Full light three-dimensional scanning confocal fluorescence microscopic imaging device and implementation method thereof

Abstract

This invention discloses an all-optical three-dimensional scanning confocal fluorescence microscopy imaging device and its implementation method. The invention employs a deep learning-driven adaptive control method, using CNN to process spatially distributed data, dynamically controlling the phase hologram, laser intensity, and phase compensation to ensure aberration-free axial displacement of the excitation light focus. It also optimizes the bidirectional parallel optical scanning trajectory of the two-dimensional scanning system to achieve all-optical three-dimensional scanning. Combining LSTM and TCN, it obtains time-dependent relationships and integrates control of the laser source, adjustable aperture, electrically adjustable focusing lens, and photodetector. This enables efficient synchronous and automated operation and high-speed axial focusing adjustment, eliminating errors and time-series mismatches, automatically adapting to different sample optical characteristics and environmental interference, maintaining high-quality imaging, and improving system robustness and applicability. This invention is used for real-time tracking of embryos in model organisms, dynamic visualization of intracellular signaling molecules, observation of cell membrane protein migration and aggregation, and tracking of intracellular organelle interactions.

Description

Technical Field

[0001] This invention relates to the field of microscopic imaging, specifically to an all-optical three-dimensional scanning confocal fluorescence microscopic imaging device and its implementation method. Background Technology

[0002] Optical fluorescence microscopy, as an important branch of modern optics, is constantly innovating with breakthroughs in optical engineering, demonstrating irreplaceable application value in biomedical research. Optical fluorescence microscopy operates based on the principles of fluorescence labeling and optical detection, using excitation light of a specific wavelength to induce fluorescent molecules in the sample to emit fluorescence signals. This technology has become an indispensable tool in modern biomedical research, providing key technical support for revealing the fundamental laws of life activities. Contemporary biological imaging places three core requirements on optical microscopy systems: wide field-of-view coverage, high spatiotemporal resolution, and three-dimensional volumetric imaging capabilities. These technical requirements essentially stem from the fundamental needs of in vivo biological research—only with high-fidelity optical imaging methods can precise observation and recording of subcellular structures and their dynamic biological processes be achieved. Especially when studying rapidly changing life phenomena, these imaging parameters often constrain each other, constituting a technical bottleneck that urgently needs to be overcome in the current field of biological optical imaging. In the existing microscopy technology system, wide-field fluorescence microscopy, with its single-shot two-dimensional imaging characteristics, can acquire a series of two-dimensional images when combined with three-dimensional scanning of the sample. However, this technology has inherent physical limitations. The imaging process involves primary volumetric excitation, which cannot effectively distinguish fluorescence signals between the focal and defocus planes, resulting in images containing significant background noise. This technical deficiency directly prevents it from achieving three-dimensional volumetric imaging and thus cannot be applied to research requiring three-dimensional structural analysis. These limitations make traditional wide-field fluorescence microscopy unable to meet the stringent imaging quality requirements of modern biological research.

