Axial imaging technologies in multiphoton microscopy

MPM provides intrinsic 3D spatial selectivity through nonlinear excitation, offering unique advantages for imaging deep structures and dynamic processes in scattering tissues. In neuroscience, neurons and their activities are distributed across three-dimensional volumes ranging from tens to hundreds of micrometers or more, making a single 2D plane insufficient for fully characterizing complex biological processes. Conventional methods acquire images layer by layer by moving the objective, sample, or focal plane along the Z-axis. Although they provide reliable optical sectioning, sequential acquisition limits volumetric imaging speed. With advances in rapid optical modulation, light-field control, and computational imaging, various approaches have emerged for axial information acquisition. Here, we classify them into three categories: single-point axial scanning, multifocal axial imaging, and continuous-beam axial imaging (Fig. 1). These approaches correspond to sequential layer-by-layer scanning, discrete parallel acquisition, and continuous axial coverage, respectively, providing a unified framework for 3D MPM.

In single-point axial scanning, a diffraction-limited focus is sequentially moved to different depths, representing the most direct and mature approach to 3D imaging (Fig. 2). Early methods relied mainly on mechanically moving the objective or sample for Z-scanning, offering simple structures, high image quality, and accurate axial positioning, but with speeds limited by mechanical inertia, vibration, and settling time. To overcome these limitations, inertia-free optical focusing methods based on electrically tunable lenses (ETLs), spatial light modulators (SLMs), deformable mirrors (DMs), and remote focusing have been developed to rapidly shift the focal plane by dynamically controlling wavefront curvature. Axial scanning has further evolved from frame-by-frame switching to rapid modulation within a pixel dwell time. TAG lenses, beam splitting with delayed recombination, and reverberation loops (Light Beads) introduce different temporal delays and wavefront curvatures, allowing excitation pulses to focus at different axial positions during a single lateral scan. Thus, single-point axial scanning has progressed from mechanical layer-by-layer imaging to high-speed temporal multiplexing while maintaining optical sectioning and signal fidelity.

Compared with sequential acquisition of different depths using a single focus, multifocal axial imaging generates multiple foci at different axial positions simultaneously, enabling parallel acquisition of depth information (Fig. 3). Based on whether signals from different depths can be distinguished, these methods can be divided into 2D projection and depth-encoded imaging. In the projection mode, multiple foci simultaneously excite different depths, and fluorescence signals from different layers are integrated by a single detection channel to form a 2D projection image, enabling efficient volumetric information acquisition. For depth-resolved imaging, distinguishable optical encoding can be introduced at different axial positions. For example, TEMP generates Bessel droplets through interference between two Bessel beams with different axial wavevectors, encoding depth information into axial spatial frequencies. AIMED combines programmable binary axial illumination with compressed sensing to reconstruct multiple depth planes from fewer encoded measurements. Experiments demonstrated compressed imaging of 5 and 8 brain-tissue layers with a compression ratio of up to 60%, while simulations extended the imaging scale to 47 axial planes with an approximately eightfold increase in imaging speed. These developments establish two representative routes for multifocal axial imaging, from parallel projection acquisition to depth-encoded reconstruction.

In addition to discrete multifocal imaging, continuous-beam axial imaging provides another route to high-speed 3D imaging. This approach uses nondiffracting or quasi-nondiffracting fields such as Bessel and Airy beams to extend a localized focus along the axial direction, enabling continuous coverage over a depth range (Fig. 4). Extended-depth-of-field (EDOF) imaging can acquire volumetric projection information across the entire depth range with a single 2D scan. Continuous optical fields can also encode depth through propagation-induced lateral displacement, beam trajectories, viewing-angle differences, or intensity variations, enabling depth-resolved imaging. For example, mirrored Airy beams recover axial position from lateral displacement, dual-tilted beams establish depth mapping through geometric separation, and multi-view Airy beams combine multiple views with tomographic reconstruction. Gradient-excitation two-photon microscopy further uses two Bessel beams with complementary axial intensity profiles to recover depth information. These approaches provide multiple routes from EDOF projection to depth-encoded reconstruction.

Beyond the three approaches for axial information acquisition, the review also summarizes recent progress in improving axial imaging through excitation light-field control (Fig. 5). These methods design and control the spatial structure of the excitation field to improve lateral resolution, suppress sidelobe background, and maintain more consistent focal characteristics across different depths. For Bessel-type beams, differential imaging between different orders or modes can further compress the effective point-spread function and improve lateral resolution, while higher-order Bessel beams can suppress sidelobes and enhance image contrast. Truncated nondiffracting beams further reshape the axial field distribution, maintaining relatively consistent lateral and axial focal dimensions over a certain depth range and enabling nearly depth-independent, inertia-free 3D resolving capability. Overall, these studies demonstrate that excitation light-field control can improve key imaging metrics while providing additional control over focal evolution in three dimensions.

This review summarizes three major approaches to axial information acquisition in MPM: single-point axial scanning, multifocal axial imaging, and continuous-beam axial imaging, together with recent advances in excitation light-field control. Single-point scanning enables rapid depth acquisition through mechanical scanning, optical focusing, and temporal multiplexing. Multifocal imaging combines parallel excitation with axial encoding and computational reconstruction, while continuous-beam imaging provides continuous axial coverage and depth-resolved imaging through spatial and intensity-based encoding. Meanwhile, Bessel, Airy, and truncated nondiffracting beams offer new approaches for controlling lateral resolution, sidelobe suppression, and focal consistency across different depths.

Looking ahead, 3D MPM will continue to seek an improved balance among volumetric imaging speed, axial resolution, imaging depth, light dose, and system complexity. With advances in spatial light modulation, fast optical devices, structured light fields, and computational imaging, different axial imaging approaches may be further combined with adaptive optics and intelligent reconstruction to dynamically control 3D excitation and information acquisition. These developments may drive MPM toward high-speed 3D imaging based on the integration of light-field control, information encoding, and computational reconstruction, providing new tools for imaging deep tissues, neural activity, and complex biological dynamics.