Evaluation of a transparent cranial implant as a permanent window for cerebral blood flow imaging

Laser speckle imaging (LSI) of mouse cerebral blood flow was compared through a transparent nanocrystalline yttria-stabilized zirconia (nc-YSZ) cranial implant over time (at days 0, 14, and 28, n = 3 mice), and vs. LSI through native skull (at day 60, n = 1 mouse). The average sharpness of imaged vessels was found to remain stable, with relative change in sharpness under 7.69% ± 1.2% over 28 days. Through-implant images of vessels at day 60 appeared sharper and smaller on average, with microvessels clearly visible, compared to through-skull images where vessels appeared blurred and distorted. These results suggest that long-term imaging through this implant is feasible. © 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

The first application of LSI to cerebral blood flow imaging [8] used LSI spatial contrast analysis [7]. Subsequently, LSI temporal contrast was developed [9,10], which allowed for reduced static noise and improved spatial resolution compared to LSI spatial contrast analysis, at the expense of temporal resolution. In LSI temporal contrast imaging, a coherent source is directed onto a roughened surface, and the scattered light produces a random interference pattern called speckle [11,12]. Movement of scattering particles within the random medium causes phase shifts in the scattered light and thus changes the random interference pattern, producing temporal fluctuations in the speckle pattern [11,12]. When this speckle pattern is imaged with a camera of limited exposure time, these temporal fluctuations in the speckle pattern corresponding to dynamic regions of the sample appear smoother than the surrounding static regions [13,22]. The degree of smoothness is measured as speckle contrast, and previous studies applying LSI to cerebral blood flow have shown that this speckle contrast correlates with blood flow velocity [11,13].
This improved method of LSI temporal contrast imaging was applied to intact rat skull [14], but the poor transparency of the skull resulted in limited spatial resolution of the underlying blood flow. Previous studies have addressed this challenge in animal models by thinning [15] or polishing [16] the skull, optically clearing the skull [17], or by replacing a portion of skull with a transparent window made from glass or PDMS [23][24][25], which are powerful research techniques but which are not appropriate for human application as permanent cranial implants for patients. Skull thinning and polishing techniques affects

Animals
All experimen University of The N = 4 8housed under water.

Surgical
Craniectomy anesthetized w (80/10 mg/kg from the scalp frames to imm and the surgic times). Implan A sagittal the skull. Per surgical drill with dimensio craniectomy d four corners o exposure for 2 eventual hum Imaging w scalp was stil days (for Mo mice were ane Periosteum ov conducted in amounts of sa imaging proce was applied postoperative anesthesia. A surgery. Fig Fig. 2(b).

Experime
In this study, mice 1-3, wi implantation, of view was e Mouse 4 at 60 different expo in Fig. 2 where I x,y (n) is the intensity at pixel (x,y) in the nth image, N is the number of images acquired, and I x,y  is the mean value of intensity at pixel (x,y) over the N images.
We assessed the quality of the speckle contrast images over time in terms of signal to noise ratio (SNR) and vessel sharpness. To quantify signal to noise ratio for each exposure time and time point, the contrast intensity profile along a vertical line (across the blood vessels) was considered. The profile lines were chosen for each mouse such that 3 to 4 vessels were intersected, and remained the same between the time points and the exposure times. Figure 3(b) shows an example of the contrast intensity profile for 6 ms exposure time at day 0, 14 and 28 for Mouse 3. Equation (2) shows how SNR values were calculated for each exposure time and time point, where ΔK is the depth of the vessel peak from the baseline (mean noise) and σK n is the standard deviation of the noise. A visual example of ΔK and σK n are shown in Fig. 3(b). The SNR values were averaged over the mice (1, 2, and 3) and standard errors were calculated. The optimum SNR between the exposure times in each time point was chosen for the contrast images sharpness comparison. As an indicator of resolution, we compared the sharpness of the vessel edges in images by calculating fall distance (the number of pixels) of the edge of the vessel to go from 10% to 90% of ΔK value were calculated [38]. A shorter fall distance corresponds to greater sharpness. The same sampled contrast intensity profiles were considered for the fall distance calculation. Figure 3(b) shows an example trace and 10-90% fall distance measurement. In order to evaluate the change in sharpness over the time points in Experiment 1, relative change in fall distance over the time points day 14 and day 28 compared to day 0 baseline for each animal and vessel were evaluated using Eq. (3).
To compare the blood vessel images through-skull and through-implant in Experiment 2, absolute fall distances with respect to the vessels size (FWHM) were determined. FWHM of the vessels in profiles were taken as the vessel diameter [39] (Fig. 3(b)).