[0003] Confocal microscopy, a significant breakthrough in modern fluorescence microscopy, achieves several technological advantages through precise optical path design. Firstly, it employs a point-scan imaging method. The excitation source, focused by the objective lens, forms a diffraction-limited spot that scans the sample point by point. The fluorescence signal excited at each point is collected by the same objective lens, and the coaxial structure ensures the stability of in-situ excitation and imaging. The tightly focused source forms a highly confined excitation focal point within the sample. The focal point of the laser focused into the sample and the focal point of the fluorescence signal incident on the detector are conjugate, with the pinhole closely attached to the photodetector surface. The collected fluorescence is separated by a dichroic mirror and filtered by the conjugate pinhole to remove defocus signals, leaving only information from the focal point, thus significantly improving the system's signal-to-noise ratio and image contrast. Secondly, the pinhole filtering method of confocal microscopy effectively filters out defocus signals, giving it excellent axial resolution and enabling sub-micron level optical sectioning. This characteristic makes it particularly suitable for biological research scenarios requiring high-precision two-dimensional imaging. Furthermore, the confocal system also possesses excellent optical tomography capabilities, clearly resolving structural information at different depths within the sample and achieving three-dimensional imaging of thick samples. Therefore, confocal systems are excellent tools for 3D imaging of mesoscopic structures such as cells. However, the 3D imaging speed of confocal microscopes is also limited by their point-scanning working principle. Only a single point can be focused at a time. To achieve 3D reconstruction, the system must acquire the 3D information of the sample through point-by-point scanning, a process that significantly increases data acquisition time. Currently, the mainstream confocal 3D imaging schemes mainly adopt two technical paths: The first is a Z-axis scanning method based on mechanical displacement, which uses a precision stepper motor to drive the sample stage or objective lens to switch focal planes. While this method can ensure the quality of focal plane imaging, it significantly reduces the imaging frame rate and may introduce motion artifacts in fast dynamic process observations. More importantly, it restricts sample preparation to a fixed displacement stage, completely excluding the possibility of in vivo imaging. The second method uses the change of lens focal length to control the convergence or divergence of light entering the objective lens to achieve rapid axial scanning. Although this scheme improves imaging speed, it introduces spherical aberration that is difficult to correct and is limited by the lens focusing range, failing to meet the needs of large-scale volume imaging. In addition, both approaches face challenges of photobleaching and phototoxicity when achieving rapid 3D imaging, which to some extent limits their application in long-term live observation.The core issue with these problems lies in the fact that existing solutions generally lack the ability to adaptively perceive dynamic changes in the imaging scene and the intelligent control mechanism for multi-component collaboration: mechanical displacement solutions cannot adaptively adjust the scanning path and focal plane switching rhythm according to the real-time motion state of the sample and the characteristics of the imaging area, and the hardware is also not adapted to in vivo imaging; focal length control solutions are difficult to intelligently compensate for aberrations for optical characteristics at different axial depths, and fixed focusing logic cannot adapt to dynamic aberration changes in large-scale imaging; more importantly, neither type of solution has established an intelligent correlation between imaging accuracy requirements, light dose distribution and scanning parameter matching, and can only output light dose and scanning parameters in a fixed mode, making it difficult to balance imaging efficiency and the protection of biological sample activity, and failing to meet the needs of high-precision application scenarios such as long-term dynamic observation of living organisms and imaging of large-scale complex samples. Summary of the Invention

[0004] To address the core bottlenecks in high-speed confocal microscopy imaging, including imaging speed, aberration control, scene adaptability, and protection of biological sample activity, this invention proposes an all-optical three-dimensional scanning confocal fluorescence microscopy imaging device and its implementation method. Utilizing a depth-adaptive control device, the device dynamically optimizes light field modulation parameters, precisely compensates for aberrations, and achieves aberration-free axial large-range scanning under high numerical aperture. Relying on intelligent collaborative control of high-speed resonant mirror bidirectional parallel scanning imaging technology, the device shortens sampling time, improves temporal resolution, and precisely controls light dose output to reduce phototoxicity, thus realizing all-optical three-dimensional high-speed aberration-free confocal microscopy imaging.

[0005] One object of the present invention is to provide an all-optical three-dimensional scanning confocal fluorescence microscopy imaging device.

[0006] The all-optical three-dimensional scanning confocal fluorescence microscopy imaging device of the present invention includes: an excitation source, a depth adaptive control device, a dichroic mirror, an adjustable aperture, a two-dimensional scanning system, a telecentric scanning correction system, an objective lens, a motorized adjustable focusing lens, a pinhole camera, a photodetector, and an intelligent control system; wherein, the two-dimensional scanning system is connected to a data acquisition card, and the photodetector is connected to an image acquisition card; the excitation source, the depth adaptive control device, the adjustable aperture, the data acquisition card, the motorized adjustable focusing lens, and the image acquisition card are respectively connected to the intelligent control system;

[0007] The excitation source emits a polarized Gaussian beam as excitation light, which enters the depth adaptive control device. The spatial light modulator of the depth adaptive control device controls the phase of the excitation light by loading a phase hologram calculated by a feedback aberration-free axial scanning algorithm and adjusted in real time by an adaptive control method driven by deep learning, thereby achieving aberration-free axial focus movement behind the objective lens. After passing through a dichroic mirror, the first-order diffracted light is filtered by an adjustable aperture to the two-dimensional scanning system for two-dimensional planar optical scanning, which is then focused by the objective lens onto the surface or internal observation area of ​​the sample.

