Dimension-based Quantification of Aging-Associated Cerebral Microvasculature Determined by Optical Coherence Tomography and Two-Photon Microscopy

Abstract

Cerebral microvascular health is a key biomarker for the study of natural aging and associated neurological diseases. Our aim is to quantify aging-associated change of microvasculature at diverse dimensions in mice brain. We used optical coherence tomography (OCT) and two-photon microscopy (TPM) to obtain nonaged and aged C57BL/6J mice cerebral microvascular images in vivo. Our results indicated that artery & vein, arteriole & venule, and capillary from nonaged and aged mice showed significant differences in density, length, diameter, complexity, perimeter, and tortuosity. OCT angiography and TPM provided the comprehensive quantification for arteriole & venule via compensating the limitation of each modality alone. We further demonstrated that arteriole & venule at specific dimensions exhibited negative correlations in most quantification analyses between nonaged and aged mice, which indicated that TPM and OCT were able to offer complementary vascular information to study the change of cerebral blood vessels in aging.

Introduction

Aging is the major risk factor for a variety of human diseases, including Alzheimer’s disease, several types of cancers, and type II diabetes, which has a high impact on physiologic function and quality of life [1]. In order to better understand aging and age-related diseases, it is important to understand pathological changes in the key organ-systems of the body, such as the central nervous system and its microvasculature. Aging is often associated with cognitive decline, however, the exact mechanisms that cause its progression are not fully understood [2]. Implicated in aging-associated cognitive decline is the observed decrease in neural plasticity, which may result from alterations in cerebral microvasculature, providing metabolic support to the central nervous system and regulating neurogenesis [3]. Vasoconstriction in cerebral microvasculature causes a decrease in cerebral glucose utilization, which is linked to neural damage [4]. Arterial stiffness is another physiological change that occurs with aging and is associated with an increased risk of stroke and subsequent cognitive decline [5]. Aging has also been shown to impair myogenic adaptation to pulsatile pressure in cerebral microvasculature. This decreased ability to regulate pressure fluctuations in microvasculature likely contributes to microvascular damage in the brain [6]. Objective quantifications of microvasculature including density, size, length, and complexity are critically useful for physiological assessment [7]. Especially, microvascular tortuosity and vascular diameter distribution are critical for the study of cerebral microvasculature as they directly impact on cerebral blood flow and microcirculation [8, 9].

The blood vessel system is composed of arteries, arterioles, veins, venules, and capillaries and the size distribution of vessels plays a critical role in physiological circulation and medical assessment [10]. Meanwhile, these blood vessels with different sizes play different roles in the transportation of blood and nutrients to maintain normal physiological functions [11]. The distribution of vessel diameter sizes throughout the vasculature reflects vessel constriction or dilation, which critically alters blood flow and circulation [8]. Changes in the diameter distribution of vasculature have been demonstrated to be associated with aging and hypertension [12]. Current standards to obtain diameter size distribution is to compute the distance from points along the vessel centerline to the nearest boundary points [8, 13].

In addition to diameter distribution, vessel tortuosity is an important metric for studying cerebral microvasculature. Studies have shown that tortuosity of cerebral microvasculature shows a significant increase with aging and the microvascular tortuosity may necessitate hypertensive levels of pressure to supply blood flow to the brain [14-16]. Additionally, tortuous microvasculature is observed supplying the deep white matter of aged brains, and increased tortuosity is observed in microvasculature affected by leukoaraiosis [17]. However, current studies of vascular tortuosity and dimensions are limited in their capacity to apply accurate calculation metrics to microvasculature on a large scale. This is because these metrics are multifactorial and difficult to ascertain through imaging. As a result, clinical settings sometimes use fewer objective approaches like visual estimation [18]. This results in a need for aging studies that implement more comprehensive modern methods of computing global vessel tortuosity based on microvascular dimension distribution [19]. Except for tortuosity, there are several quantitative parameters associated with vessel skeleton density, diameter, complexity, and perimeter that have been used to analyze capillary flux and perfusion, the mechanism of vascular disease, and pathological vascular degeneration [7, 20-24]. Our previous study has demonstrated that young mice showed significantly higher vessel density and length in cerebral blood vessels compared to aged mice [25]. Nonetheless, the quantitative evaluation of cerebral blood vessels about aging in vessel skeleton density (VSD), vessel diameter index (VDI), vessel complexity index (VCI), vessel perimeter index (VPI), and vessel tortuosity index (VTI) still keeps unknown. Particularly, the quantitative analysis of cerebral microvasculature under specific vessel dimension distributions has not been reported. It is important to combine vascular dimension distributions with vascular metrics and parameters (VSD, VDI, VCI, VPI, VTI) as they contribute to the efficiency of blood flow through vasculature. By specific diameter distributions of cerebral vessels, integrating vascular quantification parameters can further characterize the difference and change in aging. Our objective is to combine these quantitative metrics in the analysis of aging cerebral microvasculature to quantify the degree to which microvascular alterations inhibiting circulation may take place in aging cerebral tissue.

