Abstract
Blood flow imaging is widely applied in photodynamic therapy (PDT) to provide vascular morphological and statistical parameters. This approach relies on the intensity of time-domain signal differences between blood vessels and background tissues; therefore, it often ignores differences within the vasculature and cannot accommodate abundant structural information. This study proposes a multi-level optical angiography (MOA) method for PDT. It can enhance capillaries and image vessels at different levels by measuring the signal frequency shift associated with red blood cell motion. The experimental results regarding the PDT-induced chorioallantoic membrane model showed that the proposed method could not only perform multi-level angiography but also provide more accurate quantitative information regarding various vascular parameters. This MOA method has potential applications in PDT studies.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Photodynamic therapy (PDT), based on the interaction between the photosensitizer, molecular oxygen, and light, induces irreversible vasoconstriction, such as vasoconstriction and blood-flow stagnation, thereby damaging the target lesion [1]. It has been successfully employed for the therapy of vasculature-related abnormalities, such as esophageal cancer and port-wine stains (PWS) [2–4]. However, the therapeutic mechanisms and biological responses to vascular damage during PDT have not yet been fully explored. Therefore, the accurate monitoring of vascular changes is important for optimizing PDT doses and further elucidating its biological mechanism.
The characteristic indices of blood vessels include morphology, branching pattern, density, and diameter. Previous studies have demonstrated that changes in the morphology of blood vessels can be used to quantify the biological response after PDT treatment. Numerous approaches, such as magnetic resonance imaging (MRI), computed tomography (CT), and ultrasound, have been applied for long-term follow-up of vascular damage and quantification of vascular effects to further optimize the therapeutic effectiveness of PDT [5–7]. MRI requires a contrast agent to achieve angiography, whereas CT presents a radiation hazard to biological samples. Owing to resolution limitations, these two methods cannot be used to image microvessels. As a noninvasive imaging technique, ultrasound is mainly applied for the imaging of large vessels [8]. Recently, several high-resolution optical methods have been developed to monitor microvascular effects. For example, confocal microscopy and two-photon microscopy have been used to measure changes in the vascular structure during PDT [9,10]. However, these methods rely on exogenous contrast agents. Some noninvasive, label-free blood-flow imaging techniques, such as optical coherence tomography (OCT), have also been used to assess the in vivo vascular response induced by PDT [11,12]. OCT can acquire structural and functional information on blood flow and has become the gold standard for living human retinal imaging in the clinical setting. Such point-by-point scanning-based blood-flow imaging methods cannot simultaneously acquire synchronous signals at different locations and provide full-field information on blood microcirculation at the same time [13]. To address this issue, several full-field optical imaging methods have been developed for blood-flow imaging. Ren et al. used laser speckle contrast imaging (LSCI) to monitor the hemoperfusion dynamics of PWS during PDT [14]. However, studies have shown that different blood vessels respond differently to PDT. Therefore, it is important to evaluate the efficacy of PDT on different levels of the vasculature.
We developed multi-level optical angiography (MOA) to assess microvascular alterations during PDT with pyropheophorbide-alpha (Ppa) on a chorioallantoic membrane (CAM) in chicken embryos. Low-coherence light illumination was used in this study to achieve the sensitivity of low-coherence interference as well as the high contrast of absorption imaging between red blood cells (RBCs) and the background tissue. The endogenic RBC signals can be extracted by analyzing the modulation frequency. As vessels of different levels have differences in RBC motion, the resulting frequency-shift signal also differs. By measuring the signal frequency shift, the proposed method can image vessels at different levels and obtain more accurate vascular morphology information. Thus, this study has potential applications in the evaluation of PDT applied to vascular diseases.
2. Materials and methods
2.1 Blood vessel model for PDT
According to the procedure that was previously described [15], the eggs were cleaned using 70% alcohol on the first day of embryo development and kept in an incubator at a constant temperature of 37.8 °C and humidity of 75% with steady, slow half-rotations of 30 min each. On the third day, a small hole was drilled in the shell using a hand driller to remove 2 to 3 mL of albumin with a syringe, and the rotation was interrupted. A window of 2 cm2 was opened in the thinner part and sealed using adhesive tape. The eggs remained in the incubator until the sixth day when the vascular network in the CAM model matured for analysis with an evident difference in blood flow.