[0008] Fluorescent signals generated on or inside the sample are collected by the same objective lens and transmitted back to the two-dimensional scanning system in the opposite direction of the excitation light. The two-dimensional scanning system performs descanning and splits the excitation light beam with a dichroic mirror. The focal length is adjusted by an electrically adjustable focusing lens according to the change of the axial position of the excitation light, so that the fluorescence signals at different axial positions in the sample can be focused on the detection surface of the photodetector. After passing through a pinhole located in front of the photodetector to filter out the defocused signal, the signal is received by the photodetector.

[0009] Through an intelligent control system, an adaptive control method driven by deep learning is adopted to obtain an accurate relationship between the phase hologram and the axial position. The laser intensity is optimized based on the axial position of the focal point, and the inherent aberrations of the optical system are eliminated. The phase hologram loaded by the depth adaptive control device, the laser intensity, and the phase compensation of the optical system are dynamically controlled to achieve aberration-free axial displacement of the excitation light at the focal point behind the high numerical aperture objective lens. Furthermore, the bidirectional parallel optical scanning trajectory of the two-dimensional scanning system is optimized in real time to achieve all-optical three-dimensional scanning.

[0010] The sample is placed on the sample holder.

[0011] It also includes a beam expander and collimator, placed after the excitation source, to expand and collimate the excitation light; the beam expander and collimator uses a double lens for beam expansion and collimation; it is precisely configured by a pair of lenses with different focal lengths; a short focal length lens with a focal length of f1 serves as the input lens, and a long focal length lens with a focal length of f2 serves as the output lens, with a beam expansion ratio M = f2 / f1, and the diameter of the expanded beam should match the liquid crystal panel of the spatial light modulator.

[0012] The depth-adaptive control device comprises a half-wave plate, a prism, and a kilohertz-level pure phase-type silicon-based liquid crystal spatial light modulator, which together achieve precise control of the light field. The half-wave plate modulates the polarization state of the incident excitation light. By precisely rotating the angle of the half-wave plate, the linearly polarized light is adjusted to a polarization state matching the optimal working polarization direction of the spatial light modulator. The prism works with the spatial light modulator to control the light field of the Gaussian light after beam expansion, collimation, and polarization modulation. The Gaussian light is reflected from one side of the prism and incident on the liquid crystal screen of the spatial light modulator. The spatial light modulator, by loading a phase hologram, changes the axial position of the excitation light focused behind the objective lens, thereby achieving light field control and completing high-speed axial optical scanning.

[0013] It also includes a 4f relay system, positioned between the dichroic mirror and the two-dimensional scanning system. The strict 4f relay structure achieves beam constriction, comprising two coaxially arranged input and output lenses, which relay the planar beam modulated by the spatial light modulator to the two-dimensional scanning system. Adjustable apertures are located at the back focal plane of the input lens and the front focal plane of the output lens, used to select the first-order diffracted light with the highest diffraction efficiency and complete phase modulation information as the excitation light.

[0014] It also includes a telecentric scanning correction system, positioned between the two-dimensional scanning system and the objective lens; this system comprises a scanning lens and a first tube mirror. Light passing through the two-dimensional scanning system first passes through the scanning lens and then through the first tube mirror, ensuring that the light continues from the two-dimensional scanning system to the rear focal plane of the objective lens. This system can maintain a constant optical magnification during scanning, ensuring that the image quality is not affected by changes in the position or distance of the scanned object.