Experiment 1: LSI image quality over time
The results from Experiment 1 are shown in Figs. 3 and 4. Figure 3(a) shows regular images and LSI temporal contrast images at each of the 4 exposure times, for days 0, 14 and 28 from Mouse 3 (data for Mice 1-2 not shown). In the regular images, some tissue regrowth is visible at the implant edges at 14 and 28 days. It should also be noted that at day 0, blood flow is expected to be altered in response to the invasive cranioplasty surgery (e.g. due to potential reactive hyperemia [40] (increased blood flow), changes in respiration, etc.). Due to these differences in the mouse biology and physiology at different imaging time points, we sought to determine the optimal exposure time for LSI imaging at each time point, so that we could compare the images of highest SNR that could be acquired at each time point. Line inten times and ima noise were de 3(b)), and sig time was aver an exposure t gave a higher Next, usin times for day distance) of th losing resolut days 14  g. 3(b) were an e intensity prof tion and illustr Data for each ti n in Fig. 4(a). A d 28 days, whi responding to ht to compare he images were ative change in n in Fig. 4(b), e fall distance d to day 0. nalyzed for all files, peak inte rated in the ins ime point and As shown in th ile 10 ms expo these optimal the sharpness e becoming blu n fall distance and was belo increased by

Experime
As the goal imaging techn with imaging Experiment 2 comparison b both ROI 1 (th over the left p images for ex Unlike Ex Experiment 2 second experi of skull. To a the ROI mid- Fig. 5 [33]. Figure 6 plotted again ofile lines, with l diameter) is s he left side of each f skull on the right cations where line of ROI 1 and 2 for intersected by the alculated the S he mean SNR emporal contra we used this native skull an we analyzed th ll to determine -max (FWHM) 6(b) shows a p nst FWHM (i. th absolute fall shown in Fig. 6 h t e r e SNR from for each ast image image to nd a longhe 3 line e the fall ) of each plot of all .e. vessel l distance 6c.

