaction between blood flow, totalhemoglobin and neural activity in an epilepsy model. Importantly, thereis a known increase in the underlying neurological activity that is asso-ciated with the transition from a quiescent to the seizure state. Therefore,several studies were conducted in order to determine the temporalrelationship between large scale neuronal activation and blood flow.

3.3.1. Seizure predictionIn order to determine the temporal relationship between blood flow

and total hemoglobin with obvious LFP activity, focal seizures were

re and after the injection of PTZ and Saline. The fractional changes of PA signal positivelyn the seizure is onset, there is about 55% and 70% increase in PA signal and BW. (B) In theEnlarged details of blood flow, total hemoglobin and LFP data inside the dashed rectangleges in gamma band and t2 is the time point when the obvious ECoG spikes appears.

L. Xi et al. NeuroImage 161 (2017) 232–240

generated in 7 anesthetized rats, and an additional three rats, receivingsaline injection, served as controls. Based on the second section in theMethod part, the PA intensity and BW changes positively reveal thechanges in total hemoglobin and velocity, respectively. Fig. 4a shows thederived fractional changes of PA signal, BW and raw data of LFP for atypical focal seizure onset and a control case. Following PTZ injection, asignificant increase of both total hemoglobin and blood flow wasobserved with the elicited spike and wave discharges. As we expected, inthe control case, no significant increase of total hemoglobin, blood flowor LFP were observed. Fig. 4b shows the changes of total hemoglobin,blood flow and power spectra density of LFP inside the dashed rectanglein Fig. 4a. We found that changes of total hemoglobin and blood flowbegan prior to obvious ECoG spikes in 7/7 rats after the injection of PTZ.The preceding time before the significant gamma power change (t1 inFig. 4b), which was reported to increase prior to the abrupt spikes in LFP(Zhang et al., 2014), was 11.8 s ± 1.6 s. Additionally, the time point whenthe change of total hemoglobin and blood flow firstly occurred was60 s ± 10 s ahead of the beginning of the abrupt ECoG spikes (t2).

3.3.2. Effect of anesthesia on seizure inductionWhile the use of urethane has been reported to block the development

of seizures in a kindled model (Yao et al., 2010), we sought to determineif urethane also blocks PTZ elicited seizures. Therefore, two groups offive rats were given the same dose of PTZ (80 mg/kg, i.p.) either in thepresence or absence of urethane anesthesia. In the freely-moving group(Fig. 5a), the changes of total hemoglobin and blood flow started9.6 s ± 0.9 s prior to the LFP change which is shorter than that of the ratsin anesthetic condition in the focal and generalized seizure models.Following PTZ administration, rodents exhibited four typical behavioralphenomena - freezing, myoclonic twitches, clonic seizures andtonic-clonic seizures (see Video 1). We observed significant changes inboth total hemoglobin and blood flow. The pattern of the ECoG recordingwas different from that in anesthetized group with focal seizure onset.

Supplementary video related to this article can be found at http://dx.doi.org/10.1016/j.neuroimage.2017.08.037.

In the urethane anesthetized group, we did not observe any signifi-cant changes of total hemoglobin, blood flow and LFP when the rats wereinjected with the 80 mg/kg dose of PTZ (Fig. 5b). In order to determine ifthis is a dose dependent phenomenon, we administered an additional80 mg/kg for a total of 160 mg/kg PTZ. Following the higher doseadministration, seizure onset was successfully induced. However, theevoked pattern of LFP was different from the pattern recorded in rats infreely-moving condition (Fig. 5c), but similar to the recording in the

Fig. 5. Fractional changes of PA signal, BW and LFP inside the capillary of the rat with injectioncorresponding functional changes between t1 and t2 after the injection of PTZ. (A, Video 1), 80(C). t1: beginning point of the video; t2: ending point of the video.

237

anesthetized focal seizure model. Another interesting observation is thatthe changes of total hemoglobin and blood flow in the anesthetized ratsare attenuated. For the anesthetized studies, we induced the generalizedseizure two times in the same rat with different doses (80 mg/kg and160 mg/kg). In order to reduce the interaction between two injections,we kept the rats anesthetized for additional 4 h after there was no changein ECoG recording.