Pyropheophorbide-alpha (Ppa, Shanghai Yuanye Bio-Technology Co., Ltd., China) with a concentration of 0.01 mg/ml was topically administered to the vascular network in the CAM model. Before PDT, a 12 mm Teflon ring was used to delimit the target site in the CAM model. For PDT, 72 µL of Ppa solution was gently placed inside the ring using a pipette. Shortly after Ppa administration, the target site was irradiated by a 660 nm semiconductor laser with an irradiance of 30 mW/cm2.
2.2 Experimental setup for PDT monitoring
Vascular changes in the CAM model were monitored using the same experimental imaging setup used in a previous study [16]. This system was used for imaging before and after each stage of the experiment, including before adding Ppa, after adding Ppa (with no laser radiation, i.e., tr = 0 min), 1st PDT (with laser radiation for 3 min, i.e., tr = 3 min), and 2nd PDT (with laser radiation for 6 min, i.e., tr = 6 min). Low-coherence light from a customized fiber-coupled light emitting diode source (central wavelength ${\lambda _0}$ = 530 nm, bandwidth = 10 nm, power = 100 mW) was equally split into four collimated beams, and it illuminated the sample via a 1 × 4 fiber splitter. A series of 500 raw blood flow images with a resolution of 600 × 600 pixels were captured using a high-performance telecentric lens (magnification:1.7×, #63-232, depth of field: 0.5 mm, Edmund Optics) and a high-speed complementary-metal–oxide–semiconductor camera (acA2000-340 km, Basler, Germany). The exposure time and sampling rate of the camera were set as 700 µs and 92 fps, respectively. The camera had a pixel size of 5.5 µm × 5.5 µm. During the experiments, the biological samples were handled carefully in accordance with the laboratory animal protocol approved by the Institutional Animal Care and Use Committee of Foshan University.
2.3 Multi-level optical angiography
A low-coherence light source with a central wavelength of 530 nm was selected to perform MOA. The absorption coefficient of the RBCs was much higher than that of the background. Therefore, the signal of any pixel recorded by a camera can be viewed as a temporal sequence of high-absorption signals corresponding to RBC signals and low-absorption signals corresponding to the signals from “gaps” (background tissue fluid). The acquired raw signal can be expressed as
where ${I_0}({x,y} )$ is the signal from the background tissue, which does not vary with time, and ${I_N}({x,y,t} )$ represents the noise signal from the system. ${I_{RBC}}({x,y,t} )$ is the signal from moving RBCs. The raw temporal signal is transferred from the time domain to the frequency domain using a fast Fourier transform (FFT), which can be expressed as follows:When a low-coherence light beam illuminates moving RBCs, a time-dependent fluctuating intensity with a shifted frequency can be observed. The modulation frequency signal introduced by the motion of particles is positively associated with the velocity of the particles. Because the blood flow in different types of blood vessels varies in velocity, a multi-level optical angiogram can be obtained by frequency division.
