1. Introduction

In recent years, the advancement of biomedical imaging technologies has significantly contributed to our understanding of complex physiological processes and the diagnosis of various health conditions [1–3]. Understanding the intricate cerebral hemodynamics of rodents is an important issue of concern in the fields of neuroscience, physiology, psychology and imaging [4,5]. Traditional imaging techniques often require the subjects to remain still or be under anesthesia [6], which restricts the study of genuine dynamic physiological processes and natural behaviors. Therefore, the development of wearable probes capable of capturing high-resolution images from the brain in freely moving and awake subjects has the potential to revolutionize neuroscience research [7,8], especially in studying cerebral blood flow, neuronal activities, and vascular dynamics.

Recently, various optical imaging techniques such as photoacoustic microscopy (PAM) [9], two-photon (2P) microscopy [10] and laser speckle contrast imaging (LSCI) [11], have been developed to investigate neurovascular relationships in freely moving animals. However, PAM faces challenges with its axial resolution, which may extend to the range of hundreds of micrometers. 2P struggles to assess large-scale blood flow due to its slow three-dimensional (3D) scan speeds and limited FOV. LSCI suffers from limited spatial resolution, being limited to two-dimensional imaging and lacking depth-resolved information. In addition to optical imaging, functional ultrasound imaging has the capability to monitor brain activities in freely moving rodents with high temporal resolution [12]. Regrettably, its spatial resolution falls short of capturing microscopic neurovascular events and the image obtained suffers from poor contrast.

Optical coherence tomography (OCT) is an emerging optical imaging modality that enables three-dimensional (3D) volumetric imaging of biological tissue microstructure [13]. Because OCT is a non-invasive, contrast agent and radiation free technique, it shows high promise across a wide range of medical and biological applications. OCT angiography (OCTA) is the application of OCT for measuring and visualizing 3D mapping of blood perfusion [14], achieved by mathematically analyzing the red blood cells’ motion-induced temporal changes of scattering signals. Compared with other optical techniques, OCTA allows rapid, high-sensitivity, contrast-free imaging with micron-scale resolution, making it a promising choice for large FOV applications like brain blood flow monitoring [15].

In order to achieve miniaturization of the device and expand its applications range, some researchers have developed handheld OCT probes [16], to a certain extent, it has achieved structural compactness, portability, and mobility. This design allows operators greater freedom to control the position and orientation of the probe to adapt to the needs of different samples. Some researchers have replaced the traditional galvo scanning mirror in handheld probes with micro-electromechanical systems (MEMS) mirror [17,18], further reducing the size and weight of the probe. Such handheld probes have achieved inspiring progress in fields such as ophthalmology [19], otolaryngology [20], dentistry [17], and dermatology [18]. However, the overall weight of handheld probes is relatively heavy, and for three-dimensional scanning, holding the probe by hand can lead to insufficient stability. As a result, it is not feasible to perform prolonged observation of rodents’ cerebral hemodynamics in a naturally awake state. Hence, developing a wearable OCTA probe becomes a promising solution for long-term imaging in freely moving mice. However, this is a challenging task. The main challenge lies in stability maintaining during usage, because OCTA involves multiple scans of the same location, utilizing the differences between scans to acquire vascular distribution, any presence of motion during scanning can greatly affect the calculation results. Another challenge is the optimization of the design for the miniature probe, ensuring high image quality, high resolution, and ease of assembly and adjustment during experimental processes. Furthermore, there’s also a need to achieve overall lightweight construction of the probe. Currently, there have been no reports of wearable OCTA probes used in freely moving mice.

In this study, we have developed a novel wearable OCT probe, which, to the best of our knowledge, is the first wearable OCTA probe that enables long-term monitoring of mouse cerebral blood flow. We demonstrated the feasibility of high-speed non-invasive imaging of the freely moving mouse brain under natural conditions, and quantitatively analyzed the differences in mouse brain vasculature between awake and anesthetized states. Additionally, we utilized this probe to monitor the cerebral hemodynamics during the acute ischemic phase after photothrombotic stroke for 4 hours, highlighting the reliability and long-term stability of our probe. We chose 4 hours monitoring because the ideal treatment window of thrombolysis is 4.5 hours following the onset of ischemia [21], therefore, providing visual and mechanistic insights into the early phase of stroke is of great importance. This work not only enhances our understanding of brain health and neurological disorders, but also paves the way for applications in early disease detection, personalized medicine, and the assessment of treatment responses, contributing to a possible new era of wearable biomedical imaging.