In Fig. 6, we show quantitative analysis of the preceding time, base-line of BW, and fractional changes of both PA signal and BW for the rats inanesthetized and freely-moving conditions. We observed that the pre-ceding time (dark gray bar in Fig. 6a) of the hemodynamic changes priorto ECoG recordings for anesthetized and freely-moving cases are11.8 s ± 1.6 s and 9.6 s ± 0.9 s, respectively. The baselines of BW are0.28 ± 0.03 and 0.41 ± 0.02 for anesthetized and freely-moving rats,respectively (light gray bar in Fig. 6a). Fig. 6b shows the fractionalchanges of PA signal (dark gray bar) and BW (dark gray bar) for theanesthetized rats with induction of focal seizure (PA: 0.53 ± 0.07; BW:0.69 ± 0.02), anesthetized rats with induction of generalized seizure (PA:0.39 ± 0.05; BW: 0.44 ± 0.03), and freely-moving rats with induction ofgeneralized seizure (PA: 0.67 ± 0.05; BW: 0.70 ± 0.03).

4. Discussion

The current setup has a variable lateral resolution from 0.9 μm to8.8 μm depending on the distance of the capillary from the fiber tip. Inthe worst scenario the capillary would be 33 μm away from the fiber tip.However, from the vascular anatomy of the rat brain, we note that thereis an extra dense vascular network in the brain, which would prevent theworst scenario from occurring due to the strong optical absorption andscattering properties of blood at 532 nm.

From the phantom evaluation of blood flow estimation, the maximummeasurable velocity is 2 mm/s. Similar to other photoacoustic imaging/sensing of blood flow using Doppler Effect, the upper limit of this methodis determined by the repetition rate of the laser source which determinesthe time interval of two adjacent A-lines and the associated mathematicalderivation (Cao et al., 2017; Fang et al., 2007; Yao and Wang, 2010; Yaoet al., 2010). In a practical situation, when the flow speed is greater than2 mm/s, sequential A-lines are uncorrelated as only one A-line photo-acoustic signal can be recorded when the same particle passes throughthe focus. From the literature, the mean blood flow in a capillary of thebrain is between 0.5 ± 0.44 mm/s and 0.77 ± 0.51 mm/s, and the meanflow velocity is a maximum of 133%–191% of the baseline value whenthe seizure is onset (Kleinfeld et al., 1998; Bode, 1992). Therefore,

of 80 mg/kg PTZ in freely-moving condition. Video 1 presents the activities of the rat andmg/kg PTZ in the anesthetized condition (B), and 160 mg/kg in the anesthetized condition

Fig. 6. Quantitative analysis of the preceding time of hemodynamic changes prior to ECoG recordings (A), baseline of BW (A), and fractional changes of PA (B) signal and BW (B) inanesthetized and freely-moving conditions. Anes.: Anesthetized; F.M.: freely-moving; F: focal seizure; G: generalized seizure.

L. Xi et al. NeuroImage 161 (2017) 232–240

current configuration is still suitable for the proposed study. However, ifwe want to investigate the hemodynamics in large blood vessels such asartery and vein, 2 mm/s is insufficient because their baseline velocity isbetween 3 mm/s and 5 mm/s. Fortunately, this is not the fundamentallimit of this technique. It has been successfully demonstrated in optical-resolution photoacoustic microscopy (ORPAM) that the maximummeasurable velocity is up to 8 mm/s and that it has the potential to reach50 mm/s using a high-repetition rate (e.g., 300 kHz) laser (Wanget al., 2016).

Theoretically, the amplitude of raw photoacoustic signals positivelyreveals the concentration of total hemoglobin in the blood. In addition totheoretical deduction from the principle of photoacoustic effect, severalphotoacoustic studies have shown that photoacoustic intensity is posi-tively related to the concentration of hemoglobin in capillaries, arterialsand veins (Xiang et al., 2013a; Tang et al., 2015, 2016a, 2016b; Yao et al.,2015; Cao et al., 2017) with a fixed photon intensity. However, thereremains a challenge to validate this fact via phantom studies because it isdifficult to simulate the change of total hemoglobin and dilation/constriction of blood vessels using phantoms. In addition, the quantita-tive analysis of the PA baseline is meaningless since the photon intensitychanges in different positions of the target.