2.4 Quantitative analysis
To quantify the performance of multi-level angiography, four major vascular structural parameters were extracted: vascular area density (VAD), vascular linear density (VLD), vessel diameter (VD), and the number of vessels (NV). Quantitative vascular analysis can mainly be divided into three parts: vessel segmentation, vascular centerline extraction, and quantitative evaluation of vascular parameters. A flowchart of vessel segmentation is shown in Fig. 1. We first used the K-means algorithm [17] to coarsely segment the vessels in the MOA image. The K-means clustering method uses an iterative refinement algorithm to cluster pixels into K subsets. The minimized objective function is defined as follows:
VAD is defined as the ratio of the number of vascular pixels to that of the vascular segmentation image, which can be expressed as follows [23]:
3. Results
3.1 Principle validation
To experimentally validate the principle of our multi-level angiography approach, an experiment was conducted using a 6-day-old CAM model. Our method assumes that MOA on living biological samples depends on blood-flow velocity. Differences in blood-flow velocity lead to different frequency distributions [24]. The proposed technique can perform multi-level angiography, which is attributed to the frequency selection of the blood-flow signals. The raw signals of the different components were analyzed, and the spectra and quantitative data are shown in Fig. 2 and Table 1. The frequency ranges of F1–F10, which are marked in Fig. 2, are $({ {0\sim 4.6} ]} \; \textrm{Hz}$, $({ {4.6\sim 7.2} ]\; } \textrm{Hz}$, …, $({ {4.6({n - 1} )\sim 4.6n} ]} \; \textrm{Hz}\; ({n = 10} )$, respectively. Because the background tissue is not affected by RBCs, signals scattered by the background tissue have the highest intensity (line A in Fig. 2(b)). Simultaneously, the existence of random noise causes the background signal to fluctuate randomly, which also leads to its frequency-domain signal being mainly distributed in the zero-frequency range (Fig. 2(c)). For capillaries with the intermittent flow of a few RBCs, the averaged intensity of the raw signal (line B in Fig. 2(b)) is similar to that of the background. As shown in Fig. 2(d), the frequency-domain signal of the capillaries has an obvious envelope from 0 to 4.6 Hz, which makes the capillaries only exist in the F1 image. The temporal signal intensity of the secondary blood vessels (line C in Fig. 2(b)) is lower than that of the capillaries (line B in Fig. 2(b)) because of the higher RBC concentration. The distribution range of the secondary vascular signals in the frequency domain (Fig. 2(e)) is wider than that of the capillaries (Fig. 2(d)). Therefore, secondary vessels appear in the F1–F6 image. For large blood vessels, owing to the higher RBC concentration and movement speed, the scattered signal has a lower intensity (line D in Fig. 2(b)), and the corresponding frequency-domain signal is distributed in the entire frequency domain (Fig. 2(f)). When the frequency-domain signal in the first level window is used to obtain an F1 image, the large blood vessels are suppressed. The same conclusion can also be obtained from quantitative data (Table 1).
3.2 Multi-level optical angiography
To experimentally validate the feasibility of our approach, the data in Fig. 2 is shown in Fig. 3, where a representative raw image of the CAM model is shown in Fig. 3(a), and the AMD image is shown in Fig. 3(b). Some capillaries have a poor resolution in the AMD images because of the lower concentration of RBCs. Figures 3(c)–(l) show the MOA images with high resolution obtained using Eq. (3). These images differ in their ability to represent different types of blood vessels: the F1 image (Fig. 3(c)) shows a high-resolution capillary network and does not show large vessels; only secondary vessels and large vessels were obtained in the F2–F6 image (Fig. 3(d)–(h)), and the F7–F10 image (Fig. 3(i)–(l)) primarily focused on large vessels. This may be caused by the difference in velocity between vessels, which usually manifests in different high-frequency signal distributions.
We also performed experiments on 3-day-old chicken embryos, and the results are shown in Fig. 4. Figure 4(a) and (b) show the raw image collected by the camera and the AMD image, respectively. The corresponding MOA images are shown in Fig. 4(c)–(l). As the stage of development advances, the peak velocity of blood flow and the complexity of the vascular network increase [25,26]. Compared with 6-day-old chicken embryo samples (Fig. 3), the vascular networks and velocity distributions of the 3-day-old chicken embryos were simpler. In other words, the modulation frequency signals of the vessels were similar. Therefore, compared with the results shown in Fig. 3(c)–(l), the multi-level imaging performance of our approach in this experiment was weak. For MOA images with adjacent levels, the vascular distribution remains similar even in the low-frequency range, as in Fig. 4(c) and (d). When the differences in blood flow levels are not significant, reducing the number of levels can ensure the imaging of multiple levels of vessels and improve computational efficiency. For example, for the 3-day chicken embryo results, the number of levels can be reduced from 10 to 5, and the computational efficiency will be doubled.