2. Materials and methods

2.1 System overview

The wearable OCT probe and our homebuilt spectral domain OCT (SD-OCT) system are shown in Fig. 1(a). The light source is a broadband super luminescent diode (SLD, S5FC1021S, Thorlabs, 1310 ± 40 nm, 12.5 mW). A 2 × 2 coupler (BXC25, Thorlabs) was used to divide the light into the sample arm and reference arm. The reflected light from the two arms was directed to a high-speed spectrometer (Cobra 1300, Wasatch Photonics), which had a detection capability spanning from 1175 nm to 1420 nm with a maximum speed of 76 kHz. The system was set up to capture A-lines at a frequency of 32 kHz, with each B-scan consisting of 1024 A-lines with a depth of 1024 pixels.

 

Fig. 1. Experimental system configuration and its performance. (a) Setup of the imaging system and a detailed optical configuration of the wearable OCT probe. SLD: superluminescent diode; MEMS: micro-electromechanical systems. (b) Experimental estimation of the axial resolution. (c) Experimental estimation of the lateral resolution. The smallest lines which can be resolved are group 6, element 4. This corresponds to 90.5 line pairs (lp/mm) and a lateral resolution of approximately 5.5 µm.

Download Full Size | PDF

In our customized wearable probe, the beam from the optical fiber was collimated by a collimator and then directed to a two-axis MEMS scanning mirror (A7B1.1, Mirrorcle Tech) which was put on the front focal plane of an achromatic doublet lens (GCL-010618; diameter, 12.7 mm; focal length, 15 mm) for telecentric scanning. This achromatic lens consists of a positive lens with a low refractive index (crown glass) and a negative lens with a high refractive index (flint glass), which is computer-optimized and designed to minimize chromatic aberration and correct spherical aberration to the greatest extent. The utilization of a two-axis MEMS scanner is the key to reducing the overall probe volume. For a traditional galvo-scanner pair, the X and Y scan mirrors are separated, which is not conducive to miniaturization. The MEMS scanner incorporates a single scanning mirror capable of deflection in both the X and Y directions, sharing a common pivot point, which enables dual telecentric scanning in a much smaller space. Mechanically, our compact design considers the miniaturization of the probe while maintaining high optical performance. The lightweight construction ensures that the probe does not hinder the natural behavior of freely moving mice, thus contributing to reliable imaging under natural conditions. The miniaturized probe consists of only three key components: the collimator, the MEMS mirror, and the objective. These components are integrated through a customized lightweight one-piece 3D printed mount which is specifically designed and precisely machined to simplify assembly and alignment of the components as well as removal and replacement. This one-piece 3D printed mount enhances the stability of the probe during mouse movement, and minimizes motion artifacts during imaging. Our probe measures 31 × 15 × 25 mm and weighs 8 g. The lens barrel deploys a lens and a precision-machined lens spacer for easy adjustment. The probe is connected to a fixed aluminum post on the mouse head by threads and precision optical spacers, which facilitate the assembly and adjustment of the probe during experiments. These features ensure that the probe maintains optimal performance during imaging. The stability of the probe, the ease of assembly and adjustment, and the consideration of motion artifacts highlight its suitability for wearable imaging. Extensive experimental validation was performed to demonstrate the effectiveness of our wearable probe, a photograph of the probe is shown at bottom right of Fig. 1(a), Visualization 1 and Visualization 2 in the Supplementary Materials show the behavior of a mouse wearing the probe.

The axial resolution was evaluated by measuring the signal of a mirror, shown in Fig. 1(b), the full width at half-maximum (FWHM) was measured to be 12 µm. To evaluate the lateral resolution, we imaged the 1951 USAF resolution target, the smallest lines which can be resolved are group 6, element 4. This corresponds to 90.5 line pairs (lp/mm) and a lateral resolution of approximately 5.5 µm.