It is believed that the hemodynamic changes in the vasculaturenetwork are resulted from synchronized neuronal activities, thusbecoming detectable with a time delay (B�enar et al., 2002). For the firsttime, Makiranta et al. reported that BOLD changes could precede theoccurrence of evoked ECoG spike in a pig model of epilepsy (M€akirantaet al., 2005). Previous studies showed that BOLD changes occurred asearly as 9 s prior to the actual ECoG event in humans with generalizedand focal seizures using ECoG-fMRI (Moeller et al., 2008; Jacobs et al.,2009). In addition, Zhang et al. (2014). and Tang et al. (2015). studiedhemodynamic neurovascular coupling of drug induced generalized epi-lepsy in rats using mountable photoacoustic computed tomography anddiffuse optical tomography, and found that hemodynamic changes pre-ceded the occurrence of ECoG spikes. In our study, the results of theexperiments with induction of focal seizures in 7 rats under anesthetizedcondition indicated that both total hemoglobin and blood flow in capil-lary changed prior to the observation of ECoG spikes. Interestingly,accumulated experimental evidences have shown that the high frequencyoscillations in the gamma and fast gamma ranges of ECoG recording(40–120 Hz) play an important role in the beginning of seizure onset andthat they could be used to precede the neocortical seizures (Niessinget al., 2005; Scheeringa et al., 2011; Staceya et al., 2011). We derived thepower spectral density map of the ECoG signals, and found that thepreceding time of hemodynamic changes prior to significant gammapower change was 11.8 s ± 1.6 s. These early changes of hemodynamicscould be explained as follows: 1) ECoG discharges could be the result ofsuch earlier hemodynamic changes which would reverse the principalunderstanding that neuron activities are responsible for the

238

hemodynamic changes (Hawco et al., 2007); however, there is no directevidence supporting this hypothesis so far; 2) the earlier hemodynamicchanges could be resulted from neuron activities invisible to currentECoG recording and further ECoG studies would be needed to distinguishif these events are invisible to the ECoG as a result of their limitedspatiotemporal resolution or the neuro-physiological characteristics (Taoet al., 2005). Additionally, in our studies, we found that the precedingtime for the rats in freely-moving state was about 2.2 s shorter than thatin anesthetic state, and higher dose of drug were required to inducegeneralized seizures for the rats under anesthetic state. It is believed thatanesthesia has broad effect on brain cells such as changes in neuralprocessing, vascular reactivity, spontaneous neural activity, cerebralenergy metabolism and cerebral blood flow. In neurovascular couplingstudies, anesthetic drugs have suppressive actions on neuron activities,and thus may cause onset time delay of neural response to physiologicalstimulations (Masamoto and Kanno, 2012). In addition, urethane wasreported to suppress the cortical excitability and block the developmentof seizures in a kindled model (Cain et al., 1989). Martin et al. observedreduced hemodynamic responses using simultaneous recording ofevoked neural and vascular responses in rats under anesthesia (urethane,1.25 g/kg) (Martin et al., 2006). Due to the suppressive action of ure-thane on the development of seizures, higher dose of PTZ is requiredwhen the rats are anesthetized using urethane. Interestingly, when theseizure was successfully induced using a higher dose of PTZ, we observedsignificant local potential discharges, but the hemodynamic change wasnot remarkable. The total hemoglobin and blood flow changes of thefreely-moving rats with generalized seizure were nearly 1.7 times of thechanges of the anesthetized ones. Besides the inhibition of urethane onhemodynamic changes (Martin et al., 2006), another possibility is thatthe intensity of the ECoG recording using current system was not pro-portional to the hemodynamic changes. The amplitude of the ECoG spikeis associated with the location of the electrode, number of neurons in theregion of interest, and seizure lesions over the entire brain. However,different from ECoG recording, due to the high spatial resolving ability ofphotoacoustic sensing, we could only observe the hemodynamics withina confined volume.

Based on quantitative analysis of baselines in total hemoglobin andblood flow, we note that the baseline for freely-moving rats is about 1.5times of the baseline for the anesthetized rats. Besides the influence ofanesthesia, the pain-induced stress from the surgery and discomfort inthe restricted space may cause increase in total hemoglobin and bloodflow for freely-moving rats. Furthermore, the latency of seizure onsetpost the injection is different for focal and generalized epilepsy. The focalseizure has a shorter latency compared with the generalized one becausethe concentration of the PTZ used in focal epilepsy model is much higherand the location is almost the same as the position for photo-acoustic detection.