To show the superposition effect of multi-level blood vessels, a set of multi-level MOA fusion images (F1, F2, F6, and F10) are provided in Fig. 5(a) and (b), where the data is from Figs. 3 and 4, respectively. Interestingly, the higher the angiogram level, the smaller the VD of large vessels. The modulated frequency signal introduced by RBC motion is positively correlated with blood flow velocity, but the imaging parameter of MOA does not reflect the true velocity distribution. The MOA method obtains a projected image of blood flow in a multi-level frequency range. We can measure the superposition of RBCs at different depths in large vessels or the superposition of RBC motion in capillaries at different depths. For large vessels, the velocity in the centerline of the vessel is greater than the velocity around the vessel wall. Therefore, the diameter of large vessels in MOA images decreases as the frequency range increases (Fig. 5). For capillaries, the blood flow velocities are similar; therefore, the superposition of RBCs at different depths does not significantly affect imaging. In addition, theoretically, imaging is affected by out-of-focus blood flow information. However, the impact on imaging is marginal because the defocused blood flow is located at deeper depths, where its signal intensity is low and dynamic information is difficult to display.
3.3 Multi-level optical angiography for photodynamic therapy
To evaluate the performance of MOA in monitoring vascular effects, a PDT test was performed on the CAM model. The target site with an area of $2.0\; \textrm{mm} \times 2.1\; \textrm{mm}$ ($620 \times 650$ pixels) in the CAM model was employed for imaging. A high-speed camera recorded raw images before Ppa addition, after Ppa addition, and at tr = 3 min and 6 min. Significant vessel damage was observed in the CAM model under the Ppa-mediated PDT; small blood vessels in the irradiated field were absent, and the larger vessels were constricted and static [27,28]. Multi-level angiograms of this experiment are shown in Fig. 6(a)–(d), illustrating the changes in smaller blood vessels in the four abovementioned PDT stages. Combined with the corresponding partial magnification images (Fig. 6(e)–(h)), it is evident that the diameter and density of the blood vessels decreased, which suggests that they were destroyed. Furthermore, F5 images under different conditions (Fig. 6(i)–(l)) show that the blood flow and diameter of larger vessels decreased owing to vascular damage, such as coagulation, vasoconstriction, and blood stasis. The vascular changes observed in the CAM model are consistent with the results of Liu et al. [29]. Therefore, the proposed blood-flow imaging method can accurately reveal the vascular effects of PDT.
3.4 Quantitative measurement of morphological parameters
In light of our findings on angiograms at different levels, an advantage of the MOA algorithm is identified, which is the measurement of vascular morphological parameters by modulating frequency-shift signals from different blood vessels. Multi-level angiograms and AMD images from the first experiment in Section 3.2 were quantitatively evaluated to compare their ability to reflect the vascular parameters. In this study, the angiograms obtained using the AIFM method are only segmented on the basis of AMD images, and the segmentation method was consistent with that of the MOA image. Figure 2 shows that the AIFM method can obtain an angiogram with higher contrast than the raw image, but the capillaries embody poor contrast. MOA can not only perform hierarchical imaging of blood vessels but can also improve the contrast of capillaries. We quantified the performance of the compared methods with regard to blood-flow imaging based on the four previously mentioned evaluation metrics: VAD, VLD, VD, and NV (Table 2). The existence of large blood vessels affects the vascular segmentation results and restricts the vascular morphological parameters of the AIFM. As shown in Table 2, the MOA method can provide higher VAD, VLD, and NV parameters while inhibiting large vessels in the F1 image, indicating that it can obtain more capillary and secondary vessel information. For large vessels, the VD obtained from the F2 images is closer to the actual value than that obtained using the AIFM method. In addition, MOA images show different vessel diameters, which can further indicate the velocity distribution of blood flow. In conclusion, the proposed method can be used to obtain more vascular information.