2.2 Animal preparation

All procedures involving mice were approved by the Institutional Animal Care and Use Committee of Tsinghua University. Wild-type mice (C57BL/6, 9 weeks to 3 months old; weight, 20-30 g) were used in all the experiments.

The craniotomy surgery was performed in the following manner: Initially, mice were anesthetized with isoflurane (3% for induction, 1.5% during surgery). Subsequently, a toe pinch test was carried out to ensure the depth of anesthesia. After confirming the mouse's proper anesthesia, it was carefully positioned on a stereotactic frame while maintaining its body temperature with a heating pad. Before commencing the surgery, ophthalmic ointment was applied topically to safeguard the corneal surfaces from drying and potential damage due to extended exposure to the illuminating light source. A scalp incision was made, followed by conducting a craniotomy using a drill [22]. Once the skull was carefully removed, the brain was covered with 2% agarose and then a cover glass (with a thickness of 0.17 mm and a diameter of 6 mm) was affixed to the skull to minimize any motion of the exposed brain. Additionally, a custom aluminum head post was securely attached to the skull using dental cement.

2.3 Photothrombotic vascular occlusion

After the mice recovered from the craniotomy surgery, photothrombotic (PT) occlusion was induced as follows: An intraperitoneal injection of Rose Bengal solution (Sigma-Aldrich, 2.5 mg per 100 g body weight, at a concentration of 7.5 mg/mL) was administered to the mouse. Following the injection, the mouse was provided with a 30-minute resting period to facilitate the absorption of the solution into the systemic circulation. Then the occlusion was induced by focal illumination (1 mm diameter focal spot, 30 mW/mm²) with 532 nm laser (MSL-III-532, CNI) on the middle cerebral arteries (MCA) of the mouse brain for 30 minutes [23]. After PT occlusion was induced, we securely attached our wearable probe to the mouse's head and conducted continuous monitoring of post-stroke cerebral hemodynamics for up to four hours.

2.4 Data acquisition and processing

In this work, we employed a stepwise raster scanning procedure to collect volumetric data. The slow scanner (y) was driven by a step waveform with a total of 800 steps. At each step, five consecutive B-scans were acquired to analyze dynamic flow signals. Each of these B-scans consisted of 1024 A-lines in the fast-scanning direction. Adjacent B-scans were paired for blood flow extraction at each scanning step y. The original spectral interference fringe signal S (k, x, t) underwent a Fourier transformation along the wavenumber direction, resulting in the backscattered profile C_OCT(z, x, t) in the depth domain. This profile included both amplitude I and phase Φ information:

The cross-sectional angiogram was then created by subtracting the amplitude component I (z, x, t) between paired B-scans:

Then the OCTA signals at each step y were averaged to improve flow contrast. Performing this process for all y values, OCTA volumetric data containing microvasculature flow information can be obtained. The en face angiograms are obtained by preforming maximum intensity projection (MIP) along the axial direction.

3. Results

3.1 OCT structural imaging performance

The feasibility of using this probe for studying cerebral blood flow is determined by the system stability and the image quality of OCT structural imaging. To demonstrate the imaging performance of this probe, we secured the probe onto a holder and conducted a non-wearing OCT structural imaging experiment, in which we obtained high-resolution structural images of the mouse head with a 4 mm × 4 mm lateral FOV. Figure 2(a) shows a photograph of the mouse we used in this experiment, the fur on its head was removed by depilatory cream. We can see a blood plasma scab on the scalp, which wasn’t intentionally created; rather, it occurred as a result of the mouse scratching itself or engaging in fights with other mice. Figure 2(b) shows the OCT image of the mouse ear, from the image, we can observe the layered structures of the stratum corneum, the epidermal cell layers and the dermal layer.

 

Fig. 2. OCT structural imaging performance of the probe. (a) A photograph of a mouse, with a blood plasma scab on the scalp. (b) OCT image of the mouse ear, scanning position indicated by red dashed line in (a). (c) OCT image of the plasma scab on the scalp, scanning position indicated by red dashed line in (a). (d) En face OCT image of the plasma scab. (e) 3D OCT image of the plasma scab. Scale bars at bottom left of (b), (c), (d) and (e), 500 µm.