In this study, 18 months old was set to separate nonaged and aged mice based on the mouse nature life span [26]. We implemented vessel dimensional analysis in combination with multiple vessel metrics in density, diameter, complexity, perimeter, and tortuosity in cerebral microvasculature obtained from mice with Optical Coherence Tomography Angiography (OCTA) and Two-Photon Microscopy (TPM). OCTA and TPM were implemented to obtain scans of the cerebral microvasculature, which were analyzed through dimensional distributions obtained via midline analysis with global tortuosity metrics and quantification parameters computation, accounting for various parameters pertaining to vessel branches. VSD, VDI, VCI, VPI, and VTI were analyzed under the dimension distribution of microvasculature in nonaged and aged mice. This study will aid in probing the relationship between certain vascular alterations that may take place in the process of aging, which would reveal relationships between aging and age-associated disorders.

Methods and Materials

Animal Model

<18 months old (nonaged, N=9 for OCT and TPM) and >18 months old (aged, N=14 for OCT and N=9 for TPM) male C57BL/6J mice purchased from the National Institute on Aging at Charles River Laboratory (Wilmington, MA) were used for this study. Mice were housed in the Rodent Barrier Facility at University of Oklahoma Health Sciences Center with a controlled photoperiod (12 hours light and 12 hours darkness) and unlimited access to water and standard AIN-93G diet (ad libitum) and under specific pathogen-free conditions. All procedures were approved by the Institutional Animal Use and Care Committees of University of Oklahoma Health Sciences Center [25].

Chronic Cranial Window Surgery

For the maintenance of the experimental cohort, animals have been transferred to conventional animal housing facility. The detailed surgical procedures have been described in the previous study [25] In brief mice were anesthetized by inhalation of 2-3 % isoflurane (ISOTHESIA, Henry Schein Animal Health, OH) with Surgivet Classic T3 vaporizer (Smiths Medical, Minneapolis, MN). The status of anesthesia was monitored through observing eye blink, toe, and tail pinch reflexes. The experimental mouse was placed on a heating pad to maintain body temperature during the surgery. The mouse head was fixed using the ear and nose bars of the stereotaxic frame. Eye ointment was utilized on both eyes and hair removal cream was employed to remove the hair on the top of the mouse head. After removing the scalp skin, lidocaine drops were applied and the skull was thinned with a pneumatic dental drill in a circular manner to generate craniotomy. After removal of the skull bone, a special glass coverslip was used to cover the window and acrylic glue and dental cement were used to secure it. Animals have been treated with systemic analgesic and antibiotic. The mice were placed back in the animal cage and closely monitored until consciousness was regained from the anesthesia after surgery. All intravital imaging studies were conducted at least 2 weeks after surgery.

Optical Coherence Tomography Angiography

A swept-source OCT imaging system (VEG220, Thorlabs) with a swept-source laser of 1310 nm center wavelength and 100 nm spectral bandwidth was employed to provide an axial resolution of 14 μm in the air (10 μm in mouse brain) and lateral resolution of 20 μm. The wavelength-swept frequency and sensitivity in this system were 200 kHz and 98 dB, respectively [27-37]. The OCT-based optical microangiography (OMAG) algorithm was utilized to acquire in vivo volumetric angiography in mouse cerebral cortex by coherently analyzing the intrinsic scattering property of moving red blood cells (RBCs) in blood vessels [15, 25]. A field-of-view (FOV) of 2.5 × 2.5 x 1 mm³ (X by Y by Z) with 400 A-scans (fast-scanning axis) and 400 B-scans (slow-scanning axis) and 160 pixels in depth was used to provide a 6.25 μm× 6.25 μm x 6.25 μm sampling resolution. Eight images were acquired to generate a cross-sectional vascular map at each B-scan sampling position. A summed intensity projection (SIP) method was used to provide enface projected images for cerebral blood vessels. During the experiment, the focusing depth for OCT acquisition was slightly below the brain surface to avoid reflection. The depth region for OCTA analysis was 50-150 μm which matched the depth region of TPM. Figure S1 showed the scanning protocol of OCT imaging for the mouse brain.

Two-Photon Microscopy

A FluoView 1000 MPE two-photon microscope (Olympus, Tokyo, Japan) with a 690-1040 nm MaiTai HP DeepSee-OL laser (Spectra-Physics, San Jose, CA) and a 1.05 numeric aperture XLPLN25XWMP×25 water immersion objective (Olympus, Tokyo, Japan) were applied for intravital imaging. Alexa Fluor 594 Conjugate (4 mL/kg body wt of 1 mg/mL WGA-A594; Thermo Fisher Scientific) was injected retro-orbitally to label the vascular glycocalyx. The excitation light was 800 nm, and the emission light was collected by detector with bandpass filter of 575-630 nm. One center imaging location were performed via the cranial window of the mouse head top based on observable microvasculature. A FOV of 508 × 508 μm² (X × Y) at the depth of 0-150 μm (Z) with pixel numbers of 512 × 512 × 31 (X × Y × Z) was used to provide TPM z-stacks with the spatial resolution of 1 × 1 × 5 μm. The full description of the imaging details using TPM for mice was reported in the previous study [25]. The scanning protocol of TPM scanning for the mouse brain was shown in Figure S1.