[0015] It also includes a bright-field imaging system, comprising a white light source, a beam splitter, and a camera. The white light source, located in front of the objective lens, emits white illumination light that uniformly illuminates the sample surface. The light signal is collected by the objective lens and precisely transmitted to the first tube mirror, then reaches the camera's detection surface via the beam splitter. The camera collects the reflected light signal from the sample surface, obtaining two-dimensional planar information about the sample surface. Throughout this process, the system strictly adheres to the 4f optical transmission principle, ensuring faithful transmission of sample information from the front focal plane of the objective lens to the rear focal plane of the first tube mirror. Finally, the light signal is received by a high-sensitivity scientific-grade camera, achieving bright-field imaging of the sample's two-dimensional structure.

[0016] The photodetector includes a focusing lens, a pinhole, and a photomultiplier tube. The pinhole allows only the in-focus light signal at the focal plane of the objective lens to pass through the pinhole and enter the photomultiplier tube to participate in imaging, while the fluorescence signal in the out-of-focus area is blocked outside the detection pinhole. The photomultiplier tube receives the weak fluorescence signal from the focal point of the objective lens, and finally reconstructs the three-dimensional structure of the sample through three-dimensional point-by-point scanning.

[0017] The two-dimensional scanning system comprises a resonant mirror and a high-precision galvanometer mirror. The resonant mirror achieves high-speed scanning in the x-axis, with its operating frequency stabilized in the kilohertz range (typically 8 kHz), utilizing mechanical resonance to achieve efficient cosine trajectory scanning. The high-precision galvanometer mirror in the y-axis enables programmable positioning scanning. Precise mechanical calibration of the two-axis scanning mechanism ensures the orthogonality of the optical path and the flatness of the scanning plane. The two-dimensional scanning system is connected to an intelligent control system via a data acquisition card. The data acquisition card outputs an analog voltage signal to the resonant mirror, controlling it to oscillate at a fixed high frequency with a cosine waveform at a set scanning angle, completing a high-speed scan within a fixed range in the x-axis. Similarly, the data acquisition card outputs a triangular wave analog voltage signal to the galvanometer mirror, controlling it to complete a scan within a fixed range in the y-axis. While oscillating at a kilohertz frequency, the resonant mirror simultaneously outputs a square wave signal synchronized with the mechanical oscillation, with its rising and falling edges corresponding to the two endpoints of the scanning trajectory. After these synchronization signals are captured by the data acquisition card, they are used as row trigger signals to control the data acquisition card to send a specified voltage signal to the galvanometer mirror according to the triangular wave waveform, so as to accurately trigger the stepping action of the galvanometer mirror in the y direction, thereby achieving strict timing coordination and completing the row-by-row x-scan in the y direction.

[0018] Axial focus adjustment is achieved using a phase-type spatial light modulator. The spatial light modulator loads a phase hologram calculated by a feedback-based aberration-free axial scanning algorithm and adjusted in real-time by a deep learning-driven adaptive control method to control the laser's axial focus position along the z-axis, achieving non-mechanical focus control. This, combined with the aforementioned two-dimensional scanning, ensures spatiotemporal consistency of the three-dimensional data acquisition. Using a kilohertz-level spatial light modulator allows for rapid switching of the focus position without moving the sample or objective lens. After the two-dimensional scanning system completes the scan, the data acquisition card outputs a frame trigger signal to the spatial light modulator. Upon receiving this signal, the spatial light modulator loads and refreshes a phase hologram to change the laser's axial focus position along the z-axis, i.e., switching to the next layer for scanning.

[0019] High-fidelity acquisition relies on a motorized adjustable focusing lens controlled by a data acquisition card. After each focus switch along the z-axis, a trigger signal is output to the lens to precisely adjust its diopter, ensuring the focus remains on the photodetector's surface and preventing distortion due to defocusing. The aforementioned timing logic significantly improves the system's frame rate by utilizing the high-speed characteristics of a resonant mirror; avoids delays caused by motion inertia through non-mechanical z-axis adjustment; and ensures precise registration of 3D data sampling through a rigorous synchronization mechanism. These characteristics make this invention particularly suitable for dynamic observation of live samples, providing a powerful tool for biomedical research.

[0020] Another objective of this invention is to provide a method for implementing an all-optical three-dimensional scanning confocal fluorescence microscopy imaging device.

