Transcranial real-time in vivo recording of electrophysiological neural activity in the rodent brain with near-infrared photoacoustic voltage-sensitive dye imaging

a Whiting School of Engineering, Johns Hopkins University, Baltimore, MD 21218, USA b Department of Neurology and Neurosurgery, Hugo W. Moser Research Institute at Kennedy Krieger, Johns Hopkins Medical Institutions, Baltimore, MD 21205, USA c Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins Medical Institutions, Baltimore, MD 21287, USA d Department of Neuroscience, University of Copenhagen, Copenhagen 2200, Denmark e Department of Cell Biology, University of Connecticut Health, Farmington, CT 06030, USA f Solomon H. Snyder Department of Neuroscience, Johns Hopkins Medical Institutions, Baltimore, MD 21205, USA g Department of Psychiatry and Behavioral Sciences, Johns Hopkins Medical Institutions, Baltimore, MD 21287, USA h Department of Neurology, Johns Hopkins Medical Institutions, Baltimore, MD 21205, USA


Introduction
The quantification and monitoring of brain function is a major goal of neuroscience and research into the underlying mechanisms of the working brain. [1][2][3][4][5] Towards this objective, several modalities have been introduced for the purpose of appropriate neuroimaging; however, existing methods have limitations. Positron emission tomography (PET) provides high molecular resolution and pharmacological specificity, but suffers from low spatial and temporal resolution. [6][7][8] Functional magnetic resonance imaging (fMRI) provides higher spatial resolution of brain activity; however, the record is a complex blood-oxygenation level dependent (BOLD) signal with comparatively low temporal resolution and uncertain interpretation. 9,10 Optical imaging approaches have been used to monitor the brain function of small animals but have limited dynamic ranges and cover only superficial tissue depths because of light scattering and absorbance during penetration of biological tissue in vivo. 11,12 These approaches require invasive craniotomy, with problematic long-term consequences such as dural regrowth, greater likelihood of inflammatory cascade initiation, and lack of practicality of translation to non-human primate and ultimately to human studies, including those for neuropsychiatric disorders. 13 In addition, real-time imaging simultaneously with deep penetration has not been demonstrated. Near-infrared spectroscopy (NIRS) noninvasively monitors brain function in real-time (~ 1ms) for deep biological tissues (~several mm), but suffers from poor spatial resolution (~ 1 cm) at those depths. 14,15 Therefore, minimally-invasive monitoring of electrophysiological brain activities in realtime remains a task at hand in neuroimaging, with the aim to quantify brain functions at high spatial resolution in the depths of brain tissue, without need for invasive craniotomy.
To overcome the current challenges, photoacoustic (PA) imaging has been investigated as a promising hybrid modality that provides the molecular contrast of brain function with acoustic transcranial penetration and high spatial resolution. 16,17 In PA imaging, radio-frequency (RF) acoustic pressure is generated, depending on the thermo-elastic property and light absorbance of a target illuminated by pulsed laser, and it is detected by an ultrasound transducer. Based on this mechanism, several studies have presented the capability of transcranial PA imaging. [18][19][20] Additionally, several PA approaches have been recently applied to detect electrophysiological brain activities in both tomographic and microscopic imaging modes;  22 However, these studies used voltage sensing in the visible spectral range (488 nm and 530 nm for GCaMP5G; 500 nm and 570 nm for DPA), which is not suitable for recording deep tissue neural activity because of the significant optical attenuation by blood.
Here, we present transcranial recording of electrophysiological neural activity in vivo with near-infrared PA voltage-sensitive dye (VSD) imaging during chemoconvulsant seizures in the rat brain with intact scalp. As a step towards minimally-invasive external imaging in primates and human brains, the results demonstrate that PA imaging of fluorescence quenching-based VSD is a promising approach to the recording deep brain activities in rat brain, without need for craniotomy.

Methods
Photoacoustic imaging setup. For the recording of electrophysiological brain activities in vivo, real-time PA data acquisition system was built; an ultrasound research system was utilized that consisted of ultrasound linear array transducer (L14-5/38) connected to a real-time data acquisition system (SonixDAQ, Ultrasonix Medical Corp., Canada). The center frequency of the transducer was 7.1 MHz to obtain the sensing depth for neural activities in a deep brain tissue. To induce the PA signals, pulsed laser light was generated by a second-harmonic (532 nm) Nd:YAG laser pumping an optical parametric oscillator (OPO) system (Phocus Inline, Opotek Inc., USA). The tunable range of the laser system was 690-900 nm and the maximum pulse repetition frequency was 20 Hz.