In Section 3.3, we demonstrate that the MOA method can be used to perform hemodynamic monitoring of the CAM model based on PDT. We conducted a quantitative analysis of the results shown in Fig. 6 to explore the capability of the proposed method to quantify vascular effects, and the quantitative results are shown in Fig. 7. Blood vessel growth under different conditions was recorded using the four previously mentioned quantified parameters. Among them, the VAD, VLD, and NV parameters in the MOA method were measured from the region of interest (Fig. 6(e)–(h)) in the F1 images (Fig. 6(a)–(d)), and the VD parameter of the large vessel was obtained from F2 images. After the first PDT (tr = 3 min), the VAD, VLD, and NV in the ROIs decreased by 24.47%, 20.34%, and 21.30%, respectively. Furthermore, the second PDT treatment reduced these parameters by 35.54%, 35.81%, and 37.06%, respectively. The reason for this phenomenon is that most capillaries and sub-vessels were damaged after the second PDT treatment. In addition, an interesting phenomenon was observed: the large vessels conformed to severe contractions in the first PDT treatment, and as the treatment continued, relaxation of the large blood vessels occurred (Fig. 7(d)). This is consistent with the results of Fyodorov et al. [30]. It can also be seen from Fig. 7(a)–(c) that compared with AIFM, the results obtained by MOA are closer to the manual labeling results, which indicates that the proposed method can better reflect the vascular effect during PDT.
4. Discussion and conclusion
This paper presents an MOA method for evaluating vascular grading in live chicken embryo samples. The capillaries, secondary vessels, and large vessels were reconstructed based on the selective modulation of endogenous hemodynamic characteristics from low-coherence speckle. The proposed method was used to study the vascular effects in the CAM model for PDT. The experimental results demonstrate that the MOA can provide quantitative structural information at different vascular levels. As such, this method can promote the study of vascular mechanisms and the selection of treatment parameters for PDT therapy.
We noticed that different types of vessel distribution respond differently to the same imaging system parameter settings, such as magnification, pixel size, sampling rate, exposure time, and the number of levels. The parameter settings of the imaging system can be divided into two categories, one for spatial parameters and the other for temporal parameters. In our study, the spatial parameters were as follows: the magnification of the lens was 1.7, and the pixel size of the camera was $5.5\; \mathrm{\mu} \textrm{m}\; \times {\; }5.5\; \mathrm{\mu} \textrm{m}$. An RBC with a diameter of 7 µm occupies approximately 4 pixels, which ensures that the camera can capture the fluctuation of the absorption intensity caused by the RBC motion. The temporal parameters, which were the exposure time and sampling rate of the camera, were set to 700 µs and 92 fps, respectively. When the flow velocity was less than 5 mm/s, the camera could capture the intensity fluctuation signal with high contrast. When the flow velocity is greater than 15 mm/s, the intensity fluctuations are blurred due to the relatively long integration time set by the camera. PDT acts extensively on capillaries and small vessels with flow velocities less than 15 mm/s [31]. Therefore, the current exposure time can theoretically guarantee the use of MOA in clinical applications. When the sampling rate and number of levels are set to 92 fps and 10, respectively, the frequency-domain resolution of each level is 4.6 Hz, corresponding to 0.47 mm/s. When the flow velocity is higher than the maximum measurement velocity of 4.7 mm/s, as in the signal in Fig. 2(f), the blood flow signal is randomly distributed among the frequency domain signals of each level. For the 3-day-old CAM model, the capillary network with the low-velocity flow has not yet grown. Therefore, even if ten frequency-domain levels are set, only the grown main vessels and secondary vessels can be resolved. However, for the 6-day-old CAM model, capillaries, secondary vessels, and main vessels can be distinguished.
In basic PDT studies, MOA can provide a label-free, high-quality image of the vasculature compared to existing methods such as OCT and display blood flows at all levels. This significantly improves the accuracy of quantifying the morphological parameters of the vasculature and is useful in assessing the amount of photosensitizer injected and the effect of photodynamic forces on the vasculature, especially the capillaries [32,33]. However, some of its limitations should also be mentioned. When biological samples change, the frequency-domain levels should be redefined according to the vascular distribution. For example, for the 3-day-old CAM model, five frequency-domain levels can replace ten frequency-domain levels to improve computational efficiency. In addition, the sampling rate can be increased to improve the higher level of blood vessel classification (faster flow velocity). In future research, we will further investigate the optimal temporal parameters selection of MOA. The imaging based on absorption fluctuation is only suitable for near-transparent samples, such as chicken embryos and zebrafish. For turbid tissues, the raw absorption image will be dimmed due to high scattering, and its signal-to-noise ratio will be reduced. For biological research, this limitation will be addressed in future work by using optical clearing technology. Currently, MOA can be used to study the therapeutic mechanisms and biological responses to vascular injury during PDT. However, this approach has potential in clinical applications, such as those involving skin, ophthalmology, and other organs (in combination with endoscopy). To the best of our knowledge, OCT and LSCI, which have limitations regarding penetration depth and samples, have been used in these clinical applications [34,35]. In addition, multi-level speckle angiography should be possible using a multi-level imaging principle similar to MOA with coherent light instead of low-coherence light, depending on the fluctuating characteristics of speckle intensity.