Although the results are exciting, there are still several areas that need

L. Xi et al. NeuroImage 161 (2017) 232–240

to be improved in future.First, the temporal resolution is limited by the repetition rate of the

pulsed laser which is 20 Hz in this study. This limitation, however, can beeasily rectified by employing commercially available, high-repetitionrate lasers (50 kHz–500 kHz) capable of recording at the rate of theLFP (50 kHz). In addition, we realize that current axial resolution(370 μm) is much larger than the actual size of the capillary, whichresulted in an overestimation of the imaged axial size of the capillary. Ahigh-frequency ultrasound transducer could improve the axial resolution;however, since the attenuation of acoustic signals of brain skull and scalpis significantly high for high-frequency acoustic signals, we would needto select a suitable transducer based on the test at an expected penetra-tion depth. For instance, we could use a 50 MHz transducer to detectgenerated photoacoustic signals within 3 mm penetration depth with a15 μm axial resolution, and a 20 MHz transducer for the detection oftargets located within 5–10 mm under the surface with a 75 μm axialresolution. During the onset of seizure, the capillary will dilate andconstrict to allow more flux of red blood cells compared with that in anormal status. Since the capillary size is relatively small, a transducerwith ultra-high center frequency and ultra-wide bandwidth is required tocapture the morphological changes of the capillary. To reduce the severeacoustic attenuation by the skull, a feasible way is to remove or polish theskull to allow the transmission of ultra-high frequency photoacousticsignals before the experiment.

Second, functional imaging/sensing is one of the major advantagesfor photoacoustic detection, because it provides more functional pa-rameters. By using a multi-wavelength strategy, it will be possible toderive absolute concentrations of oxygen hemoglobin and de-oxygenhemoglobin, cerebral metabolic rate of oxygen, and oxygen saturationin the blood vessels (Xi et al., 2014; Huang et al., 2016; Razansky et al.,2009) which can significantly broaden the applications of this technique.For example, it will enable studies of physiological behaviors such aspain, fear conditioning, objective recognition, and social interactions,and of brain disorders such as stroke and Alzheimer's disease.

Third, optogenetics is a powerful tool to investigate brain functions.Our technique can be easily combined with fluorescence detection andoptogenetics using double cladding optical fibers. In addition, the majorlimitation of this technique is a restricted sensing region and lack ofimaging capability. Similar to other optical microscopic models, it ispossible to develop a miniaturized photoacoustic microendoscopic probeusing gradient index (GRIN) lenses.

5. Conclusions

In summary, using a high spatiotemporal resolution miniature devicecapable of recording the capillary-level changes of total hemoglobin andblood flow as well as surrounding neural activities in real-time, we haveprovided a platform by which neural activity and hemodynamics can becorrelated in freely-moving rodents. The potential impact of this tech-nology could span from basic science questions-such as “how do micro-circuit changes alter blood flow?” to studies that seek to bridge BOLDdata collected in humans with neurophysiology data collected in rodents.

Acknowledgments

This work was sponsored by National Natural Science Foundation ofChina (81571722 and 61528401), State International CollaborationProgram from Sichuan (2016HH0019), and startup grant(A03012023601011) from University of Electronic Science and Tech-nology of China.

References

Akemann, W., Mutoh, H., Perron, A., Rossier, J., Knopfel, T., 2010. Imaging brain electricsignals with genetically targeted voltage-sensitive fluorescent proteins. Nat. Methods7, 643–649.

Beard, P., 2011. Biomedical photoacoustic imaging. Interface Focus 1, 602–631.

239

B�enar, C.G., Gross, D.W., Wang, Y., Petre, V., Pike, B., Dubeau, F., Gotman, J., 2002.BOLD response to interictal epileptiform discharges. NeuroImage 17, 1182–1192.

Binding, J., Arous, J.B., L�eger, J.F., Gigan, S., Boccara, C., Bourdieu, L., 2011. Brainrefractive index measured in vivo with high-NA defocus-corrected full-field OCT andconsequences for two-photon microscopy. Opt. Express 19, 4833–4847.

Bode, H., 1992. Intracranial blood flow velocities during seizures and generalizedepileptic discharges. Eur. J. Pediatr. 151, 706–709.

Briers, J.D., 2011. Laser Doppler, speckle and related techniques for blood perfusionmapping and imaging. Physiol. Meas. 22, 35–66.