The laser pulse was delivered into the probe through bifurcated fiber optic bundles, each 40 mm long and 0.88 mm wide (Fig. 1a). The customized, 3-D printed shell fixes the PA probe between the outlets of the bifurcated fiber optic bundles for evenly distributed laser energy density in lateral direction on rat scalp surface. The alignment of outlets was focused specifically at 20 mm depth. The PA probe was located at around the Interaural 11.2 mm and Bregma 2.2 mm to obtain the cross-section of motor cortexes (Fig. 1c). The distance between PA probe and rat skin surface was 20 mm, and the resultant energy density was at 3.5 mJ/cm 2 ,which is far below the maximum permissible exposure (MPE) of skin to laser radiation by the ANSI safety standards. 23 A wavelength of 790 nm was used, at which the light energy was sufficiently absorbed by the near-infrared VSD, i.e., IR780. Probing at this wavelength avoided the undesired time-variant change of oxygen saturation, being at the isosbestic point of Hb and HbO 2 absorption spectra in the in vivo setup (Fig. 1b). Fig. 1c presents a representative crosssectional PA image of a rat brain. The outlines for the brain and motor cortex were drawn based on the rat brain atlas 24 (Fig. 1c). We conducted the in vivo experiments according to the protocol shown in Fig. 1d, with a design composed of seizure, control, and negative control groups. Seizure was not induced in the control group subject to VSD + Lexiscan administration. The negative control group served mainly to monitor the hemodynamic change induced by Lexiscan injection and seizure induction without VSD administration.
Fluorescence quenching-based near-infrared voltage-sensitive dye. Several cyanine VSDs have been proposed as markers for real-time electrical signal detection 25 , and applied for optical imaging of the mitochondrial membrane potential in tumors 26 and fluorescence tracking of electrical signal propagation on a heart. 27 Recently we presented the mechanism of action of a cyanine VSD on the lipid vesicle model. 28 The discussed mechanism of VSD proposes a suppressive PA contrast when neuronal depolarization occurs, while yielding an enhanced contrast for fluorescence. In the present proof-of-principle study, we used the fluorescence quenching-based nearinfrared cyanine VSD, IR780 perchlorate (576409, Sigma-Aldrich Co. LLC, MO, United States) with the analogous chemical structure of PAVSD800-2 in our previous study. 28 This VSD yields fluorescence emission leading to a reciprocal PA contrast with nonradiative relaxation of absorbed energy. μ L, and 10 μ L, respectively), while maintaining the Na+ buffer. We expected the logarithmic change in membrane potential levels to the upper estimation at 100-fold K+ gradient (i.e., -83 mV, -102mV, and -120 mV based on the Nernst equation 28,29 ). This will yield a corresponding suppressive PA intensity change. The hair was shaved from the scalp of each rat for improved optical/acoustic coupling for transcranial PA recording. The head of the anesthetized rat was fixed to a stable position using a standard stereotaxic device. This fixation procedure was required to prevent any unpredictable movement during PA recording of neural activities. Pharmacological treatment for VSD delivery into blood-brain-barrier. The lumen of the brain microvasculature consists of brain endothelial cells, and the blood-brain barrier (BBB) is comprised of their tight junctions to control the chemical exchange between neural cells and cerebral nervous system (CNS). In this study, the penetration through BBB were achieved with a pharmacological method using FDA-approved regadenoson Criteria for selecting region-of-interest. The regions-of-interest (ROI) were selected from left and right motor cortices in a PA image. In PA image, the signals from SSS and SCV would be higher than those in brain tissue region -Brain tissue region naturally contains less blood quantity than those at SSS and SCV since the blood supply is performed by capillaries, not any large vasculatures. Therefore, we used our criteria to select regions-of-interest which can neglect the influence of blood signal at SSS and SCV into cortical regions: (1) The minimal size of the ROIs on each hemisphere was set to 1.86 x 1.54 mm 2 within the motor cortex region whose overall dimension is approximately 3 x 2.5 mm 2 based on the anatomy of the rat brain atlas shown in Fig.   1c. 24 (2) The positions of ROIs were at least 2.75 mm below the scalp surface to avoid the regions of several other layers covering a brain, i.e., periosteum, skull, dura mater, arachnoid, subarachnoid space, and pia mater. 35 The distance is much farther than the axial spatial resolution given by ultrasound transducer (i.e., 479.5 µm of mean FWHM at the depth-of-interest, Fig. 3), thereby the VSD sensing at brain cortical regions could be conducted without any interferences from the superficial dominant clutter regions of SSS and SCV.