Vessel Diameter Assessment

Imaged vessels (Figure 1A) were first converted to binary images (Figure 1B). The edge of vessels (Figure 1C) was extracted from binary images. Vessel skeletons (Figure 1D) were then obtained through use of Zhang-Suen Thinning, the skeletonized vessels served as midlines of the original vessels. The image was then iterated through and distance transform was applied to every pixel along the vessel skeleton, calculating the distance in pixels from the midline to the nearest non vessel pixel, serving as vessel radius. This value was multiplied by 6.25 μm in OCTA and 1.00 μm in TPM to reflect the distance in micrometers and doubled to serve as vessel diameter. The vessel diameters taken at each pixel of the skeleton were averaged to reflect the mean vessel diameter in micrometers of the imaged vasculature. A color organized image was generated based upon pre-determined color ranges to visualize the distribution of vessel diameter within the image. Figure 1E and 1F showed cerebral blood vessels and their corresponding skeleton with different diameter dimensions that were labeled with rainbow colors from a representative OCT angiography image.

Figure 1.

Image processing of blood vessel dimension segmentation and quantification. A, original OCTA image. B, binary. C, edge. D, skeleton. E, color-labeled dimension-separated blood vessels. F, skeleton color-labeled dimension- separated blood vessels. G, skeleton color-labeled branch- separated blood vessels. The color is to label individual vascular branch and has no specific value indication. H, vessel tortuosity quantification of color-labeled branches. The color is to label individual vascular branch and has no specific value indication. The number is used to count the number of branches. I, vessel skeleton density map. J, vessel diameter index map. K, vessel complexity index map. L, vessel perimeter index map. M-P, skeleton density, diameter, complexity, and perimeter maps with corresponding binary vessels. Scale bar is 250 μm.

Junction and End-Point Identification

In order to assess the tortuosity of imaged vasculature, the vasculature was first divided into individual vessel segments. For every pixel in the vessel skeleton, the 3x3 square of pixels centered on that pixel was analyzed to identify vessel junctions and end points. For each pixels, the value of each of the 8 surrounding pixels were passed into an array (vessel pixels were white, background pixels were black), starting with the center left pixel at position 0, and progressing clockwise to the bottom left pixel at position 7 (Figure S2A). This array was analyzed to produce the number of neighboring branches attached to the center pixel. A white pixel indicates a neighboring branch, however, if a white orthogonal neighbor (positions 0, 2, 4, and 6) (Figure S2B), is adjacent to a white diagonal neighbor (positions 1, 3, 5, and 7, Figure S2C), these 2 pixels corresponded to only 1 neighboring branch (Figure S2D). This is because 2 adjacent neighboring pixels could correpond to either a single branch approaching the center pixel from a corner, or 2 branches approaching from different directions. At the point where they are adjacent, they are treated as one branch, and then labelled as separate branches at a point where they diverge (Figure S2E). If a pixel has only one neighboring branch, it is identified as a vessel endpoint. If it has two neighboring branches, it is identified as a continuation of one vessel, and if it has greater than two neighboring branches, it is identified as a junction vessels (Figure S2F). Junctions and endpoints are color coded in blue.

Segment Identification

To identify vessel segments, all pixels in the image were iterated through, stopping at junctions and endpoints. At each junction or endpoint, the eight neighboring pixels were iterated through. At every white neighboring pixel, a vessel stepping function was initiated. A list to hold the pixel coordinates of the segment was created, with the original endpoint or junction coordinates as the first index and the identified white pixel’s coordinates as the next index. This pixel was then analyzed: the available ‘Rook Moves’ (white 4-neighbors) were accounted. If there were no available Rook Moves, ‘Bishop Moves’ (white 8-neighbors) were accounted. When an available move was identified, its pixel coordinates were appended to the list of segment pixel coordinates, and it was then analyzed in the same manner. Pixels that had already been added to the lists of coordinates were not eligible moves. This resulted in the program counting down the vessel segment, until it reached another end point or junction, at which it was terminated. This process was iterated through all branches in the image, resulting in a list of lists containing pixel coordinates for every vessel segment, which was exported to MATLAB for further processing.