Buzsaki, G., 2006. Rhythms of the Brain.Cain, D.P., Raithby, A., Corcoran, M.E., 1989. Urethane anesthesia blocks the

development and expression of kindled seizures. Life Sci. 44, 1201–1206.Cao, R., Li, J., Ning, B., Sun, N., Wang, T., Zuo, Z., Hu, S., 2017. Functional and oxygen-

metabolic photoacoustic microscopy of the awake mouse brain. Neuroimage 150,77–87.

Cauli, B., Tong, X.K., Rancillac, A., Serluca, N., Lambolez, B., Rossier, J., Hamel, E., 2004.Cortical GABA interneurons in neurovascular coupling: relays for subcorticalvasoactive pathways. J. Neurosci. 24, 8940–8949.

Chen, Z., Milner, T.E., Dave, D., Nelson, J.S., 1997. Optical Doppler tomographic imagingof fluid flow velocity in highly scattering media. Opt. Lett. 22, 64–66.

Chen, S.L., Ling, T., Huang, S.W., Baac, H.W., Guo, L.J., 2010. Photoacoustic correlationspectroscopy and its application to low-speed flow measurement. Opt. Lett. 35,1200–1202.

Chen, S.L., Xie, Z., Carson, P.L., Wang, X., Guo, L.J., 2011. In vivo flow speedmeasurement of capillaries by photoacoustic correlation spectroscopy. Opt. Lett. 36,4017–4019.

Devor, A., Sakad�zi�c, S., Srinivasan, V.J., Yaseen, M.A., Nizar, K., Saisan, P.A., Tian, P.,Dale, M.A., Vinogradov, S.A., Franceschini, M.A., Boas, D.A., 2012. Frontiers inoptical imaging of cerebral blood flow and metabolism. J. Cereb. Blood Flow. Metab.32, 1259–1276.

Fang, H., Maslov, K., Wang, L.V., 2007. Photoacoustic doppler effect from flowing smalllight-absorbing particles. Phys. Rev. Lett. 99, 184501.

Fox, M.D., Raichle, M.E., 2007. Spontaneous fluctuations in brain activity observed withfunctional magnetic resonance imaging. Nat. Rev. Neurosci. 8, 700–711.

Grinvald, A., Lieke, E., Frostig, R.D., Gilbert, C.D., Wiesel, T.N., 1986. Functionalarchitecture of cortex revealed by optical imaging of intrinsic signals. Nature 324,361–364.

Hawco, C.S., Bagshaw, A.P., Lu, Y., Dubeau, F., Gotman, J., 2007. BOLD changes occurprior to epileptic spikes seen on scalp EEG. Neuroimage 35, 1450–1458.

Huang, D., Swanson, E.A., Lin, C.P., Schuman, J.S., Stinson, W.G., Chang, W., Hee, M.R.,Flotte, T., Gregory, K., Puliafito, C.A., Fujimoto, J.G., 1991. Optical coherencetomography. Science 254, 1178–1181.

Huang, N., Guo, H., Qi, W., Zhang, Z., Rong, J., Yuan, Z., Ge, W., Jiang, H., Xi, L., 2016.Whole-body multispectral photoacoustic imaging of adult zebrafish. Biomed. Opt.Express 7, 3543–3550.

Huchzermeyer, C., Albus, K., Gabriel, H., Ot�ahal, J., Taubenberger, N., Heinemann, U.,Kov�acs, R., Kann, O., 2008. Gamma oscillations and spontaneous network activity inthe hippocampus are highly sensitive to decreases in pO2 and concomitant changes inmitochondrial redox state. J. Neurosci. 28, 1153–1162.

Jacobs, J., LeVan, P., Moeller, F., Boor, R., Stephani, U., Gotmanc, J., Siniatchkin, M.,2009. Hemodynamic changes preceding the interictal EEG spike in patients withfocal epilepsy investigated using simultaneous EEG-fMRI. NeuroImage 45,1220–1231.

Jobsis, F.F., 1977. Noninvasive, infrared monitoring of cerebral and myocardial oxygensufficiency and circulatory parameters. Science 198, 1264–1267.