(3) Skin surface information may not be sufficient to decide the ROI positions as the melanin contents, a major absorber in scalp, has decreasing absorbance as the wavelength become longer 36 ; The 790-nm wavelength used have only around 24% of extinction coefficient compared to those frequently used by other groups for brain imaging all at visible light range (e.g., 500 nm and 570 nm 22 , 532 nm 37 , 488 nm and 530 nm 21 ). Therefore, we are also considering the relative distance from the superior sagittal sinus for ROI positioning. The relative location of SSS and brain tissue involves four layers with different thicknesses. The SSS is in the middle of dura mater (300 μm) which is on top of the arachnoid (75 μm), subarachnoid space (750 μm), and pia mater (75 μm) layers covering a brain. 35 Therefore, the expectable thickness between SSS and brain is 1,050 μm, and motor cortex is extended down to 3-4 mm from the brain tissue surface (bregma 2.2 -3.2 mm, interaural 11.2 -12.2 mm 38 ). From this justification, the substantial overlap between our ROIs and motor cortex could be ensured since our ROIs are also posed at 2 -4 mm below the SSS location where (x, z) is the lateral and axial Cartesian coordinates in the PA image; S SSS is the PA intensity at SSS (x = 0; z = 0); z is the depth index below SSS region in mm scale; S signal (z) is the PA intensity measured from the rat brain image; S noise (z) is the PA intensity measured from the noise image without any absorber in the identical field-of-view size (in the water-filled tank); and σ noise (z) is the standard deviation measured from the noise image at each depth index. The maximal CNR values was projected in lateral direction with 10-mm width centered at SSS (e.g., blue dotted box in Fig. 3d). The photobleaching in vivo was also evaluated. From the ROIs selected at motor cortex region (e.g., white dotted boxes in Fig. 3d), the PA intensity was sequentially measured over time, and normalized with the base intensity at the time range from 0 min to 1 min. Note that the control group (N = 3) was used for these investigations on CNR over the depthof-interest and photo-bleaching over time.
EEG validation of neural seizure activity. To obtain the EEG records of electrical spike discharges that originated from motor cortex, sub-dermal scalp EEG recording electrodes were located at the corresponding locations on motor cortex (Xs in Fig. 9a), the schematic of the rat cranium (three electrodes, 1 recording and 1 reference over motor cortex, 1 ground electrode over rostrum). The EEG signal at motor cortex was recorded with the identical preparation procedures in PA imaging including animal preparation, administration of IR780, Lexiscan, and PTZ, time duration for recording, and interval between sequences in the protocol. Data acquisition was done using  9a). Increasing concentrations of VSD were tested in the same rat at temporally spaced time-points. Rats were anesthetized with IP injection to ketamine/xylazine and a cranial window made over the right motor cortex. After recording a baseline EEG in the rat for 10-min duration with the craniotomy, the follow-on EEG recording continued to record EEG following application of increasing concentrations of vehicle alone and VSD + vehicle for the same duration of EEG recordings (i.e.; 10 min) allowing comparisons of EEG responses to each increasing gradient of VSD on cortical activity as compared to baseline EEG signature in the same rat. depth from the SSS position. This is sufficient imaging depth since the brain surface starts from 1 mm depth from the SSS position, and 4-mm of brain tissue region can be imaged with the imaging system used. We also evaluated the photo-bleaching over time with the given energy density (i.e., 3.5 mJ/cm 2 ). Fig. 3e present the normalized PA intensity over time at our ROI for neural activity detection. Note that the control group was used for this investigation (N = 3). No significant decrease in PA intensity was found at cortical region with the given energy density. literature. 28 The median value in the estimated quantum yield range for each K+ gradient level presents a proportionally-increasing trend as depolarized. Note that the non-specific quantum yield at 25-fold K+ gradient is due to a limited sensitivity to differentiate the subtle membrane potential variation -The specificity of the estimation becomes proportionally improved as more K+ gradient is given.

