2.1.

Two-Photon Fluorescence Microscope with a Bessel/Gaussian Module

A schematic diagram of the optical setup is provided in Fig. 1(a). A femtosecond laser (Axon920TPC, Coherent) delivered pulses of up to 1 W and 150-fs pulses at 920 nm with a repetition rate of 80 MHz. After expansion and collimation by an optical 4f system, the laser pulse was directed to a beam-switching device that alternated between the Gaussian and Bessel modules using flip mirrors (FM1 and FM2). In the Bessel module, various orders and DOFs of Bessel beams were generated by a phase-only spatial light modulator (SLM, Pluto NIR-015, Holoeye) and then transformed into optical rings of diverse radii by L3. A custom annular mask²⁵ was placed at the back focal plane of L3 to filter out stray light outside the Bessel beam. The optical rings were relayed to a pair of galvanometer scanning mirrors (GVS011/M, Thorlabs) through optical 4f systems (L4 to L7), with the scanning mirrors conjugated with the rear pupil of the microscope objective using a scan lens and a tube lens. Objective1 (Plan Apo Lambda 20×, Nikon) was used to measure fluorescent beads, and objective2 (XLPLN25XWMP2, 25×, NA 1.05, Olympus) was used for in vivo experiments. The fluorescence emission from the sample was collected by the same objective and passed through a dichroic mirror (ZT670rdc-xxrxt, Chroma), a short-pass filter (ET680sp-2p8, Chroma), and a collection lens to finally reach the photomultiplier tube (PMT2001/M, Thorlabs).

2.2.

Sample Preparation

All animal operations complied with the protocols approved by the Institutional Ethical Committee of Animal Experimentation of Shenzhen University. C57 mice were purchased from Guangdong Medical Laboratory Animal Center. Before 2-photon imaging, craniotomies were performed on C57 mice (female, 8 to 10 weeks old, weighing 20 to 21 g) centered at 2 mm lateral and posterior to the bregma point. A cover glass was used to seal the cranial window. To fix the mice, a home-made alloy ring was stuck onto the skull by dental cement. During surgery and imaging, a gaseous anesthesia system (Matrix VIP 3000, Midmark) and a heating blanket were used to anesthetize mice and maintain their body temperature, respectively. A prepared aqueous solution ($200\mu \mathrm{L}$) of fluorescent probes TPETPABT²⁶ was injected through the orbit to label the brain vasculature.

Zebrafish embryos of the Tg(kdrl:EGFP) strain were obtained from the China Zebrafish Resource Center and raised in E3 solution containing 0.003% $N$-phenylthiourea (Sigma) to inhibit pigmentation after 20 h postfertilization. Vascular epithelial cells were labeled with EGFP. Prior to live imaging, zebrafish were anesthetized with $600\mu \mathrm{M}$ Tricaine (Sigma, Cat. # E10521) and mounted in 1% low-melting-point agarose (NuSieve GTG, Cambrex BioScience, Cat. # 50080) for imaging to minimize movement artifacts during imaging.

2.3.

SLM-Based Bessel Module

Herein, we used a phase-only SLM²⁷ instead of an axicon to generate Bessel beams. The SLM can produce tunable Bessel beams, allowing for modulation of the order, lateral resolution, and depth of focus to adapt to different samples and applications. To generate Bessel beams using an SLM, a grayscale image must be loaded onto the device. This image can be a hologram encoded with either an axicon or an annular aperture, with the former having a higher beam conversion efficiency:²⁸

The transmission function of the axicon²⁹ is given as

In Fig. 1, the Fourier spectral plane of the Bessel beam (i.e., the mask) is conjugated to the pupil plane of the objective. The effective numerical aperture ${\mathrm{NA}}_{\mathrm{eff}}$ of the Bessel focus scanned on the sample depends on the annular radius $R_{\text{pupil}}$ on the pupil plane of the objective:

Here $f_{\mathrm{obj}}$ is the effective focal length of the objective, $m$ is the magnification from the mask to the pupil of the objective, $f_3$ is the focal length of the lens after the SLM, and $\beta ={\sin }^{-1}(n\sin \alpha )-\alpha$.