[0021] The method for implementing the all-optical three-dimensional scanning confocal fluorescence microscopy imaging device of the present invention includes the following steps:

[0022] 1) The excitation source emits a polarized Gaussian beam as the excitation light, which enters the depth adaptive control device;

[0023] 2) The spatial light modulator of the depth adaptive control device loads a phase hologram calculated by a feedback aberration-free axial scanning algorithm and adjusted in real time by an adaptive control method driven by deep learning to control the phase of the excitation light, thereby achieving aberration-free axial focus movement behind the objective lens.

[0024] 3) After passing through a dichroic mirror, the first-order diffracted light is filtered through an adjustable aperture to a two-dimensional scanning system for two-dimensional planar optical scanning. The objective lens focuses the light onto the surface or internal area of ​​the sample to be observed.

[0025] 4) Fluorescent signals are generated on or inside the sample surface. They are collected by the same objective lens and transmitted back to the two-dimensional scanning system in the opposite direction of the excitation light. The two-dimensional scanning system performs the descanning and splits the excitation light beam by a dichroic mirror.

[0026] 5) The focal length is adjusted by an electrically adjustable lens according to the change of the position of the excitation light along the axis, so that the fluorescence signals at different axial positions in the sample can be focused on the detection surface of the photodetector.

[0027] 6) After passing through the pinhole located in front of the photodetector, the out-of-focus signal is filtered out and then received by the photodetector;

[0028] 7) Through the intelligent control system, the relationship between the accurate phase hologram and the axial position is obtained by adopting the deep learning-driven adaptive control method. The laser intensity is optimized according to the axial position of the focal point, and the inherent aberrations of the optical system are eliminated. The phase distribution of the depth adaptive control device, the laser intensity and phase compensation are dynamically controlled, so that the excitation light has an aberration-free axial displacement at the focal point of the high numerical aperture objective lens. The bidirectional parallel optical scanning trajectory of the two-dimensional scanning system is optimized in real time to realize all-optical three-dimensional scanning.

[0029] Among them, the deep learning-driven adaptive control method is based on a neural network architecture constrained by a physical model. It achieves autonomous optimization and real-time adjustment of the optical scanning system through end-to-end training. The neural network architecture adopts a multi-scale feature fusion convolutional neural network (CNN). The CNN processes spatially distributed data to optimize the laser intensity and the phase hologram calculated by the feedback-based aberration-free axial scanning algorithm loaded on the spatial light modulator. A multi-input multi-output CNN is used, with three types of spatial data as inputs. The CNN extracts local distortion features of the phase hologram, sidelobe patterns of the intensity distribution, and spatial distribution patterns of phase error through convolutional layers. After dimensionality compression by pooling layers, the outputs are three types of static correction values. The first type of input is a phase hologram loaded by a spatial light modulator (SLM), and the corresponding first type of output is a phase hologram optimized for spatial residuals. This is used to correct the nonlinear response of SLM pixels caused by residuals due to objective conditions of devices or environment, and to obtain an accurate relationship between the phase hologram and the axial position. The second type of input is a focal region intensity distribution image, used to improve the consistency of focal intensity at different axial positions. The corresponding second type of output is an intensity distribution equalization parameter, which optimizes the laser intensity according to the axial position of the focal point. The third type of input is a phase residual map, i.e., the distribution of the difference between the actual phase and the theoretical phase, to eliminate the inherent aberrations of the optical system. The corresponding third type of output is a static phase compensation matrix, which performs corresponding phase compensation for the inherent phase aberrations of different optical systems. The first type of dataset for convolutional neural networks loads theoretical phase holograms covering the entire phase range and different spatial frequency modes into the SLM. The phase distribution is solved inversely by using the light field intensity distribution obtained by the camera. This phase distribution is considered to be the actual phase. The difference between the actual phase and the theoretical phase is calculated, which is the nonlinear residual of the SLM pixel. The output residual inverse compensation amount is the phase hologram that optimizes the spatial residual. The dataset covers the entire pixel area of ​​the SLM and different operating temperatures.