Results
With the confirmation from the in vitro lipid vesicle model, we conducted the in vivo validation for transcranial sensing of electrophysiological neural activity in the rat brain according to the protocol shown in Fig. 1d. The PA probe was located on the crosssection at around the Interaural 11.2 mm, Bregma 2.2 mm to monitor the PA signal change originating from motor cortex (Fig. 1c) 24 Fig. 5a, significant differences between seizure and control groups were also recorded (Fig. 5b). Fig. 5c shows the fractional change of the neural activity index measured from motor cortexes We validated the chemoconvulsant-induced seizure activity in the in vivo protocol with EEG recording. Using a well-established model of chemoconvulsant-induced status epilepticus, we replicated the classic evolution of chemoconvulsant-induced status epilepticus using PTZ (Fig. 7). 30 These evolutions as related to bursts of synchronized neural activity in vivo were assessed in two similar experimental protocols mirrored for the EEG and PA experiments. We recorded vEEGs of seizure inductions using PTZ (45mg/kg IP injections) in anesthetized rats. EEG baseline recording continued until a stable seizure induction profile (i.e., continuous burst discharges indicating synchronized neuronal depolarization-related action potentials) was recorded using subdermal EEG scalp electrodes. The seizure activity in EEG was associated with tonicclonic movements in the fore-and hind-limbs of the anesthetized rats, indicating motor cortex involvement (Video 4) recorded on synchronous video during EEG acquisition.
The PTZ evolution of status on EEG did not alter with IV VSD treatment.

Discussion
Here, we present a transcranial PA recording of electrophysiological neural activity in vivo using near-infrared VSD for chemoconvulsant seizure in rat brain. In the lipid vesicle phantom experiment, the near-infrared VSD, IR780, clearly revealed the signature of the VSD mechanism in polarization/depolarization events induced by valinomycin and gramicidin (Fig. 4). Based on the validated VSD, the in vivo validation study demonstrated that the global seizure activity of the motor cortex was clearly differentiated from the activities of the control and negative control groups (Fig. 5, Video 5, 6, 7). The pharmacological enhancement of VSD delivery into the cortical region by increased permeability of the BBB was confirmed by the histopathological microscopic validation (Fig. 6). The results also strongly agreed with the electrophysiological activities observed by EEG measurement, with an identical experimental setup and protocol (Fig. 7). Our research is intentionally designed to use a low-frequency clinical ultrasound imaging transducer (center frequency, f c = 7.1 MHz), which allows totally non-intrusive imaging without any additional invasive procedures such as scalp removal, skull thinning, or craniotomy usually required for high-frequency imaging 22,38 and does not require any sophisticated hardware configuration for narrower isotropic spatial resolution. 21 These previous methods obliged to remove scalp and skull of rat or to use very small-scale subject (i.e., zebrafish). Also, our cortex-of-interest, the motor cortex, is widely extended at frontal brain region down to 5 mm depth (bregma 4.2 mm to bregma -0.92 mm) and up to 5-mm width at each hemisphere (at bregma 3.2 mm), so the cortical region can be monitored with currently-available spatial resolution and sensing depth (Fig. 3). No photo-bleaching effect was identified at the ROI selected.
From this specification, the cortical-scale neural depolarization events evoked by chemoconvulsant seizure could be identified with high contrast resolution.
The potentially confounding factors in the experimental setup and protocol employed need to be carefully considered and eliminated. The change in cerebral blood volume (CBV) during chemoconvulsant seizure can generate fluctuations of PA intensity over time that can be misinterpreted as the suppressive VSD response. [41][42][43] To address this concern, we adjusted two considerations in the in vivo protocol and analysis: (1) we allocated 5-10 min of the time duration for hemodynamic stabilization before collecting the PA data, and (2) normalized the STFT spectrogram in both the frequency and time dimensions. Zhang et al. suggested that the total hemoglobin began to change in the pre-ictal period and remained stable after the initiation of tonic-clonic seizure, and the time length from PTZ injection to seizure onset was ~2 min on average, 44 but it was sufficiently covered by our stabilization period in the in vivo protocol. The neural activity map could be stabilized with respect to the CBV change, because the bias on the STFT spectrogram could be rejected during normalization procedures. The negative control group in our in vivo protocol is mainly served to test whether these considerations successfully would work. The PA data obtained without any VSD administration went through an identical analysis method to monitor the chemoconvulsant variation of total hemoglobin concentration and capillary CBV. As shown in Fig. 6a (right column), comparable neural activity was obtained between the baseline and seizure phases in the negative control group, and there was no significant gradual change of hemodynamics over time, demonstrating that the hemodynamic interferences were successfully rejected. Moreover, instantaneous blood flow perturbation due to heart beating would not affect the results, as every individual PA frame was compounded for two seconds that include 11-16 heart cycles of a rat (typically 5.5-8 beats per second).