Download Full Size | PDF

To demonstrate the probe's capability in detecting abnormal structures, we imaged minor blood plasma scabs on the mouse scalp, clearly revealing the differences between the blood plasma scabs and normal skin, shown in Fig. 2(c). The en face OCT image shown in Fig. 2(d) was obtained by 3D scanning of the scabs and its surrounding area, and compressing the 3D data along the z-axis direction, the plasma scab can be clearly distinguished, as indicated by the red arrow. The volumetric data of the sample was displayed in 3D in Fig. 2(e), the raised contours of the plasma scab area can be clearly distinguished, indicated by the red arrow. Note that for this large FOV, due to hardware limitations of our existing system, the existing sampling strategy would result in undersampling, according to Nyquist sampling theorem, the image resolution for this FOV will decrease. However, undersampling is not present in the 2.3 mm FOV in the OCTA experiments below.

3.2 OCTA imaging performance

To verify the capability of our probe for conducting OCTA experiments, we pumped intralipid solution into a polytetrafluoroethylene (PTFE) tube with constant speed and then acquired 3D data sets. We took 800 B-scans in the slow scanner direction, at each step, we repeated the B-scan five times, each of these B-scans consisted of 1024 A-lines in the fast-scanning direction, the FOV is set to be 2.5 mm × 2.5 mm. The OCTA signals were obtained through the steps described above. Figure 3(a) and 3(d) are the cross-sectional images of the vascular model in OCT and OCTA mode, respectively. In Fig. 3(a) the signal from the tube is very prominent, whereas in Fig. 3(d), the signal from the tube is significantly attenuated, but the signal representing blood flow within the tube remains unaffected. Figure 3(b) and 3(e) are the comparison of en face images, the tube signal is significantly attenuated in the OCTA image, indicated by the red arrow. Figure 3(c) and 3(f) are the comparison of 3D images, showing the same results, note that the highlighted area at the top of the tube is due to the strong reflection caused by the smooth tube surface. This experiment confirms the feasibility of using this probe for OCTA experiments in vivo.

 

Fig. 3. OCTA imaging performance of the probe by using a vascular model. Blood vessel was simulated by a PTFE tube, blood was simulated by intralipid solution, and a syringe pump was utilized to drive the solution to simulate blood flow. (a) and (d) Cross-sectional of the vascular model, the tube is clearly visible in the OCT image (a), but it is significantly attenuated in the OCTA image (d), indicated by the red arrow. Meanwhile, the blood signal is well-preserved. (b) and (e) En face image of the vascular model, the tube signal is significantly attenuated in the OCTA image, indicated by the red arrow. (c) and (f) 3D image of the vascular model, the tube signal is significantly attenuated in the OCTA image, indicated by the red arrow. Scale bars at bottom left of (d), (e) and (f), 200 µm, same for (a), (b) and (c).

Download Full Size | PDF

We initially conducted a conventional non-wearing experiment by securing the probe onto a holder and imaging an anesthetized mouse placed beneath the probe. Figure 4(a) shows a photograph of the mouse brain, Fig. 4(b) and Fig. 4(c) are the OCTA images of the yellow and white square region in Fig. 4(a), respectively, representing 1.3 mm × 1.3 mm and 2.3 mm × 2.3 mm FOV. The cross-sectional OCTA images of the red dash lines in Fig. 4(b) and Fig. 4(c) are shown in Fig. 4(d) and Fig. 4(e), the depth of these images looks shallow, but actually it’s not, the reasons are as follows: First, these cross-sectional images are OCTA images, where most part of the static tissue is removed during the differential process, only the signals from blood vessels are kept. As a result, the signals in the deeper layers appear weaker. Second, we applied grayscale stretching to enhance the vascular signals on the original OCTA data, further attenuating residual structural signals and noise, which also makes the depth appear shallow (same reason for the cross-sectional images of Fig. 5). Note that we did not attach the probe to the mouse's head in this experiment. Wearable experiments results are shown in Figs. 5 and 6.