Vessel Tortuosity Index

Global tortuosity is the measure of the tendency of a vessel to twist and turn throughout its length, this affects the efficiency with which blood can circulate through the tissue [9]. In MATLAB, the list of vessel segment pixel coordinates was imported and processed using a cubic smoothing spline with a regularization parameter of 0.5. This converted each segment’s list of pixel coordinates into smoothed line segments, as shown in Figure 1G. These Line segments were then processed with the Vessel Tortuosity Index (VTI) tool by Maz M Khansari, as shown in Figure 1H. Obtained VTI values for every segment in the image were then averaged to reflect the mean VTI of the imaged vasculature [38].

Vessel Skeleton Density (VSD, Figure 1I and 1M) was defined as the unit-less ratio of the total image area of the vessel length to the total image area in the skeletonized maps.

$$VSD=\frac{{\sum }_{i=1,j=1}^n{S}_{(i,j)}}{{\sum }_{i=1,j=1}^n{X}_{(i,j)}}$$

Where, $S_{(i,j)}$ represented the pixel of skeletonized vessel area (white pixels on Figure 1D), and $X_{(i,j)}$ represented all the pixels in the skeletonized image (all pixels on Figure 1D). VSD was an assessment of the vessel length density regardless of the vessel diameter [7, 21].

Vessel Diameter Index (VDI, Figure 1J and 1N) was obtained by comparing the binarized map to the skeletonized map to yield the average vessel caliber.

$$VDI=\frac{{\sum }_{i=1,j=1}^n{A}_{(i,j)}}{{\sum }_{i=1,j=1}^n{S}_{(i,j)}}$$

Where, $A_{(i,j)}$ represented the pixel of binarized vessel area (white pixels on Figure 1B). and $S_{(i,j)}$ represented the pixels of the skeletonized vessel length (white pixels on Figure 1D). VDI quantified the averaged vessel caliber of blood vessels within the image and presented the vessel size information regardless of the vessel area and length [7, 21].

Vessel Complexity Index (VCI, Figure 1K and 1O) was defined using the perimeter map and the binarized image to describe the complexity of the morphology of blood vessels.

$$VCI=\frac{{({\sum }_{i=1,j=1}^n{P}_{(i,j)})}^2}{4\pi {\sum }_{i=1,j=1}^n{A}_{(i,j)}}$$

Where, $P_{(i,j)}$ represented the pixel of edged vessel area in the perimeter map (white pixels on Figure 1C), and $A_{(i,j)}$ represented the pixels of the binarized vessel area (white pixels on Figure 1B). VCI was also a unit-less ratio to quantify the morphological characteristics of blood vessels [7, 21].

Vessel Perimeter Index (VPI, Figure 1L and 1P) was calculated as a unit-less ratio of the vessel perimeter area to the total area in the perimeter maps.

$$VPI=\frac{{\sum }_{i=1,j=1}^n{P}_{(i,j)}}{{\sum }_{i=1,j=1}^n{X}_{(i,j)}}$$

Where, $P_{(i,j)}$ represented the pixel of edged vessel area in the perimeter map (white pixels on Figure 1C), and $X_{(i,j)}$ represented all the pixels of in the perimeter map (all pixels on Figure 1C). VPI estimated the total size information of blood vessels as it took the vessel length and diameter into consideration [7, 21].

Results

Overall Vasculature in OCTA and TPM between Nonaged and Aged Mice

We obtained SIP images of cerebral blood vessels at 50-150 μm depth from OCTA and 50-150 μm depth from TPM. We showed representative images of nonaged and aged mice from OCTA (Figure 2A₁ and 2B₁) and TPM (Figure 2C₁ and 2D₁). Both OCT angiography and TPM showed that there were larger blood vessels in nonaged mice (Figure 2A₁ and 2C₁) compared to aged mice (Figure 2B₁ and 2D₁). Binary images (Figure 2A₂-D₂) from the original OCTA and TPM were applied to separate different vessel dimensions and perform the quantification analysis. With the dimension-separation processing, we labeled vessels with different diameters which allowed us to further analyze the distribution of cerebral blood vessels with various sizes, as shown in Figure 2A₃-D₃. Figure 2A₄-D₄ showed vessel tortuosity quantification analysis after we separated the vessels into branches from vessel skeletons. Figure 2E₁-E₅ and 2F₁-F₅ were quantitative assessment of cerebral blood vessels in skeleton, diameter, complexity, and perimeter from OCTA and TPM, respectively. We observed that all the five parameters showed no significant difference between nonaged and aged groups. VSD, VPI, and VTI of nonaged mice were higher in OCTA but lower in TPM compared to aged mice. In contrast with these, VDI and VCI of nonaged mice, were higher in OCTA and TPM. OCTA and TPM had consistent results in VDI and VCI but opposite results in VSD, VPI, and VTI in global vascular quantification for cerebral blood vessels of nonaged and aged mice.

Figure 2.