Funding
Key-Area Research and Development Program of Guangdong Province (2020B1111040001); National Natural Science Foundation of China (62075042, 62271148); Research Fund of Guangdong-Hong Kong-Macao Joint Laboratory for Intelligent Micro-Nano Optoelectronic Technology (2020B1212030010).
Disclosures
The authors declare no conflicts of interest.
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.
References
1. H. Wu, T. Minamide, and T. Yano, “Role of photodynamic therapy in the treatment of esophageal cancer,” Dig Endosc 31(5), 508–516 (2019). [CrossRef]
2. S. J. Bakri, J. E. Thorne, A. C. Ho, J. P. Ehlers, S. D. Schoenberger, S. Yeh, and S. J. Kim, “Safety and efficacy of anti-vascular endothelial growth factor therapies for neovascular age-related macular degeneration: a report by the American Academy of Ophthalmology,” Ophthalmology 126(1), 55–63 (2019). [CrossRef]
3. D. Chen, J. Ren, Y. Wang, H. Zhao, B. Li, and Y. Gu, “Relationship between the blood perfusion values determined by laser speckle imaging and laser Doppler imaging in normal skin and port wine stains,” Photodiagn. Photodyn. Ther. 13, 1–9 (2016). [CrossRef]
4. V. Gottfried, R. Davidi, C. Averguj, and S. Kimel, “In vivo damage to chorioallantoic membrane blood vessels by porphycene-induced photodynamic therapy,” J. Photochem. Photobiol. B, Biol. 30(2-3), 115–121 (1995). [CrossRef]
5. K. E. Emblem, K. Mouridsen, A. Bjornerud, C. T. Farrar, D. Jennings, R. J. Borra, P. Y. Wen, P. Ivy, T. T. Batchelor, and B. R. Rosen, “Vessel architectural imaging identifies cancer patient responders to anti-angiogenic therapy,” Nat. Med. 19(9), 1178–1183 (2013). [CrossRef]
6. D. Sampath, J. Oeh, S. K. Wyatt, T. C. Cao, H. Koeppen, J. Eastham-Anderson, L. Robillard, C. C. K. Ho, J. Ross, G. Zhuang, H. B. Reslan, P. Vitorino, K. H. Barck, S. E. Ungersma, J. M. Vernes, M. Caunt, N. V. Bruggen, W. Ye, U. Vijapurkar, Y. G. Meng, N. Ferrara, L. S. Friedman, and R. A. D. Carano, “Multimodal microvascular imaging reveals that selective inhibition of class I PI3 K is sufficient to induce an antivascular response,” Neoplasia 15(7), 694–IN4 (2013). [CrossRef]
7. B. J. Vakoc, R. M. Lanning, J. A. Tyrrell, T. P. Padera, L. A. Bartlett, T. Stylianopoulos, L. L. Munn, G. J. Tearney, D. Fukumura, R. K. Jain, and B. E. Bouma, “Three-dimensional microscopy of the tumor microenvironment in vivo using optical frequency domain imaging,” Nat. Med. 15(10), 1219–1223 (2009). [CrossRef]
8. J. H. Cha, W. K. Moon, N. Cho, S. Y. Chung, S. H. Park, J. M. Park, B. K. Han, Y. H. Choe, G. Cho, and J. G. Im, “Differentiation of benign from malignant solid breast masses: conventional US versus spatial compound imaging,” Radiology 237(3), 841–846 (2005). [CrossRef]
9. T. Suzuki, M. Tanaka, M. Sasaki, H. Ichikawa, H. Nishie, and H. Kataoka, “Vascular shutdown by photodynamic therapy using talaporfin sodium,” Cancers 12(9), 2369 (2020). [CrossRef]