KimEmail, M.N., Durduran, T., Frangos, S., Edlow, B.L., Buckley, E.M., Moss, H.E.,Zhou, C., Yu, G., Choe, R., Maloney-Wilensky, E., Wolf, R.L., Grady, M.S.,Greenberg, J.H., Levine, J.M., Yodh, A.G., Detre, J.A., Kofke, W.A., Kim, M.N., 2010.Noninvasive measurement of cerebral blood flow and blood oxygenation usingnearinfrared and diffuse correlation spectroscopies in critically brain-injured adults.Neurocrit Care 12, 173–180.

Kleinfeld, D., Mitra, P.P., Helmchen, F., Denk, W., 1998. Fluctuations and stimulus-induced changes in blood flow observed in individual capillaries in layers 2 through 4of rat neocortex. Proc. Natl. Acad. Sci. 95, 15741–15746.

Kuhn, B., Denk, W., Bruno, R.M., 2008. In vivo two-photon voltage-sensitive dye imagingreveals top-down control of cortical layers 1 and 2 during wakefulness. Proc. Natl.Acad. Sci. U. S. A. 105, 7588–7593.

Liao, L.D., Lic, M.L., Lai, H.Y., Shihd, Y.Y.I., Loe, Y.C., Tsang, S., Chao, P.C.P., Lin, C.T.,Jaw, F.S., Chen, Y.Y., 2010. Imaging brain hemodynamic changes during rat forepawelectrical stimulation using functional photoacoustic microscopy. Neuroimage 52,562–570.

Logothetis, N.K., 2002. The neural basis of the blood-oxygen-level-dependent functionalmagnetic resonance imaging signal. Philos. Trans. R. Soc. Lond B Biol. Sci. 357,1003–1037.

Logothetis, N.K., 2003. The underpinnings of the BOLD functional magnetic resonanceimaging signal. J. Neurosci. 23, 3963–3971.

Logothetis, N.K., Pauls, J., Augath, M., Trinath, T., Oeltermann, A., 2001.Neurophysiological investigation of the basis of the fMRI signal. Nature 412,150–157.

M€akiranta, M., Ruohonen, J., Suominen, K., Niinim€aki, J., Sonkaj€arvi, E., Kiviniemi, V.,Sepp€anen, T., Alahuhta, S., J€antti, V., Tervonen, O., 2005. BOLD signal increaseprecedes EEG spike activity—a dynamic penicillin induced focal epilepsy in deepaesthesia. NeuroImage 27, 715–724.

Martin, C., Martindale, J., Berwick, J., Mayhew, J., 2006. Investigating neural-hemodynamic coupling and the hemodynamic response function in the awake rat.Neuroimage 32, 33–48.

L. Xi et al. NeuroImage 161 (2017) 232–240

Masamoto, K., Kanno, I., 2012. Anesthesia and the quantitative evaluation ofneurovascular coupling. J. Cereb. Blood Flow Metabolism 32, 1233–1247.

Moeller, F., Siebner, H.R., Wolff, S., Muhle, H., Boore, R., Granert, O., Jansen, O.,Stephani, U., Siniatchkin, M., 2008. Changes in activity of striato-thalamo-corticalnetwork precede generalized spike wave discharges. NeuroImage 39, 1839–1849.

Niessing, J., Ebisch, B., Schmidt, K.E., Niessing, M., Singer, W., Galuske, R.A.W., 2005.Hemodynamic signals correlate tightly with synchronized gamma oscillations.Science 309, 948–951.

Paulson, O.B., Hasselbalch, S.G., Rostrup, E., Knudsen, G.M., Pelligrino, D., 2010.Cerebral blood flow response to functional activation. J. Cereb. Blood Flow. Metab.30, 2–14.

Razansky, D., Distel, M., Vinegoni, C., Ma, R., Perrimon, N., K€oster, R.W., Ntziachristos,2009. Multispectral opto-acoustic tomography of deep-seared fluorescent proteins invivo. Nat. Photonics 3, 412–417.

Sakad�zi�c, S., Yuan, S., Dilekoz, E., Ruvinskaya, S., Vinogradov, S.A., Ayata, C., Boas, D.A.,2009. Simultaneous imaging of cerebral partial pressure of oxygen and blood flowduring functional activation and cortical spreading depression. Appl. Opt. 48,169–177.

Scheeringa, R., Fries, P., Petersson, K.M., Oostenveld, R., Grothe, I., Norris, D.G.,Hagoort, P., Bastiaansen, M.C.M., 2011. Neuronal dynamics underlying high- andlow-frequency EEG oscillations contribute independently to the human BOLD signal.Neuron 69, 572–583.