Representative images and quantitative comparisons of cerebral blood vessels in OCTA and TPM for nonaged and aged mice. A₁-A₄, the representative nonaged mouse’s original, binary, color-labeled dimension-segmented, and tortuosity quantification OCTA images. B₁-B₄, the representative aged mouse’s original, binary, color-labeled dimension-segmented, and tortuosity quantification OCTA images. C₁-C₄, the representative nonaged mouse’s original, binary, color-labeled dimension-segmented, and tortuosity quantification TPM images. D₁-D₄, the representative aged mouse’s original, binary, color-labeled dimension-segmented, and tortuosity quantification TPM images. E₁-E₅, quantitative comparisons of VSD, VDI, VCI, VPI, and VTI of OCT angiography images between nonaged and aged mice. F₁-F₅, quantitative comparisons of VSD, VDI, VCI, VPI, and VTI of TPM angiography images nonaged and aged mice. The color in A₄-D₄ is to label individual vascular branch and has no specific value indication. The number in A₄-D₄ is to count the number of branches. N=9 for OCT and TPM in nonaged and N=14 for OCT and N=9 for TPM in aged. Scale bar is 250 μm in OCTA and 50 μm in TPM.

Anatomy-based Vasculature Splitting in OCT and TPM between Nonaged and Aged Mice

To explore the difference in cerebral blood vessels between nonaged and aged mice under specific vascular dimensions, we applied an anatomy-based threshold for splitting artery & vein and arteriole & venule in OCTA and splitting arteriole & venule and capillaries in TPM. The threshold for separating artery & vein from arteriole & venule was 100 μm and the threshold for separating arteriole & venule from capillaries was 13 μm [10]. With the anatomy-based separation, we provided the representative artery & vein (Figure 3A₁ and 3B₁) and arteriole & venule (Figure 3A₂ and 3B₂) images from OCTA as well as arteriole & venule (Figure 3C₁ and 3D₁) and capillaries (Figure 3C₂ and 3D₂) images from TPM. Additionally, tortuosity quantification of OCTA (Figure 3A₃-A₄ and 3B₃-B₄) and TPM (Figure 3C₃-C₄ and 3D₃-D₄) was employed to assess the difference between nonaged and aged mice under the anatomy-based segmentation. We observed that arteriole & venule occupied most blood vessels in OCTA and there were limited artery & vein to provide quantification analysis. In comparison, both arteriole & venule and capillary were abundant in TPM to make quantification assessment. In OCTA, both artery & vein and arteriole & venule of nonaged mice displayed higher VSD, VDI, VCI, VPI, and VTI values than that of aged mice, as shown in Figure 3E₁-E₅ and 3F₁-F₅, which indicated that artery & vein kept consistent with arteriole & venule in all the quantification parameters from OCTA. However, arteriole & venule only kept consistent with capillaries in VSD (Figure 3G₁ and 3H₁) and VPI (Figure 3G₄ and 3H₄) from TPM. Nonaged mice showed lower VSD and VPI than aged mice. Nonaged mice showed higher arteriole & venule but lower capillaries in VDI (Figure 3G₂ and 3H₂) and VTI (Figure 3G₅ and 3H₅) from TPM compared to aged mice. Particularly, the VCI of nonaged mice was significantly lower in arteriole & venule but significantly higher in capillary than aged mice, as shown in Figure 3G₃ and 3H₃.

Figure 3.

Quantitative assessment of anatomy-based vascular separation in OCTA and TPM between nonaged and aged mice. A₁-A₄, artery & vein from OCTA, arteriole & venule from OCTA, artery & vein tortuosity from OCTA, and arteriole & venule tortuosity from OCTA in nonaged mice. B₁-B₄, artery & vein from OCTA, arteriole & venule from OCTA, artery & vein tortuosity from OCTA, and arteriole & venule tortuosity from OCTA in aged mice. C₁-C₄, artery & vein from TPM, arteriole & venule from TPM, artery & vein tortuosity from TPM, and arteriole & venule tortuosity from TPM in nonaged mice. D₁-D₄, artery & vein from TPM, arteriole & venule from TPM, artery & vein tortuosity from TPM, and arteriole & venule tortuosity from TPM in aged mice. E₁-E₅, VSD, VDI, VCI, VPI, and VTI of artery & vein in OCTA. F₁-F₅, VSD, VDI, VCI, VPI, and VTI of arteriole & venule in OCTA. G₁-G₅, VSD, VDI, VCI, VPI, and VTI of arteriole & venule in TPM. H₁-H₅, VSD, VDI, VCI, VPI, and VTI of capillaries in TPM. The color in A₃-D₃ and A₄-D₄ is to label individual vascular branch and has no specific value indication. The number in A₃-D₃ and A₄-D₄ is to count the number of branches. Threshold between artery & vein and arteriole & venule is 100 μm. Threshold between arteriole & venule and capillary is 13 μm. N=9 for OCT and TPM in nonaged and N=14 for OCT and N=9 for TPM in aged. *, P-value < 0.05, **, P-value < 0.01. Scale bar is 250 μm in OCTA and 50 μm in TPM.