The lateral resolution of the generated Bessel focus is given as

Both lateral resolution and the DOF are inversely proportional to the ${\mathrm{NA}}_{\mathrm{eff}}$, i.e., they counterbalance each other. The DOF of the Bessel focus can be modulated by the hologram loaded on SLM, the lens positioned after the SLM, the magnification from the mask to the objective’s pupil, and the objective itself. Although the enhancement can exceed 30-fold, it would necessitate a substantial increase in power usage. Moreover, an overly aggressive enhancement could potentially lead to signal overlap. Thus we produced a Bessel focus with a lateral resolution slightly lower than that of the Gaussian focus but with a DOF approximately seven times greater [Fig. 1(b)], wherein $\alpha =1\mathrm{deg}$, $f_3=75\mathrm{mm}$, $m=6$, and objective 1 were used.

3.1.

Rapid Volumetric Imaging with the Bessel-TPFM System

To demonstrate the capabilities of fast volumetric imaging, we used our home-built Bessel-TPFM system for performing volumetric imaging of $200\mu \mathrm{m}\times 200\mu \mathrm{m}$ areas of vasculature in the mouse brain and transgenic zebrafish using both Gaussian and Bessel foci (Fig. 2). The 3D projection images were obtained by axial scanning with a step size of $1\mu \mathrm{m}$ using a Gaussian focus, with depth information color-coded to enable post hoc assignment of depth information to data from the Bessel module. Obviously, a 2D image obtained by scanning with a Gaussian focus can only capture the transverse section of a blood vessel, so axial scanning is required to capture the complete blood vessel signal. For in vivo volumetric imaging, we scanned a volume of $200\mu \mathrm{m}\times 200\mu \mathrm{m}\times 30\mu \mathrm{m}$ using a Gaussian focus and acquired a stack of 30 images, with each $512\times 512$ frame taking 5 s, which is not sufficient for real-time monitoring in in vivo imaging. By contrast, a single-frame obtained by 2D scanning with a Bessel focus can capture all signals within the volume, increasing the imaging throughput by 30 times. The speed enhancement can be quantified as $N$, which is the ratio of the DOF of the Bessel focus by the axial step of the Gaussian focus. In this study, the DOF of the Bessel focus is $37\mu \mathrm{m}$ and the axial step of the Gaussian focus is $1\mu \mathrm{m}$. Considering the problem of weak signal loss, the speed of the Bessel-TPFM system can be augmented by more than 30 times. It should be mentioned that the imaging speed here is limited by the hardware conditions of the galvanometer, whereas the fast volumetric imaging capability of the Bessel-TPFM system is compatible with fast scanning devices, such as resonant scanners and acousto-optic deflectors. In addition, during tomographic scanning, the step size causes some signals to not be detected by the Gaussian projection, such as in the ROI 1 of Figs. 2(a)–2(d), resulting in the loss of some detailed biological information. The postobjective power utilized in Figs. 2(a)–2(f) was 15, 45, 40, 80, 10, and 52 mW, respectively. Bessel foci require significantly more power to generate the same intensity as they distribute the majority of their energy into their sidelobes and have extended DOFs. However, due to the large fraction of power in the sidelobes of the Bessel beam, the signal-to-background ratio (SBR) of the Bessel frame in the same areas is lower than the Gaussian projection that averages dozens of frames, such as ROI 2 of Figs. 2(a)–2(d). From the imaging results of zebrafish [ROIs of Figs. 2(e)–2(f)], the image quality of the Bessel frame is better than that of Gaussian projection. This may be due to the fact that zebrafish embryos, as transparent samples, have less scattering and sufficient power (5 times that of Gaussian projection) to excite a strong enough signal, making the background produced by sidelobes less noticeable. However, the vast majority of biological samples are strongly scattering and opaque substances, and high power is limited in biological microscopy. Therefore, it is necessary to further suppress side lobes that introduce background noise.

