The imaging system used in this study is rooted in a swept-source OCT configuration, featuring a line scan rate of 100 kHz. The swept source (SL131090, Thorlabs, Inc.) operated at a central wavelength of 1,300 nm with a spectral bandwidth of 100 nm, thereby delivering a measured axial resolution of 16 µm in air. As shown in Figure 1, the light source was bifurcated into two beams via Coupler1, subsequently traversing the sample and reference arms separately, which converged and interfered in Coupler2. Ultimately, a balanced detector was utilized to register interference signals and mitigate common-mode noise. Within the sample arm, an x–y galvanometer was incorporated for 3D volume scanning, and a telecentric scan lens (f = 36 mm) was used to concentrate the probing light beam onto the area of interest. This configuration achieved a lateral resolution of approximately 23.4 µm and a depth of focus of approximately 108 µm.

For volumetric imaging, a stepwise raster scanning protocol (z-x-y) was implemented, consisting of 2,048 pixels in each A-scan, 350 A-lines per B-scan (fast-scan, x direction), and 1,750 B-scans per volume (slow-scan, y direction). At 350 tomographic positions, five successive B-frames were obtained to compute dynamic flow signals and reduce the decorrelation noise in blood flow imaging. Each volume dataset required an acquisition time of 13.4 s and spanned a 7*7 mm² spatial region in the x–y plane.

We procured male Sprague–Dawley rats, aged 7 weeks and weighing between 160 and 200 g, from Hangzhou Medical College (certificate no. SCXK2019-0002). All rats were accommodated in individually ventilated cages within an experimental animal room (22°C, 55% humidity, and 12/12-h light/dark cycle) and were provided with a standard laboratory diet and water. The Animal Experimentation Committee of The First Affiliated Hospital of Zhejiang University Medical College approved this study (no. 2022-1,125). After taking the measurements, the rats were humanely euthanized via excess anesthesia, effectively minimizing their discomfort.

Following the establishment of general anesthesia with chloral hydrate (5%, 8 mL/kg), the surgical area was shaved and the skin was cleansed with 70% ethanol. The dorsal region of the cranium was shaven and prepared aseptically for the surgical procedure. Skull bone exposure was achieved through a midline incision and spreading of the skin, fascial, and periosteal layers overlaying the calvarium’s sagittal suture [32]. In rats, 5-mm calvarial defects are classified as critical-sized bone defects [33]. Initially, a 5-mm circular bone defect was created on the calvarial bone using a dental surgical drilling unit, consistently cooled with sterile saline. Subsequently, the calvarial bone was gently elevated to prevent dura tears. Concurrently, the midsagittal suture was circumvented to eliminate its contribution to bone restoration and minimize the risk of superior sagittal sinus damage 11. After thoroughly cleaning the area with physiological saline to remove all bone debris, the periosteal layer was meticulously sutured with 4-0 silk sutures (Ethicon, Somerville, NJ, United States) and the skin was closed using 3-0 silk sutures (Ethicon, Somerville, NJ, United States). The surgery day was designated as day 0.

On days 7,14, and 28, rats were individually restrained in the stereotaxic instrument under chloral hydrate (5%, 8 mL/kg) anesthesia. The calvarial bone was exposed by reopening the midline incision and spreading the skin, fascia, and periosteal layers covering the sagittal line of the skull. Subsequently, the established OCT system was utilized to capture microvessel images at different time stages.

Upon acquiring the raw spectral interferogram data S (k, x, y, t), a series of preprocessing steps and motion contrast OCTA were implemented to extract dynamic blood signals from the critical-sized calvarial defect rat model. Here, k denotes the wave number, x signifies the fast scan direction, y represents the slow scan direction, and t stands for the temporal index of adjacent B-scans at the same location. In detail, by executing a Fourier transformation of the k-direction, we can obtain the depth-resolved complex signal A (z, x, y, t) and generate a structural cross-sectional image via the amplitude component of A. By applying cross-correlation registration between adjacent B-frames, we compensate for minor skull displacement due to respiration [34]. Subsequently, we calculate the inter-frame complex decorrelation using a local kernel to achieve the OCTA motion contrast. Additionally, by merging the features of the inverse signal-to-noise ratio (SNR) and decorrelation (ID), we apply a vascular morphological mask with the adaptive signal-to-noise ratio to decorrelation mapping, thereby procuring angiography (ID-OCTA) [18].