Discussio
The WttB hol variety of bra potential to en the need for platform for u long-term, it assessed LSI imaged throug In LSI tem on image qua image increas increase with including phy statistical unc these sources with respect t (including noi that is changi such as heart fluctuations in following imp components t the physiolog optimized dur we compared 6(a)), and ea Indeed, our re points in Exp 10 ms at day points. At day blood flow co provide suffic no longer be may be need remains low regrowth, rev 6. a) Mean SNR of separate exposur ness (respectively) nt and skull for the on lds the transfor ain pathologies nable delivery repeated high understanding t is important temporal con gh the implant mporal contrast ality. As expo se [21]. While h exposure tim ysiological no certainties. The and the relati to the exposur ise due to impl ing considerab tbeat or respir n the cerebral plantation. Thi to the measure gical noises w ring each imag the SNR of im ach imaging se esults found th eriment 1, with 14. This may y 0, altered blo ould give high cient contrast o modified in re ded to increase compared to vascularization, f contrast intensity re times (error b ) vs FWHM for a e LSI temporal con rmative potenti s and neurolog to and/or acqu hly-invasive cr the brain. As th that the optica ntrast imaging up to 60 days t imaging, the sure time incr e the contrast me. Several so oise, hardware e total noise of ive contributio re time [21]. H lantation surge bly over time. ration-induced blood flow, a s motion and b ed signal and t will be differe ging session. To mages taken at ession proceed hat the optimal h maximum SN be explained b ood flow due to h contrast to th of the blood flo esponse to the e contrast in t day 0, but no , etc.), and thu best compromise between increased contrast and reduced noise. It should be noted, however, that the modification of blood flow at the different time points, as evidenced by the different optimal exposure times, may also be due to several other confounding factors. Anesthetic agents (including isoflurane and ketamine used in this study) are known to modulate cerebral blood flow, and the concentration of these drugs in the blood stream was not kept constant over the imaging sessions. Respiration can cause noise in LSI, and its rate, along with arterial blood gas levels, can significantly alter blood flow. These variables, in addition to the small sample size used in this study, make it difficult to attribute the observed differences in blood flow between the time points to a particular cause.
Image resolution (i.e. sharpness) was found to decrease by up to 13% over the 28 days compared to day 0 baseline in the 9 vessels analyzed from Mice 1-3 ( Fig. 4(b)). The relative increases in average fall distance between day 14 and day 28 compared to day 0 were 6.56% ± 1.03% and 7.69% ± 1.2%, respectively. It should be noted that while day 0 is considered as the baseline for comparison, it does not represent an unperturbed state of normal cerebral blood flow. It is likely that blood flow was altered in response to the invasive cranioplasty surgery, and thus the higher SNR and sharpness of the day 0 images compared with days 14 and 28 may be due in part to an elevated signal at day 0, rather than a loss of signal at days 14 and 28. SNR and sharpness stayed more stable between 14 and 28 days, consistent with this hypothesis. Additionally, new microvessels which were not visible in day 0 are visible at these later time points, which suggests revascularization may be occurring near the implant surface. Together, these results appear to indicate that image quality remains high up to 28 days post-implantation, but additional longer time points are needed to confirm this is the case. In Experiment 2, a day 60 time point shows that SNR is similar to the SNRs obtained at days 14 and 28 in Experiment 1 (SNR between 3 to 4 in both Fig. 4(a) and Fig. 6(a)), and microvessels are clearly visible at this later time point (Fig. 5(a)).
Comparing imaging through the skull and a long-term (60 days) implant ( Fig. 5(a)), the skull had a higher SNR for lower exposure times (1 and 2 ms), and the implant SNR was higher at longer exposure times (6 and 10 ms) ( Fig. 6(a)). The higher SNR of the throughskull images compared with implant at short exposure times may be explained by the natural filtering-out of smaller vessels imaged through the skull. Imaging through the window, on the other hand, resolves small vessels which are more sensitive to exposure time. At low exposure time, the contrast of these small vessels is very low, contributing to the lower SNR through the window compared to skull. Exposure time of 6 ms gave the highest SNR for both skull and implant and was used for comparing SNR and sharpness of vessels imaged beneath each. Because this comparison was made between two different spatial regions of the brain, different vessels of varying diameter were compared. Thus, SNR and sharpness of vessels (fall distance) are plotted against the vessel diameter (FWHM of vessels in intensity profiles). As mentioned earlier a number of small vessels, associated with lower blood flow and accordingly SNR, were detected through the implant which were naturally filtered-out by the skull (Fig. 6(b)). These detected small vessels decreased the averaged SNR shown in Fig.  6(a). In Fig. 6(c), interestingly, the data from the skull and implant fell into separate clusters, with vessels imaged through the implant having smaller FWHM and shorter fall distances, while vessels imaged through the skull have larger FWHM and higher fall distances. This data shows that vessels imaged through the implant appear smaller in diameter and have sharper borders compared to those imaged through the skull. As the size of vessels and velocity of blood flow is not expected to differ on average between the two hemispheres of the mouse brain, the apparent increase in vessel diameter imaged through the skull vs implant is likely due to blurring of the image through the skull. In general, the light transmission to the brain and reflection from the brain are lower in the through-skull image. Additionally, the skull scattering disorders the speckle pattern that was created by the brain hemodynamics, making a less accurate speckle pattern. Both of the defects are caused by skull and the skull texture is not homogeneous, making image correction or enhancement very complex [14].
The reduction in border sharpness of the vessels imaged through the skull vs implant is consistent with this explanation, and the blurred nature of the through-skull image compared to the through-implant image is visually apparent in Fig. 5(a). It should also be noted that while the SNR and sharpness of the through-implant images are an improvement over the through-skull images, the mouse skull is inherently transparent itself [14], which is not the case in larger animals or humans.
LSI temporal contrast imaging has become a useful tool for brain blood flow applications as the images it produces contain functional information (i.e. relative blood velocity) in addition to showing structure of the vessel networks. Figure 7 contains a color map version of the 6 ms LSI image from Mouse 4 ( Fig. 5(a)), where minimum relative velocity (i.e. static regions) appears blue and maximum relative velocity (i.e. maximum blood flow velocity) appears red. As seen in Fig. 7, much more precise velocity information, particularly of microvessels, can be detected through the implant, while the reduced sharpness of the vessels imaged through skull obscures precise flow determinations. Imaging microvessels (usually involving small arteries of less than 150 μm in diameter, arterioles, capillaries, and venules [41,42]) is important for many applications, as microcirculation plays a critical role in physiological processes such as tissue oxygenation and nutritional exchange [43]. Monitoring the spatio-temporal characteristics of microcirculation is crucial for studying the normal and pathophysiologic conditions of tissue metabolism. It is also useful for assessing microcirculatory dysfunction due to disease conditions such as type 2 diabetes, peripheral vascular disease (PVD), atherosclerotic coronary artery disease, obesity, heart failure, Alzheimer's, schizophrenia and hypertension, among others [44][45][46]. In addition, quantification of dynamic blood flow, angiogenesis, and vessel density are critical for monitoring the progression of wound healing [47]. Although high resolution vascular network mapping is possible using imaging modalities such as computed tomography (CT), these approaches require injection of contrast agents and pose disadvantages such as radiation exposure. Existing non-invasive methodologies (including LSI through skull) are inadequate to study blood flow at microvessel resolution [48]. Windows such as the WttB implant are thus important tools for research, and in the case of the WttB implant, can become important enablers of clinical diagnostics and therapy involving cerebral microvessels.
As recent studies show, creating novel windows for brain studies has been gaining attention recently [25,[49][50][51]. Some of these studies, involving optical clearing agents (OCAs) applied to the scalp overlying native skull, have shown limited success due to optical losses and scattering in the skull. These skin optical clearing strategies could work in conjunction with the WttB implant, to allow for imaging of cerebral blood flow in awake and behaving animals through closed scalp and implant, for example to study cerebrovascular hemodynamic responses and tissue oxygenation and other stimuli. Future studies by our group will explore this combined OCA-WttB strategy, as well as including additional imaging modalities which can couple with LSI to provide additive information, such as OCT and intrinsic optical signal (IOS). There are and n = 1), findings. Whi at later time p such as infect

Conclusio
In conclusion up to 28