[0030] The second type of dataset for convolutional neural networks achieves axial scanning by controlling the loading of different phase holograms by SLM. The camera records the light intensity distribution of the focal region under different axial positions and different sample scattering characteristics. The light intensity deviation is calculated based on the light field intensity of the objective lens focal plane without axial displacement, and the mapping relationship between the light intensity deviation and the laser power adjustment is calibrated through experiments. The light intensity distribution equalization parameters are output. The dataset includes different sample types and scanning depths.

[0031] Different optical systems have different inherent phase aberrations. The third type of dataset of the convolutional neural network is input to the phase residual map into the SLM. The camera detects the actual three-dimensional intensity distribution at different axial depths and field positions, calculates the difference between it and the ideal light field distribution, and further solves the corresponding phase distribution from the difference. The output is a static phase compensation matrix. The dataset covers the entire field of view of the objective lens and the scanning depth. It is tested in actual optical systems to perform corresponding phase compensation for the inherent phase aberrations of different optical systems.

[0032] The convolutional neural network was trained using the first to third types of training datasets to obtain the trained convolutional neural network.

[0033] Furthermore, the neural network architecture also includes a Long Short-Term Memory (LSTM) network and a Temporal Convolutional Network (TCN) to obtain temporal dependencies. This integrates control of the laser source, adjustable aperture, motorized adjustable focus lens, and photodetector, enabling efficient synchronization and automated operation of multiple devices and high-speed axial focusing adjustment. The LSTM network is used to model temporal dependencies. The three types of static correction values ​​output by the Convolutional Neural Network (CNN) are arranged in time series and used as input to the LSTM. A historical position error sequence reflecting dynamic drift trends is also included. The LSTM captures temporal dependencies through a gating mechanism, such as the linear evolution of phase drift caused by temperature changes over time and the cumulative pattern of position errors during continuous scanning. It outputs dynamic correction parameters: a temporally smoothed phase adjustment sequence corrects dynamic phase deviations to suppress abrupt changes in light intensity at adjacent positions; real-time phase compensation increments ensure temporal consistency of light intensity, i.e., tracking dynamic errors; and position prediction correction values ​​offset accumulated temporal deviations.

[0034] The Long Short-Term Memory (LSTM) network dataset is obtained by collecting two types of temporal data at the scanning frame rate of an all-optical 3D scanning confocal fluorescence microscopy device: First, three types of static correction sequences output by the CNN, including a phase hologram optimizing the spatial residual, light intensity distribution equalization parameters, and a static phase compensation matrix; each sequence covers the entire scanning cycle. Second, the focal position error temporal data recorded by the camera, with ambient temperature and optical path vibration parameters simultaneously collected to ensure temporal alignment with the static correction sequence. The focal position error temporal data is used to temporally sort and correct the static correction sequence output by the convolutional neural network, resulting in three types of dynamic correction parameters: a temporally smoothed phase adjustment sequence obtained by taking the inverse compensation amount of phase abrupt changes in adjacent frames, a real-time phase compensation increment calculated from the measured phase residual, and a position prediction correction value obtained by predicting the position error and taking its inverse value. The above input and output data are time-aligned and normalized, and the dataset is expanded by adding dynamic interference to form training samples for LSTM network training.

[0035] The Temporal Convolutional Network (TCN) models the temporal dependencies between devices, optimizes the control command sequence, and ensures precise synchronization of each component. The dynamic correction parameters output by the LSTM, along with device control sequences such as the power adjustment command of the laser source, the aperture control signal of the adjustable aperture, the scanning trajectory command of the 2D scanning system, the focal length parameter of the motorized adjustable lens, and the sampling frequency of the photodetector, are used as inputs to the TCN. The TCN models the temporal dependencies between multiple devices (such as the hysteresis characteristics of laser power changes and light intensity response, and the synchronization accuracy of the focusing lens movement and focal position) through causal convolution and dilated convolution. It optimizes and generates a collaborative control command sequence, outputting a multi-device collaborative control command sequence to ensure precise synchronization of the laser source power, the aperture of the adjustable aperture, the scanning trajectory of the 2D scanning system, the focal length of the motorized adjustable lens, and the phase modulation of the photodetector and SLM, achieving efficient collaborative control of the laser source, adjustable aperture, 2D scanning system, motorized adjustable lens, and photodetector.