Staceya, W., Van Quyenc, M.L., Mormannd, F., 2011. Andreas Schulze-Bonhagee, what isthe present-day EEG evidence for a preictal state? Epilepsy Res. 97, 243–251.

Sun, J., Lee, S.J., Wu, L., Sarntinoranont, M., Xie, H., 2012. Refractive index measurementof acute rat brain tissue slices using optical coherence tomography. Opt. Express 20,1084–1095.

Tang, J., Xi, L., Zhou, J., Huang, H., Zhang, T., Carney, P.R., Jiang, H., 2015. Noninvasivehigh-speed photoacoustic tomography of cerebral hemodynamics in awake-movingrats. J. Cereb. Blood Flow. Metab. 35, 1224–1232.

Tang, J., Coleman, J.E., Dai, X., Jiang, H., 2016. Wearable 3-D photoacoustic tomographyfor functional brain imaging in behaving rats. Sci. Rep. 6, 25470.

Tang, J., Dai, X., Jiang, H., 2016. Wearable scanning photoacoustic brain imaging inbehaving rats. J. Biophot. 9, 570–575.

240

Tao, J.X., Ray, A., Hawes-Ebersole, S., Ebersole, J.S., 2005. Intracranial EEG substrates ofscalp EEG interictal spikes. Epilepsia 46, 669–676.

Wang, L.V., Hu, S., 2012. Photoacoustic tomography: in vivo imaging from organelles toorgans. Science 335, 1458–1462.

Wang, X., Pang, Y., Ku, G., Xie, X., Stoica, G., Wang, L.V., 2013. Noninvasive laser-induced photoacoustic tomography for structural and functional in vivo imaging ofthe brain. Nat. Biotechnol. 21, 803–806.

Wang, T., Sun, N., Cao, R., Ning, B., Chen, R., Zhou, Q., Hu, S., 2016. Multiparametricphotoacoustic microscopy of the mouse brain with 300-kHz A-line rate. Neurophoton3, 045006.

Xi, L., Grobmyer, S.R., Wu, L., Chen, R., Zhou, G., Gutwein, L.G., Sun, J., Liao, W.,Zhou, Q., Xie, H., Jiang, H., 2012. Evaluation of breast tumor margins in vivo withintraoperative photoacoustic imaging. Opt. Express 20, 8726–8731.

Xi, L., Grobmyer, S., Zhou, G., Qian, W., Yang, L., Jiang, H., 2014. Molecularphotoacoustic tomography of breast cancer using receptor targeted magnetic ironoxide nanoparticles as contrast agents. J. Biophot. 7, 401–409.

Xiang, L., Jia, L., Zhang, T., Wang, B., Yang, J., Zhang, Q., Jiang, M.S., Zhou, J.,Carney, P.R., Jiang, H., 2013. Noninvasive real time tomographic imaging ofepileptic foci and networks. Neuroimage 66, 240–248.

Xiang, L., Wang, B., Ji, L., Jiang, H., 2013. 4-D photoacoustic tomography. Sci. Rep. 3,1113.

Yao, J., Wang, L.V., 2010. Transverse flow imaging based on photoacoustic Dopplerbandwidth broadening. J. Biomed. Opts. 15, 021304.

Yao, J., Maslov, K.I., Shi, Y., Taber, L.A., Wang, L.V., 2010. In vivo photoacoustic imagingof transverse blood flow by using Doppler broadening of bandwidth. Opt. Lett. 35,1419–1421.

Yao, J., Xia, J., Maslova, K.I., Nasiriavanakia, M., Tsytsarevb, V., Demchenkoc, A.V.,Wang, L.V., 2013. Noninvasive photoacoustic computed tomography of mouse brainmetabolism in vivo. Neuroimage 64, 257–266.

Yao, J., Wang, L., Yang, J.M., Maslov, K.I., Wong, T.T.W., Li, L., Huang, C.H., Zou, J.,Wang, L.V., 2015. High-speed label-free functional photoacoustic microscopy ofmouse brain in action. Nat. Methods 12, 407–410.

Zhang, T., Zhou, J., Jiang, R., Yang, H., Carney, P.R., Jiang, H., 2014. Pre-seizure stateidentified by diffuse optical tomography. Sci. Rep. 4, 3798.