10. L. Hu, Z. Chen, Y. Liu, B. Tian, T. Guo, R. Liu, C. Wang, and L. Ying, “In vivo bioimaging and photodynamic therapy based on two-photon fluorescent conjugated polymers containing dibenzothiophene-S, S-dioxide derivatives,” ACS Appl. Mater. Interfaces 12(51), 57281–57289 (2020). [CrossRef]
11. D. Chen, W. Yuan, H. Park, and X. Li, “In vivo assessment of vascular-targeted photodynamic therapy effects on tumor microvasculature using ultrahigh-resolution functional optical coherence tomography,” Biomed. Opt. Express 11(8), 4316–4325 (2020). [CrossRef]
12. L. M. Smid, K. A. Vermeer, K. T. Wong, J. P. M. Ciriano, J. H. de Jong, V. Davidoiu, J. F. de Boer, and M. E. J. van Velthoven, “Detailed optical coherence tomography angiographic short-term response of type 3 neovascularization to combined treatment with photodynamic therapy and intravitreal bevacizumab,” Acta Ophthalmol. 99(2), 207–214 (2021). [CrossRef]
13. M. Yi, L. Wu, Q. Du, C. Guan, M. Liu, X. Li, H. Xiong, H. Tan, X. Wang, J. Zhong, D. Han, M. Wang, and Y. Zeng, “Spatiotemporal absorption fluctuation imaging based on U-Net,” J. Biomed. Opt. 27(2), 026002 (2022). [CrossRef]
14. J. Ren, P. Li, H. Zhao, D. Chen, J. Zhen, Y. Wang, Y. Wang, and Y. Gu, “Assessment of tissue perfusion changes in port wine stains after vascular targeted photodynamic therapy: a short-term follow-up study,” Lasers Med. Sci. 29(2), 781–788 (2014). [CrossRef]
15. H. H. Buzzá, L. C. Fialho de Freitas, L. T. Moriyama, R. G. Teixeira Rosa, V. S. Bagnato, and C. Kurachi, “Vascular effects of photodynamic therapy with curcumin in a chorioallantoic membrane model,” Int. J. Mol. Sci. 20(5), 1084 (2019). [CrossRef]
16. C. Guan, M. Yi, Q. Du, H. Xiong, H. Tan, M. Wang, and Y. Zeng, “Full-field optical multi-functional angiography based on endogenous hemodynamic characteristics,” J. Biophotonics 14(5), e202000411 (2021). [CrossRef]
17. P. Hilty, P. Guerci, Y. Ince, F. Toraman, and C. Ince, “MicroTools enables automated quantification of capillary density and red blood cell velocity in handheld vital microscopy,” Commun. Biol. 2(1), 217 (2019). [CrossRef]
18. N. Otsu, “A threshold selection method from gray-level histograms,” IEEE Trans. Syst. Man. Cybern. 9(1), 62–66 (1979). [CrossRef]
19. F. Zhang, M. Wang, D. Han, H. Tan, G. Yang, and Y. Zeng, “In vivo full-field functional optical hemocytometer,” J. Biophotonics 11(2), e201700039 (2018). [CrossRef]
20. A. Y. Yuan, Y. Gao, L. Peng, L. Zhou, J. Liu, S. Zhu, and W. Song, “Hybrid deep learning network for vascular segmentation in photoacoustic imaging,” Biomed. Opt. Express 11(11), 6445–6457 (2020). [CrossRef]
21. Z. Ma, N. Ding, Y. Yu, Y. Ma, X. Yuan, Y. Wang, Y. Zhao, J. Luan, and J. Liu, “Quantification of cerebral vascular perfusion density via optical coherence tomography based on locally adaptive regional growth,” Appl. Opt. 57(35), 10117–10124 (2018). [CrossRef]
22. N. Wu, L. Wang, B. Zhu, C. Guan, M. Wang, D. Han, H. Tan, and Y. Zeng, “Wide-field absolute transverse blood flow velocity mapping in vessel centerline,” J. Biomed. Opt. 23(2), 026008 (2018). [CrossRef]