To scrutinize the microvascular circulation during the longitudinal growth of the critical skull defect, the acquired angiography was bifurcated into two layers: the surface layer (from the surface to a depth of 483 µm) and the deep layer (from a depth of 483–967 μm). For the surface layer, after employing Gaussian filters to minimize noise, angiography was binarized with a global threshold. Subsequently, the parameters of the vessel area density (VAD) and maximum vascular diameter index (MVDI) of the projection view image were calculated to quantify the vascular structure alterations. V A D = ∑ x = 1 , y = 1 n A x , y n 2 (1) V D I = ∑ x = 1 , y = 1 n A x , y ∑ x = 1 , y = 1 n S x , y (2)

In this equation, A(x,y) symbolizes the white vessel pixels in the binary image, n is the side length of the image, and S x , y represents the pixels of the vascular skeleton. VAD is obtained by calculating the ratio of white pixels to total pixels within the modeling area. The largest vascular region is manually selected, and MVDI is calculated by determining the ratio of white pixels to total pixels along the vessel length.

Defect sites, along with surrounding bone and soft tissue, were dissected and preserved in 4% PFA. The specimens were then decalcified through immersion in an EDTA solution for 2 weeks and embedded in paraffin. Transverse sections were created through the center of the calvarial defect and subjected to hematoxylin and eosin (H&E) staining. Histological examination was conducted under a microscope, complemented using a morphometric system connected to a personal computer. Measurements of five random 1-mm² fields from 10 slices in each group were used to calculate the microvessel density, with each count performed three times by the same individual, and the mean value was determined.

Statistical analyses were carried out using SPSS 26 software (IBM Corp., Armonk, NY, USA). All variables were measurement data and presented as mean ± standard deviation. The Kolmogorov–Smirnov test was utilized to examine whether the data were normally distributed. For the variables of vessels with a normal distribution, the repeated measures ANOVA was performed, while the Friedman test was conducted for variables with a non-normal distribution. The bilateral test level was established at α = 0.05, and p-values less than 0.05 were deemed statistically significant.

Histological assessment of the specimens, conducted using H&E staining, revealed that, on day 7, nascent vessels had emerged from the periphery of the skull defect, accompanied by a small amount of connective tissue (Figures 2A, D) By day 14, the bone had begun to form in the defects, extending from the external bone surface toward the central region of the defect, coupled with a noticeable increase in the number of blood vessels (Figures 2B, E, G). At the 28-day mark, a heightened degree of irregular ossification was observed, with a thicker layer of connective tissue encompassing the critical-sized defects, particularly at the margins (Figures 2C, F). The microvessel density on day 28 showed a marked increase compared to day 7 (Figure 2G). However, there was no significant difference in the number of blood vessels on day 28 compared to day 14 (Figure 2G).

Initially, Figure 3A presents the schematic diagram of the areas imaged. As shown in the structure cross section in Figure 3B, which aligns with the green dotted line cross section in 3D OCTA (Figure 3E), the region between the yellow dotted lines marks the area of the critical-sized defect model. Due to the removal of the skull from the defect area in this model, the OCT probe light within the defect area was not attenuated by fibrous bone, thereby making deeper-layer structural information observable. An OCTA cross section was subsequently generated by analyzing the temporal dynamics of OCT scattering signals, resulting in the visualization of the enriched blood vessels in the deep layer. Figures 3C, D showcase the projection view of the surface and deep layers in the depth direction, representing, respectively, the surface (from the surface to a depth of 483 μm) and deep (from 483 μm to a depth of 967 μm) layers. As is evident, the vessel network in the surface tissue primarily consists of capillaries and neovascularization, while the vessel network at the deeper level comprises diving arterioles and ascending venules. A 3D perfusion map (Figure 3E) was further accomplished by stacking a series of sequential cross-sectional angiograms. Figure 3F shows the projection view of the 3D perfusion map in the depth direction, where the surface layer is represented by the pale yellow channel and the deep layer by the red-orange channel. Therefore, given its micron-scale vessel resolution and depth information, OCTA imaging merges as a robust technical method for monitoring 3D dynamic changes in the vascular structure within the critical-sized calvarial defect.

We monitored the process of angiogenesis in areas exhibiting critical-sized defects on days 7, 14, and 28 post-surgery, utilizing OCTA scanning (Figure 4). On day 7, diminutive blood vessels radiated from the lateral boundaries of the circular defect; however, the capillaries on one side barely connected with those on the opposite side (Figure 4A). By day 14, a conspicuous increase was observed in both the number and size of vessels, which demonstrated extensive convergence within the circular skull defect (Figure 4B). Conversely, the vascular images on day 28 did not reflect a continuous growth in the number and dimensions of vessels (Figure 4C).

Additionally, vascular changes were quantified objectively by OCTA in critical-sized models at distinct time stages. Microvascular density was assessed using the VAD metric, while vessel morphology was quantified using the MVDI (Figure 5). The calculated blood vessel density significantly increased to 0.509 on day 14 from 0.368 on day 7. However, the blood vessel density in the OCTA image significantly decreased to 0.374 on day 28 from that on day 14. Likewise, the indicators of the vessel diameter showed a significant increase to 771.50 µm on day 14 from 284.55 µm on day 7. The vessel diameter on day 28 (313.20 µm) significantly decreased compared to day 14.