 

Fig. 4. OCTA imaging performance of the probe. (a) A photograph of the mouse brain, scale bar at bottom left, 1 mm. (b) OCTA image of the yellow square region in (a), scale bar at bottom left, 200 µm. (c) OCTA image of the white square region in (a), scale bar at bottom left, 400 µm. (d) and (e) Cross-sectional OCTA images of the red dash lines in (b) and (c), respectively, scale bar at bottom left, 200 µm.

Download Full Size | PDF

 

Fig. 5. Four-hour dynamic blood flow observation after PT stroke using our wearable probe, 2.3 mm × 2.3 mm FOV. (a)-(f) This is a process of acute ischemia, the ischemic area expanded significantly, with the vascular network disappearing remarkably. Scale bar at bottom left of (d), 400 µm. (g)-(l) Cross-sectional OCTA images of (a)-(f), respectively, the selected cross-section is indicated by red dash line in (b). Scale bar at bottom left of (j), 300 µm.

Download Full Size | PDF

 

Fig. 6. Freely moving imaging results of mouse brain, 2.3 mm × 2.3 mm FOV. (a) OCTA image of an anesthetized mouse brain. Scale bar at bottom left, 400 µm. (b) OCTA image, 5 minutes after the cessation of anesthesia. (c) OCTA image, 15 minutes after the cessation of anesthesia. (d) OCTA image, 25 minutes after the cessation of anesthesia. (e) OCTA image of fully awake and freely moving mouse, some stripes appeared due to the movement of the mouse. (f) The stripes in (e) were removed during post-processing. (g) and (h) are the binary images of the anesthetized state (a) and awake state (f), respectively. (i) Comparison of vessel diameters of anesthetized state and awake state of 6 selected vessels marked in (g) and (h).

Download Full Size | PDF

3.3 Four-hour stroke monitoring with wearable OCTA probe

Stroke is the second highest cause of death globally and a leading cause of disability. It has shown increasing incidence in developing countries, where 70% strokes are of ischemic origin [24]. For ischemic stroke, the rapid and efficient restoration of cerebral blood flow is crucial in halting pathological processes and minimizing functional impairment [25]. Consequently, thrombolysis plays a vital role in the early treatment of ischemic stroke [26], with an ideal treatment window of 4.5 hours following the onset of ischemia [21]. However, a significant number of patients fail to meet this critical time frame. Hence, a rapid and high-resolution in vivo imaging technique is of great importance for visually assessing blood perfusion and brain tissue damage, as well as for longitudinally monitoring the ischemic region to study the progress and mechanisms of post-stroke reperfusion.

Wearing the wearable probe on the mouse's head allows for continuous and long-term dynamic monitoring of cerebral blood flow, which is beneficial for studying dynamic physiological and pathological processes. Moreover, it eliminates the cumbersome step of position registration required for multiple imaging of the same location at different times. In this study, we used our probe to monitor the mouse brain for up to four hours after PT stroke, the stroke was induced as described in 2.3. After laser illumination, we placed the probe on the mouse's head. Figure 5(a)–5(f) are the en face OCTA images, Fig. 5(g)–5 l are the corresponding cross-sectional OCTA images. Figure 5(a) shows the cerebral vascular distribution 20 minutes after MCA occlusion. The blue arrow points to the edge of the MCA that hasn't been completely blocked yet. This region becomes fully occluded after 120 minutes, as shown in Fig. 5(d). Blood vessels marked by green and yellow arrows in Fig. 5(a) are fully occluded after 60 and 240 minutes, respectively, as shown in Fig. 5(c) and 5(f). The vascular network within the red box in Fig. 5(a) gradually diminishes, with only a few remaining unblocked after 240 minutes, as depicted in Fig. 5(f). From the cross-sectional images, it is also noticeable that some blood vessels gradually disappear during this process, as indicated by the red arrows in Fig. 5 g. Note that the cross-sectional OCTA images of Fig. 4 and Fig. 5 are different, this is because the major blood vessels in the central region have been occluded in Fig. 5, leaving only some small peripheral vessels. And there are bright non-vascular areas in Fig. 5(g)–5 l, this is because the core region is in the acute ischemic phase, and non-vascular areas are also in a dynamic process, which results in differential signals.