[0036] The dataset for the temporal convolutional network is obtained by synchronously collecting two types of time-series data at the frame rate as input: one is the original control sequence of laser source power command, adjustable aperture signal, motorized adjustable focus lens focal length parameter and photodetector sampling frequency; the other is the dynamic correction parameters output by LSTM, which outputs the optimal control command sequence that each device can execute, and these are used to form training samples for training the temporal convolutional network.

[0037] The network training employs a physically constrained loss function to ensure that the prediction results conform to the laws of optical transmission: L = αL MSE +βL Physics +γL Smoothness , where L MSE To predict error loss, L Physics L is a constraint term in the Helmholtz equation. Smoothness This is the phase smoothing regularization term. The adaptive control employs a reinforcement learning framework, using scan quality (such as focus sharpness, aberration compensation, and scan efficiency) as reward signals. Control parameters are dynamically adjusted through a policy gradient method to adapt to different sample characteristics and environmental disturbances.

[0038] The feedback-based aberration-free axial scanning algorithm calculates the phase hologram as follows: According to the vector light field diffraction theory, the process of the incident light field being focused by the objective lens is mainly a diffraction process; the intensity distribution of the incident light field incident on the spatial light modulator is I0, and the desired intensity distribution of the target light field is I. t Using a spatial light modulator that is purely phase-based, only the phase distribution to be modulated on the incident light field is needed to obtain the corresponding target light field. The feedback-based aberration-free axial scanning algorithm of this invention is as follows: First, a random phase distribution Φ is added to the incident light field with a known intensity distribution. r Incident light field Li0 (I0,Φ r After one diffraction calculation, the target light field U is obtained after the first calculation. t1 (I t1 ,Φ t1 ), I t1 and Φ t1 These are the intensity and phase distributions of the target light field calculated in the first diffraction; the intensity distribution I of the target light field calculated in the first diffraction is... t1 Replace with the desired intensity distribution of the target light field I t Then, inverse diffraction calculations were performed to obtain the incident light field L corresponding to the first inverse diffraction calculation. i1 (I i1 ,Φ i1 ), I i1 and Φ i1 The intensity and phase distributions of the incident light field calculated from the first inverse diffraction are respectively; the intensity distribution I of the incident light field calculated from the first diffraction is... i1 After replacing the incident light field with its intensity distribution I0, a second diffraction calculation is performed to obtain the target light field U from the second diffraction calculation. t2 (I t2 ,Φ t2 ), I t2 and Φ t2 The intensity and phase distributions of the target light field calculated in the second diffraction are respectively; the intensity distribution I of the target light field calculated in the second diffraction is... t2 Replace with the intensity distribution I of the target light field t Then, inverse diffraction calculations were performed to obtain the incident light field L corresponding to the second inverse diffraction calculation. i2 (I i2 ,Φ i2 ), I i2 and Φ i2 The intensity and phase distributions of the incident light field are calculated separately for the second inverse diffraction; after repeating the above steps a total of k times, the target light field U calculated for the kth diffraction is obtained. tk (I tk ,Φ tk ), I tk and Φ tk Let L be the intensity and phase distribution of the target light field calculated by the k-th diffraction, and L be the incident light field calculated by the k-th inverse diffraction. ik (I ik ,Φ ik ), I ik and Φ ik Let I represent the intensity distribution and phase distribution of the incident light field calculated by the k-th inverse diffraction, where k is a natural number ≥ 1. The intensity distribution of the target light field is used as the evaluation criterion for the feedback mechanism. When the intensity distribution of the target light field calculated by the k-th diffraction is I...tk With the target light field I t When the difference between the intensity distributions is within a set threshold or reaches a set number of iterations, the phase distribution Φ of the incident light field calculated in the k-th diffraction is... ik This refers to the phase distribution that needs to be modulated on the spatial light modulator.