The stability of stereotaxic fixation against the induced motor seizure was also investigated. The counter-hypothesis of this concern was an abrupt disorientation of rat brain due to motor seizure that will induce instantaneous decorrelation between adjacent PA frames. Also, based on the behavioral observation during seizure (Video 4), we anticipated the decorrelation within a sub-second time scale, if it happened. For these hypotheses, we calculated the cross-correlation maps throughout PA frames obtained for 8 minutes (1920 frames, 240 frames/min). Three different time intervals were tested: 0.25 sec, 0.5 sec and 1 sec, which respectively correspond to 4, 2 and 1 frame intervals. For each interval, the minimal correlation projection (MCP) map was composed by finding the minimal value per pixel in temporal direction of the entire stack (Fig. 8). The PA frames with seizure indicated no significant decorrelation between adjacent PA frames compared to those obtained without seizure. Therefore, the interference by motor seizure could be rejected as potential cause of artifacts in the results.
Toxic CNS effects of VSD is another factor that can alter brain activity. We tested the protocols described in Fig. 1d with varying VSD concentration in rats as a direct application to the cortex (Fig. 9). Results for VSD IR780 with cortical application with cranial windows used in six male rats yielded reliable and reproducible EEG signatures to swell and bleb. 45,46 We also found that recent PA neuroimaging researches presented their electrophysiological contrast on HEK-293 cells using KCl administration. 22 Though it enabled an artificial level of membrane potential (i.e., 0 -100 mV), the employed extracellular KCl concentration was much higher (i.e., 12.7, 61.4, 134.8, and 296.0 mM) than the range reported for neuronal injury and death (~ 50 -90 mM 45,46 ). In addition, we do not have the sensitivity to measure PA signals from single patched neurons due to the limited spatial resolution of the commercially-available linear array transducer used for non-invasive deep brain imaging (Figure 3a). In this regard, we believe that the presented lipid vesicle model is the most reliable methodology to characterize our fluorescence quenching yield-changing VSD distributed in both intracellular and extracellular regions. Furthermore, we have good control of the amplitude of the membrane potential across the vesicle membrane by use of a K + diffusion potential mediated by the K + -specific ionophore valinomycin and reversed by the non-selective ionophore gramicidin ( Figure 4). Also, there have been several lipid models (e.g., monolayers, liposome, vesicles, and planar bilayers) as stable methodologies in physicochemical and biophysical studies to better understand the interaction of exogenous molecules with biological membranes. [47][48][49] Here, we demonstrated the first proof-of-concept of transcranial PA sensing of neural activity with near-infrared VSD, using a chemoconvulsant seizure model of the rat brain. We plan a number of follow-up efforts to further advance the concept. For instance, the monitoring of seizure activity involved in the hippocampus would be our next step of experimental validation as the region is usually the strongest source of seizure-related activity. The step-by-step approach is required along with the parallel efforts to overcome the obvious technical challenges related to the optical and acoustic attenuations. We have been successfully using 3.5mJ/cm 2 as the energy density to obtain sufficient sensitivity while not inducing photo-bleaching in our measurements, but the confidence level of our transcranial measurement is extended down to 5 mm depth (Fig. 3d,e) , which is not sufficient for hippocampus imaging (5 -7 mm depth at bregma -1.88). We will further investigate the optimal energy density to increase the penetration depth. Also, contrast enhancing algorithms such as adaptive beamforming could be the solution for this approach.
In addition, we expect that improved signal processing for extracting neural activity from the ubiquitous blood context will enable better characterization of brain function.
The present in vivo experiments confirmed the possibility of background suppression, but still have artifacts in the sensing area (Baselines in Fig. 5a). Enhanced signal processing and/or use of multi-spectral wavelengths may allow significantly improved spectral unmixing of electrophysiological activities in the brain, leading to development of novel quantitative metrics for real-time brain characterization. Also, there is an abnormal appearance of seizure activity more on a single-side hemisphere (Fig. 5a), and it arise a question of biological factor on the VSD delivery. In our animal preparation, we are using a jugular vein port for VSD + Lexiscan administration. The injected VSD + Lexiscan directly hits a single hemisphere first and perfused through body to reach out to another hemisphere. We are concerned about varying efficiency of BBB opening by Lexiscan concentration, though the amount of VSD would be eventually be equal for both hemispheres after complete body perfusion. To address to this concern, we will change our VSD administration method to tail-vein IV injection for the complete systematic VSD delivery, and we will also conduct comprehensive investigation on the pharmacological mechanism of BBB opening by Lexiscan. These results will be also presented in the following papers.