Customized-based Vasculature Splitting in OCTA and TPM between Nonaged and Aged Mice

We further specified the customized-based dimension separation for vascular segmentation to further investigate the difference between nonaged and aged mice from OCTA and TPM. In OCTA, we set 20 μm as the dimension interval to split cerebral blood vessels (supplement Figure S3), since artery & vein only occupied a minority of vascular information, there was no meaningful comparison of customized dimension separations for artery & vein. Therefore, we split artery & vein (Figure 4A₁ and 4B₁) and arteriole & venule (Figure 4A₂ and 4B₂) images from the original OCTA image (Figure 4A and 4B). Then, we obtained separated images for arteriole & venule (<100 μm) in OCTA, as shown in Figure 4A₃-A₇ and 4B₃-4B₇. We found that most blood vessels were in 40-100 μm dimensions from OCTA (Figure 4A₃-A₅ and 4B₃-B₅) but blood vessels in 0-40 μm range were scattered and unable to provide sufficient vascular information for quantification assessment as shown in the representative OCTA images (Figure 4A₆-A₇ and 4B₆-B₇). In TPM, we split artery & venule in increments of 20 μm and capillary in increments of 6 μm from the original image (Figure 4C-D and supplement Figure S3). To match the customized dimension of 20-40 μm in OCTA and TPM, a customized dimension of 13-20 μm in TPM was built to provide the quantification assessment for arteriole & venule, and the detailed principle of the threshold-selecting was provided in supplement Figure S3. We observed that there was only a minority of blood vessels larger than 40 μm in arteriole & venule that can provide quantification analysis from TPM. Moreover, the data from capillaries that were smaller than 6 μm was unable to offer quantification analysis (supplement Figure S3).

Figure 4.

Quantitative assessment of customize-based vascular segmentation in OCT and TPM between nonaged and aged mice. A-A₇, the original, artery & vein, arteriole & venule, 80-100, 60-80, 40-60, 20-40, and 0-20 μm OCTA images in nonaged mice. B-B₇, the original, artery & vein, arteriole & venule, 80-100, 60-80, 40-60, 20-40, and 0-20 μm OCTA images in aged mice. C-C₄, the original, arteriole & venule, capillary, 20-40, and 13-20 μm TPM images in nonaged mice. D-D₄, the original, arteriole & venule, capillary, 20-40, and 13-20 μm TPM images in aged mice. N=9 for OCT and TPM in nonaged and N=14 for OCT and N=9 for TPM in aged. Scale bar is 250 μm in OCTA and 50 μm in TPM.

Overlapped-Dimension Vasculature Comparison in OCTA and TPM between Nonaged and Aged Mice

OCTA and TPM images with the customized dimension separation implied there was probably an association of vascular quantification assessments between nonaged and aged mice. We compared quantitative assessments of customized dimension separation of blood vessels within arteriole & venule from OCTA and TPM and particularly concentrated on the comparison of overlapped-dimension blood vessels with 0-40 μm in OCTA and 13-40 μm in TPM. We found that all quantitative parameters of cerebral blood vessels in 40-100 μm from OCTA (except for VDI in 60-80 and 80-100 μm) and 13-40 μm from TPM (except VCI in 13-20 μm) between nonaged and aged mice were significantly different except for the VTI (Figure 5). Our results showed that vessel separations of 40-100 μm in OCTA and 13-40 μm in TPM showed a significant difference of cerebral blood vessels associated with aging between nonaged and aged mice. Additionally, with the comparison of 0-40 μm in OCTA and 13-40 μm in TPM, all quantitative parameters showed opposite difference tendency between nonaged and aged mice except for VDI and VTI, but more significant differences were monitored from TPM. Considering the limited spatial resolution of OCTA, TPM was more effective to provide the quantification assessment of cerebral blood vessels in 13-40 μm than OCTA and could compensate for the limited data on arteriole & venule with the dimension below 40 μm in OCTA.

Figure 5.

Vascular quantification analysis in customized dimension segmentation in OCT and TPM between nonaged and aged mice. VSD, vessel skeleton density. VDI, vessel diameter index. VCI, vessel complexity index. VPI, vessel perimeter index. VTI, vessel tortuosity index. N=9 for OCT and TPM in nonaged and N=14 for OCT and N=9 for TPM in aged. *, P-value < 0.05, **, P-value < 0.01, ***, P-value < 0.001.