This process is not only a part of the acute ischemic phase but also represents the optimal time window for thrombolysis. The introduction of our wearable probe helps to monitor the dynamic evolution of blood perfusion during this process and provides valuable guidance for thrombolytic therapy.

3.4 Freely moving OCTA imaging

The data obtained in the freely moving state mice is closer to the natural physiological conditions, avoiding the stress of fixation or anesthesia required in traditional experiments. This approach assists in more accurately observing the cerebral hemodynamics of mice. Figure 6 illustrates the morphological changes in blood vessels obtained using the wearable probe, from the anesthetized state to the fully awake state. Figure 6(a), 6(b), 6(c), 6(d), and 6e correspond to the anesthesia state, 5 minutes after anesthesia cessation, 15 minutes after anesthesia cessation, 25 minutes after anesthesia cessation, and the fully awake state with free movement, respectively. We can observe that as the mouse gradually awakened and its physical activities increased, there is a progressively increasing presence of stripes in the OCTA images. These stripes result from the movements of the mouse's body between two successive B-scans at the same location, which disrupts the results of the differential operation. This study employed a simple procedure to remove these stripes: subtracting the mean value of each row from Fig. 6(e), yielding the result shown in image Fig. 6(f), effectively suppressed the horizontal stripes. Visualization 3 in Supplementary Materials shows the data acquisition process of the wearable probe on a fully awake mouse, the jitter seen in the video is a result of the mouse's movement.

Additionally, we performed quantitative analysis of blood vessel diameters in the anesthetized and awake states. Figure 6 g represents the binarized OCTA image of the anesthetized state, while Fig. 6 h shows the binarized image of the awake state. In these binarized images, we selected six blood vessels in the same positions and compared their diameters between the anesthetized and awake states. The results are shown in Fig. 6(i), note that the diameters of the blood vessels in the awake state are set to be 1, while those in the anesthetized state have expanded to 1-2 times larger than their normal size, which is in agreement with previous works [27,28].

4. Discussion and conclusion

In this study, we have developed a novel wearable OCTA probe that achieves imaging capabilities with 12µm axial resolution, 5.5µm lateral resolution and 4 mm × 4 mm FOV. This wearable probe allows mice to move freely, enabling data collection in their natural state. This eliminates the stress induced by anesthesia or fixation, which is beneficial for studying the more authentic cerebral hemodynamics and neuronal activity in mice. We conducted a four-hour monitoring of cerebral blood flow with this probe during the acute ischemic phase after PT stroke, which deepens our understanding of the dynamic evolution of blood perfusion during this process and offers valuable guidance for thrombolytic therapy.

Furthermore, we also quantitatively analyzed the differences in mouse brain vasculature between awake and anesthetized states, promoting the transition of neuroscience experiments to natural, non-anesthetized state.

Wearable devices hold the potential for future integration with various technologies to expand their applications. For example, wearable devices can be combined with Doppler technology to measure blood vessel morphology and simultaneously obtain flow velocity information [29]. It's also possible to introduce measurements of hemoglobin absorption and scattering coefficients [30], allowing the calculation of blood oxygen saturation in a non-anesthetized, natural state. Further improvements in probe structure can incorporate photoacoustic imaging technology [31], enabling wearable multimodal imaging for more accurate disease diagnosis and monitoring. Additionally, when imaging awake mice, motion-induced stripes, as shown in Fig. 6(e), can be addressed by implementing motion compensation for the affected frames [32] or by utilizing deep learning techniques to remove stripes and enhance image quality [33].

It's worth mentioning that the probe developed in this work is a first-generation design, its weight and size can be further optimized. There are reports of photoacoustic wearable probes weighing as little as 1.8 g, and we can explore similar improvements, for example, we can build our own collimators instead of using commercially available ones with metal casings, use lighter materials to fabricate the overall frame, and consider changing the probe's solid structure to a hollow one.

We believe that this work holds promising prospects for future applications and will promote our understanding of brain health and neurological disorders.