Fig. 2

In vivo volumetric structural imaging of vasculature using the Bessel-TPFM system. Gaussian-TPFM image projection of vasculature at a depth of (a) 300 to $330\mu \mathrm{m}$ and (c) 400 to $430\mu \mathrm{m}$ below the top surface of the dura mater in the mouse cortex in vivo, color-coded by depth. (e) Gaussian TPFM image projection of vasculature in the Tg (kdr: EGPF) s843 transgenic zebrafish, expressing GFP in vascular endothelial cells, color-coded by depth. (b), (d), and (f) Scanning the Bessel focus in 2D captured all vasculature in the volumes shown in (a), (c), and (e), respectively. The regions of interest (ROIs) were enlarged from the areas within the dashed boxes in (a)–(f), and the intensity profiles were compared across the same blood vessels. Scale bars: $20\mu \mathrm{m}$. Postobjective power: (a) 15 mW, (b) 45 mW, (c) 40 mW, (d) 80 mW, (e) 10 mW, and (f) 52 mW.

3.2.

High-Contrast Imaging with the Bessel-TPFM System

We imaged fluorescent beads using the constructed Bessel-TPFM system and acquired the sidelobe proportion of Bessel beams with different effective numerical apertures ${\mathrm{NA}}_{\mathrm{eff}}$ [Fig. 3(a)]. The sidelobe proportion is represented by the energy ratio of the first-order sidelobe to the central main lobe. Although two-photon excitation can suppress the energy of sidelobes to a certain extent compared with single-photon excitation, it can be seen that, even in the case of low NA, the sidelobe proportion is still close to 20% [Fig. 3(b) shows the imaging result when ${\mathrm{NA}}_{\mathrm{eff}}=0.38$] and it rises with the increase of ${\mathrm{NA}}_{\mathrm{eff}}$. For high-resolution imaging, the situation of high ${\mathrm{NA}}_{\mathrm{eff}}$ must be considered.

Fig. 3

Proportion of energy in the side lobes and the image-difference method. (a) The proportion of energy in the first side lobe relative to the central main lobe under different ${\mathrm{NA}}_{\mathrm{eff}}$. Error dotted lines represent standard deviations. Experimental data were linearly fitted. (b) The proportion of energy in the $x-y$ cross section of the Bessel beam when ${\mathrm{NA}}_{\mathrm{eff}}=0.38$. (c) The proportion of energy in the side lobes after image difference processing with different difference coefficients. (d) The simulated results of the image-difference method. (e) The lateral and axial intensity distributions of the (e1) zeroth order, (e2) third-order Bessel beam, and (e3) the difference result. $X$, $Y$, and $Z$ scale bars: $1\mu \mathrm{m}$. Results from one $1\text{-}\mu \mathrm{m}$-diameter bead.

Because the intensity distribution of the third-order Bessel beam is roughly the same as the sidelobe distribution of the zeroth-order Bessel beam, by selecting an appropriate coefficient to perform a difference between the two, the sidelobe energy can be greatly reduced while maintaining the central spot. The difference outcome is derived from the subtraction of the third-order Bessel beam image from the zeroth-order Bessel beam image. Given the disparity in their amplitudes, it is imperative to incorporate a differential coefficient to bridge this gap. We identified the optimal difference coefficient that results in the minimization of the energy proportion of the first sidelobe. As shown in Fig. 3(c), we compared the suppression effect of sidelobes under different difference coefficients through numerical simulation. The sidelobe proportion is the energy ratio of the first-order sidelobe to the central spot after taking the absolute value of the difference results. When the difference coefficient is 0.77, the sidelobe proportion decreases from 16.8% to 6.77% [Fig. 3(d)].

We generated zeroth-order and third-order Bessel foci of ${\mathrm{NA}}_{\mathrm{eff}}=0.38$ to scan the fluorescent beads and calculated their difference result with the optimal difference coefficient [Fig. 3(e)]. It is obvious that the sidelobe distribution of the zeroth-order Bessel beam matches well with the transverse intensity distribution of the third-order Bessel beam and the axial length is also basically consistent. Therefore, the signals from the central spot are maintained whereas those from the sidelobes are effectively suppressed after image difference processing.