The Correlation of Vascular Quantification at Different Customized-based dimensions from OCTA and TPM between Nonaged an Aged Mice

To further compare the difference between vessels with different sizes and fully take advantage of the complementary results from OCTA and TPM, we split blood vessels with 40-100 μm in OCTA and 13-40 μm in TPM for quantification analysis. Moreover, we used a difference tendency defined by the slope of the fitting curve/line of the tendency change between nonaged and aged mice in quantitative parameters, to describe the correlation between vessels with different sizes (13 μm - 40 μm vs 40 μm −100 μm). The product of the slope between vessels with different sizes was utilized to characterize the status of correlation. We observed that quantifications of arteriole & venule with 40-100 μm in OCTA (Figure 6A₁-A₅) and 13-40 μm in TPM (Figure 6B₁-B₅) were significantly different in all parameters except for VTI. The combination of OCTA and TPM described the quantification analysis for the entire range of arteriole & venule dimensions. The fitting line of the difference tendency between nonaged and aged mice (Figure 6C₁-C₅) showed only VTI displayed the same tendency from OCTA and TPM. Table 1 showed the correlation value between vessels with different sizes from OCTA and TPM, which was employed to combine the correlation plot (Figure 6D₁-D₅) to investigate the association between vessels with different sizes from OCTA and TPM. The correlation of vascular quantifications between vessels with different sizes from OCTA and TPM was shown in supplement Table S1 and Figure S4 for the comparison of full range, arteriole & venule, and capillaries between nonaged and aged mice. We found that the difference tendency of quantification parameters in nonaged and aged mice between vessels with different sizes were negatively correlated except for VTI which had a positive correlation. This indicated that OCTA and TPM provided quantitative comparisons for arteriole and venule with specific vascular dimensions, but that information has not been overlapped. Therefore, OCTA and TPM offered independent quantitative assessments between nonaged and aged mice throughout arteriole & venule specifically for different sizes.

Figure 6.

Quantification of vascular dimension segmentations and the correlation of vessels with different sizes between nonaged and aged mice. A₁-A₅, quantification assessments VSD, VDI, VCI, VPI, and VTI in OCTA (40 μm - 100 μm). B₁-B₅, quantification assessments VSD, VDI, VCI, VPI, and VTI in TPM (13 μm - 40 μm). C₁-C₅, Fitting plot of difference tendency between nonaged and aged mice for quantification analyses. D₁-D₅, correlation plot of difference tendency between nonaged and aged mice for quantification analyses. Fitting and correlation plots were built based on the mean and standard deviation values under normalization. N=9 for OCT and TPM in nonaged and N=14 for OCT and N=9 for TPM in aged. **, P-value < 0.01, ***, P-value < 0.001.

Table 1.

Angles, slopes, and correlations of the difference tendency in nonaged and aged mice between OCTA and PTM. “−” indicated negative correlation and “+” indicated positive correlation.

		VSD	VDI	VCI	VPI	VTI
Angle	OCT	305.22	15.21	302.65	306.4	342.83
	TPM	50.3	304.72	47.05	54.19	309.24
Slope	OCT	−1.42	0.27	−1.56	−1.36	−0.31
	TPM	1.2	−1.44	1.07	1.39	−1.22
Correlation	Value	−	−	−	−	+

Discussion

OCT angiography and TPM were employed to monitor the differences and changes in the morphology of cerebral blood vessels in nonaged and aged mice, and the quantification from different imaging methods was explored. OCT angiography was performed by calculating decorrelation across multiple B-scans at the same location to visualize dynamic blood flow changes and reconstruct the vascular structure [39, 40]. The signal-to-noise ratio (SNR) of small capillary detection was influenced by the blood flow, which, in turn, was affected by vascular size. Additionally, capillaries situated beneath larger arteries, arterioles, veins, or venules may have their vascular signals blocked, thus affecting the detection and analysis of certain blood vessels. TPM was a 3D imaging technique via two longer [41-43]. Since the fluorescent emission only happened while two photons were absorbed by fluorophores simultaneously, TPM achieved super high resolution and high SNR imaging for extremely small capillaries compared to OCT. Nonetheless, the distribution of fluorescence dyes in scanning regions was determined by the metabolism of the animals and had an impact on the acquisition of fluorescence signal intensities. Furthermore, the limited field of view (FOV) resulting from the use of a collimated laser excitation beam with standard diffraction-limited optics restricted the imaging of large regions with TPM. TPM may also suffer from photobleaching of the fluorescence dyes, which may cause signal decrease after imaging for a certain time [44]. We utilized several quantification assessments including skeleton density, diameter, complexity, perimeter, and tortuosity to demonstrate the differences of cerebral blood vessels between nonaged and aged mice. These findings suggest that changes in vascular morphology can be used as a biomarker to assess the aging process in cerebral cortex. We observed that nonaged mice showed higher VSD, VDI, VCI, VPI, and VTI values than aged mice in OCTA, which probably resulted in higher blood flux and blood-oxygen diffusion in nonaged mice brains compared to aged mice [9, 15, 25]. Additionally, TPM stated that nonaged mice showed higher VDI and VCI (Figure 2F3-F4) but lower VAD, VSD, VPI, and VTI than aged mice, which indicated that there was an inconsistency in vascular quantifications of OCTA and TPM possibly due to differences in the regions of interest (ROI) and spatial resolutions during imaging. Specifically, since the field of view (FOV) in OCT imaging encompassed the size of the cranial window, while the FOV in TPM imaging was smaller than the cranial window, vascular quantification in TPM imaging was influenced by the choice of scanning regions within the cranial window area. Furthermore, OCT lost the quantification of certain capillaries compared to TPM due to the disparity in lateral resolutions, with OCT offering 20 μm and TPM providing 1 μm. Meanwhile, TPM also lost the quantification of certain arterioles and venules compared to OCT due to the limitation of the FOV.