23. A. Y. Kim, Z. Chu, A. Shahidzadeh, R. K. Wang, C. A. Puliafito, and A. H. Kashani, “Quantifying microvascular density and morphology in diabetic retinopathy using spectral-domain optical coherence tomography angiography,” Invest. Ophthalmol. Visual Sci. 57(9), OCT362 (2016). [CrossRef]
24. M. Wang, W. Mao, C. Guan, G. Feng, H. Tan, D. Han, and Y. Zeng, “Full-field functional optical angiography,” Opt. Lett. 42(3), 635–638 (2017). [CrossRef]
25. A. M. Oosterbaan, N. T. C. Ursem, P. C. Struijk, J. G. Bosch, A. F. W. van der Steen, and E. A. P. Steegers, “Doppler flow velocity waveforms in the embryonic chicken heart at developmental stages corresponding to 5-8 weeks of human gestation,” Ultrasound Obstet. Gynecol. 33(6), 638–644 (2009). [CrossRef]
26. M. Maibier, B. Reglin, B. Nitzsche, W. Xiang, W. Rong, B. Hoffmann, V. Djonov, T. W. Secomb, and A. R. Pries, “Structure and hemodynamics of vascular networks in the chorioallantoic membrane of the chicken,” Am. J. Physiol. Heart Circ. Physiol. 311(4), H913–H926 (2016). [CrossRef]
27. H. H. Buzzá, L. V. Silva, L. T. Moriyama, V. S. Bagnato, and C. Kurachi, “Evaluation of vascular effect of photodynamic therapy in chorioallantoic membrane using different photosensitizers,” J. Photochem. Photobiol. B, Biol. 138(1), 1–7 (2014). [CrossRef]
28. X. Chen, R. Zheng, L. Zhao, R. Kong, N. Yang, Y. Liu, A. Chen, C. Wang, H. Cheng, and S. Li, “Photodynamic therapy initiated immunotherapy of self-delivery re-educator by inducing immunogenic cell death and macrophage polarization,” Chem. Eng. J. 435, 134783 (2022). [CrossRef]
29. Q. Liu, S. Zhou, Z. Zhang, and Q. Luo, “Laser speckle contrast imaging: monitoring blood flow dynamics and vascular structure of photodynamic therapy,” in Optics in Health Care and Biomedical Optics: Diagnostics and Treatment II (2005), pp. 26–33.
30. S. N. Fyodorov, V. G. Kopaeva, J. B. Andreev, G. V. Ponomarev, and H. M. Suchin, “Vascular effect of photodynamic therapy,” Proc. SPIE 2728, 214–222 (1996). [CrossRef]
31. A. A. Linninger, I. G. Gould, T. Marinnan, C. Y. Hsu, M. Chojecki, and A. Alaraj, “Cerebral microcirculation and oxygen tension in the human secondary cortex,” Ann. Biomed. Eng. 41(11), 2264–2284 (2013). [CrossRef]
32. X. Xu, L. Lin, and B. Li, “Automatic protocol for quantifying the vasoconstriction in blood vessel images,” Biomed. Opt. Express 11(4), 2122–2136 (2020). [CrossRef]
33. R. Gao, H. Xu, L. Liu, Y. Zhang, T. Y. H. Zhou, M. Sun, N. Chen, Y. Ren, T. Chen, Y. Pan, M. Zheng, T. Y. Ohulchanskyy, R. Zheng, L. Cai, L. Song, J. Qu, and C. Liu, “Photoacoustic visualization of fluence rate dependence of photodynamic therapy,” Biomed. Opt. Express 11(8), 4203–4223 (2020). [CrossRef]
34. E. A. Swanson and J. G. Fujimoto, “Registration of free-hand OCT daughter endoscopy to 3D organ reconstruction,” Biomed. Opt. Express 7(12), 4995–5009 (2016). [CrossRef]
35. C. Regan, C. Hayakawa, and B. Choi, “Momentum transfer Monte Carlo for the simulation of laser speckle imaging and its application in the skin,” Biomed. Opt. Express 8(12), 5708–5723 (2017). [CrossRef]