Although previous research²¹ has introduced the image-difference method based on Bessel beams with successful applications in mouse brain sections, tissue slices lack rapidly changing physiological information, eliminating the need for sacrificing depth information for rapid volume imaging using the Bessel focus. Bessel-TPFM represents a technique that enhances imaging throughput at the expense of depth information and is aimed at studying rapidly changing physiological activities in vivo. To validate the effectiveness of this method in more varied and complex physiological environments, we extended its application to the in vivo imaging of mouse brains and zebrafish.

To suppress the background in the Bessel frames [Figs. 4(a), 4(d), and 4(g)], we used the image-difference method with a difference coefficient of 0.5, which is distinct from the optimal theoretical value. However, the optimal theoretical value is computed with the objective of minimizing the first sidelobes, while not taking into account the potential degradation of peak signal and the occurrence of image artifacts. The difference results exhibit clearer and more distinct vessel contours without significant signal loss [Figs. 4(b), 4(e), and 4(h)]. To further elucidate the changes in image contrast via the image-difference method, we selected ROIs at different vessels and measured the intensity profiles by drawing lines. In the difference results [ROIs of Figs. 4(b), 4(e), and 4(h)], the signal peaks of the vessels become significantly narrower, whereas the background signal is efficiently suppressed, allowing for improved separation between signal peaks. The SBR is markedly enhanced, demonstrating that the image-difference method can effectively suppress background noise generated by the side lobes of Bessel beams, thereby improving imaging contrast. Notably, we selected the same regions in both Figs. 2 and 4, enabling a comparison of the results from the Gaussian and Bessel modules, as well as the image-difference method.

Fig. 4

Image-difference method in the Bessel TPFM system. (a), (d), and (g) Image of the vasculature in mouse brain (red) and zebrafish (green) scanned by zeroth-order Bessel beam. (b), (e), and (h) The difference results. (c), (f), and (i) The intensity profiles of ROIs in (a), (b), (d), (e), and (g), (h). Scale bars: $20\mu \mathrm{m}$.

3.3.

Consistently Steady Imaging With the Bessel-TPFM System

Axial movement caused by heartbeats and breathing presents a significant challenge in in vivo measurements. The Bessel-TPFM system exhibits robustness against axial motion artifacts due to its extended DOF,³⁰ enabling it to steadily monitor changes in fluorescence signals during real-time in vivo measurements. We initially observed Bessel beams’ stability in living samples, which is a promising approach for biomedical research.

In Fig. 5, we used the Gaussian module and Bessel module to acquire image time series of the same region of zebrafish blood vessels [Fig. 5(a) shows a representative frame from Video 1, and Fig. 5(e) displays a representative frame from Video 2]. For the Gaussian module, the vertical displacement of blood vessels caused signal drift and loss of detail [Figs. 5(b) and 5(c)], resulting in measurement bias. In Fig. 5(d), we calculated the displacement and correlation coefficient between each pair of adjacent images in the image time series, demonstrating that vibration of blood vessels can cause significant measurement bias. By contrast, the Bessel module exhibited a degree of resistance to axial displacements, allowing for better preservation of signal details [Fig. 5(f)] and consistently steady monitoring of fluorescence signals [Figs. 5(g) and 5(h)].

Fig. 5

Image time series of the vasculature in zebrafish scanned by (a) Gaussian foci and (e) Bessel foci. (b) The ROIs extracted from (a). (c) The intensity profiles along the white line in (b). (d) The displacement and correlation of the ROIs in the image time series. (f) The ROIs extracted from (e). (g) The intensity profiles along the white line in (f). (h) The displacement and correlation of the ROIs in the image time series. Scale bars: $10\mu \mathrm{m}$ (Video 1, MP4, 38.8 kB [URL: https://doi.org/10.1117/1.JBO.29.1.016501.s1]; Video 2, MP4, 43.4 kB [URL: https://doi.org/10.1117/1.JBO.29.1.016501.s2]).