We then split artery & vein from arteriole & venule in OCTA and arteriole & venule from capillaries in TPM based on the anatomy dimension segmentation and found that OCTA and TPM images performed differently between nonaged and aged mice in vascular quantifications. OCTA provided few vascular information on artery & vein for statistical analyses due to the limitation of FOV and penetration. Thus, the quantification assessment from OCTA was primarily for arteriole and venule [45]. However, arteriole & venule in OCTA also included the vascular information of capillaries which showed insufficient image quality because of the limitation of resolution from OCTA, which would cause errors to quantifications for the statistics. TPM was able to provide sufficient vascular information for the quantification analysis of arteriole & venule and capillaries. Compared to OCTA, arteriole & venule and capillary showed different change tendencies of quantification assessments between nonaged and aged mice in TPM. Particularly, TPM revealed significant quantitative differences between nonaged and aged mice in arterioles & venules, which could be used in combination with OCTA to improve accuracy and minimize errors [46]. Through customized dimension segmentation, we observed significant differences in vascular quantifications between nonaged and aged mice in two different ranges: OCTA 40-100 μm arteriole and venule, and TPM 13-40 μm arteriole and venule. These results suggest it is necessary to specify the size range when studying the vessels using OCTA results in the future to get more accurate results. Moreover, these results indicate that TPM can compensate for the limitations in spatial resolution of OCTA by detecting small arterioles, venules, and capillaries, while OCTA can compensate for the limitations in the FOV of TPM by detecting large arterioles and venules. Additionally, OCTA provided wide and relatively macroscopic vascular information while TPM offered small and more microscopic vascular information to show multidimensional quantification and comparisons in mouse brain aging. With the comparison of the correlation between OCTA and TPM, there was no overlapped vascular information between OCTA and TPM, which confirmed that OCTA and TPM independently provided essential vascular information for quantifications of cerebral blood vessels. Therefore, the combination of OCT and TPM allowed for more effective quantification of cerebral blood vessels with different sizes in mice from capillaries to arterioles and venules during aging.

With the growth of neural tissues, the brain gradually expanded and remodeled the vascular networks due to the need for an adequate supply of oxygen and nutrients [47]. The vascular alteration resulted in the progressive angiogenesis and degeneration of blood vessels during the aging process [48]. Since VSD, VDI, VCI, VPI, and VTI were associated with the expansion of vascular networks, the relative change of these parameters indicated the generation or degeneration of cerebral blood vessels. Our results revealed that VSD, VDI, VCI, and VPI exhibited opposite trends between non-aged and aged groups in relation to different-sized blood vessels, as depicted in Figure 6. This suggests a correlation between the expansion and degeneration of cerebral blood vessels and their size. For example, in Figure 6, vascular diameter increased in the 40-100 μm range but decreased in the 13-40 μm range, indicating that larger cerebral blood vessels expanded while smaller ones degenerated with aging.

In this study, we demonstrated that OCTA and TPM can serve as imaging tools to effectively monitor and quantify vascular differences and changes in nonaged and aged mice. It should be noted that the sample size in this study remains a limiting factor in determining the significance of the observed differences in vascular quantifications between nonaged and aged mice, even though the trend of the differences is evident. In future studies, more mice should be involved to get more reliable statistical conclusions. Furthermore, the selection of imaging area for TPM is also a factor that may affect the quantitative analysis because the FOV of TPM is smaller than the cranial window. Therefore, this may potentially influence vascular quantification results in TPM. We suggest that future studies can set fixed imaging areas on the cranial window for each mouse to obtain data to avoid the effect from the imaging region selection.

Conclusion

In this study, we utilized OCT and TPM to monitor and quantify cerebral vascular differences between nonaged and aged mice with the anatomy- and customize based vascular segmentation strategies. Our results demonstrated that different vessel types (artery & vein, arteriole & venule, and capillaries) showed significant differences and change tendencies in vascular skeleton density, length, diameter, complexity, perimeter, and tortuosity in aging process. The negative correlation between vessels with different sizes from OCTA and TPM confirmed that arteriole & venule contained significant differences in vascular quantification assessments between nonaged and aged mice when looking at different size scales, which also indicated combining OCTA and TPM could provide comprehensive quantification analysis. Therefore, imaging methods with different spatial scales such as OCT and TPM should be considered to use as multimodal imaging modalities to achieve more accurate vascular monitoring and quantification for vascular analyses.

Data Availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.