Having isotropic resolution with 2-D PA probe would be also an interesting direction to pursue as follow up to the present work. The use of 2-D PA probe would not only allow real-time volumetric information, but also enable the suppression of off-axis interference. Even though we presented that neural activity can be successfully discerned with current 1-D PA probe, its sensitivity might be affected by off-axis interferences especially from the elevation direction because of the limited acoustic lens focusing at a fixed depth. The neuroimaging using 2-D PA probe would reject those interferences by the advanced electrical beamforming capability in axial, lateral, and elevation directions.
The use of localized, non-invasive neural stimulation will allow us to substantially expand our perspectives in real-time brain response to the external stimuli in a totally non-intrusive way. 50 In particular, we envisage that the glutamate receptor modulation using NMDA and/or direct stimulation on sensory cortex can advance the proposed VSD imaging technology. We have recently obtained promising results in NMDAevoked cortical activity and physiological neural activity imaging at visual cortex with visible light stimulation. 51 However, these results will be presented in the separated publication as its scope is beyond this publication. On the other hand, the integration with ultrasound neuromodulation may have a huge impact on the neuroscientific and clinical efforts by enabling the breakthrough beyond the passive brain investigation, while allowing additional benefits on non-pharmacological BBB opening. 52,53 The toxicity and biodegradability of IR780 perchlorate dye is another important issue that deserves further evaluation; Even though there have been no long-term and comprehensive toxicity study, we believe that the metabolic products of IR780 should be very similar to ICG, FDA-approved near-infrared cyanine dye, since its absorptive contrast is basically based on same chromophore. This strongly suggest its biocompatibility of our VSD mechanism. We will prove our hypothesis in our future works.
Furthermore, the neural sensing speed should be further improved. Current PA sensing speed is limited to 4 frames per second to obtain sufficient signal sensitivity in the deep brain cortex region with the current laser excitation scheme (20 Hz of pulse repetition rate, 3.5 mJ/cm 2 ). This speed may limit its applicability in research, as it is well known that resting electrophysiological neural activity ranges up to several tens of Hz (e.g., delta: 1-4 Hz; theta: 4-8 Hz; alpha: 8-13 Hz; beta: 13-30 Hz; gamma: 30-50 Hz). 54 We will attempt to resolve the tradeoff in sensitivity by having ~100 Hz of sensing speed. Successful investigation will substantially increase the capability of the proposed approach for understanding brain function in real-time.

Disclosures
No conflicts of interest, financial or otherwise, are declared by the authors.   Note that the outlines for brain and motor cortex in Fig. 1c was drawn based on the rat brain atlas (Interaural 11.2 mm, Bregma 2.2 mm) 24 The success of seizure induction was confirmed by tonic-clonic movements in the fore and hind-limbs of the anesthetized rat during the experiments (Video 4, MPEG, 1.8MB).    28 The median values were presented in the estimated quantum yield range for each K+ gradient level.  6. In vivo validation of VSD delivery to a rat brain: (a) Fractional changes of PA intensity depending on Adenosine receptor signaling modulation using intravenous regadenoson administration. Each PA sequences, i.e., VSD only and VSD + regadenoson, was measured from the brain tissue region (3 mm below the skin surface), and projected during last 2 min (8-10 min, 480 times points). (b) histopathological analysis on negative control (VSD-, Lexiscan-), control (VSD+, Lexiscan-), and BBB opening (VSD+, Lexiscan+) groups.    9. VSD toxicity study using EEG recordings during direct cortical applications using a cranial window in rats. (a) Schematic of experimental protocol. A rectangular cranial window drilled under anesthesia overlying unilateral motor cortex. Duramater was kept intact. Following craniotomy, a small window was made in duramater without traversing blood vessels. (b) EEG recording of baseline brain activity under anesthesia was followed by using a hamilton micro syringe to apply increasing concentrations of IR780 directly to the cortical surface via window made in duramater. Base EEG remained unaltered at lower concentrations but showed significant background suppression after applying a 100X solution. This study allowed us to determine the concentration of IR780 10X for all PA experiments. (c) EEG power spectral quantification for every 10-sec epoch of EEG over the duration of the recording confirmed EEG suppression with the 100X dose.