[0039] Advantages of this invention:

[0040] The intelligent control system of this invention employs a deep learning-driven adaptive control method, combined with a high numerical aperture objective lens, to achieve all-optical three-dimensional scanning confocal fluorescence microscopy imaging based on high-speed spatial light field modulation. This effectively filters out defocused light signals, significantly improving image signal-to-noise ratio, axial resolution, and system adaptability, enabling three-dimensional reconstruction of samples and high-contrast observation of dynamic processes. By dynamically learning the spatial and temporal characteristics of SLM pixel nonlinear response and system aberrations through CNN and LSTM, the phase hologram is optimized in real time, improving phase modulation accuracy and excitation light intensity. This effectively counteracts dynamic aberrations caused by environmental disturbances, reducing light intensity in highly sensitive areas and increasing excitation power in weak signal areas while maintaining signal-to-noise ratio, thus reducing photobleaching and extending the observation time of live samples. Duration; By modeling the temporal dependencies between devices using TCN, the control command sequence is optimized to ensure precise synchronization of each component. Simultaneous SLM phase modulation, resonant mirror scanning, adjustable lens focusing, and photodetector sampling eliminate errors and timing mismatches, automatically adapting to different sample optical characteristics and environmental interference to maintain high-quality imaging, further enhancing system robustness and applicability. The resonant mirror and spatial light modulator used in this invention are both kilohertz-level instruments, capable of high-speed three-dimensional scanning imaging. Combined with data acquisition cards and image acquisition cards, integrated control of the two-dimensional scanning system, spatial light modulator, electrically adjustable lens, and photodetector is achieved, realizing high-speed, high-resolution, and large-volume three-dimensional biological imaging. This invention employs an end-to-end feedback aberration-free axial scanning algorithm to achieve aberration-free large-range axial scanning under high numerical apertures. Compared to methods that achieve axial scanning through direct zoom, this invention successfully eliminates phase differences and achieves complete, high-quality output. During the three-dimensional scanning of the sample, this invention does not involve any mechanical movement; it completes the three-dimensional scan solely by moving the optical focus, completely avoiding the impact of motion artifacts on the final imaging effect and avoiding the slow imaging speed caused by mechanical movement. This also significantly reduces the damage to the sample caused by photobleaching and phototoxicity. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of an embodiment of the all-optical three-dimensional scanning confocal fluorescence microscopy imaging device of the present invention;

[0042] Figure 2This is a timing control diagram of an embodiment of the all-optical three-dimensional scanning confocal fluorescence microscopy imaging device of the present invention;

[0043] Figure 3 The following is a result diagram of an embodiment of the all-optical three-dimensional scanning confocal fluorescence microscopy imaging device of the present invention, wherein (a) is a field of view diagram and (b) is a resolution diagram in each direction. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0045] like Figure 1 As shown, the optical three-dimensional scanning confocal fluorescence microscopy imaging device of this embodiment includes: an excitation light source 1, a beam expander and collimator 2, a depth adaptive control device 3, a dichroic mirror 4, a 4f relay system 5, an adjustable aperture 5-1, a two-dimensional scanning system 6, a telecentric scanning correction system, a beam splitter 8, an objective lens 10, a sample holder 11, a white light source 12, a camera 13, a first reflecting mirror 14, a motorized adjustable focusing lens 15, a second reflecting mirror 16, a second tube mirror 17, a filter 18, a pinhole 19, a photodetector 20, and an intelligent control system; wherein, the two-dimensional scanning system 6 is connected to a data acquisition card, and the photodetector is connected to an image acquisition card; the excitation light source 1, the depth adaptive control device 3, the adjustable aperture 5-1, the data acquisition card, the motorized adjustable focusing lens 15, and the image acquisition card are respectively connected to the intelligent control system;

[0046] A 488 nm continuous output semiconductor laser is used as the excitation source 1 to emit a polarized Gaussian beam as the excitation light, which enters the beam expander and collimator 2.

[0047] The beam expander and collimator 2 expands and collimates the excitation light using a dual-lens system. It consists of a pair of lenses with different focal lengths precisely configured. A short focal length lens with a focal length of f1 serves as the input lens, and a long focal length lens with a focal length of f2 serves as the output lens. The beam expansion ratio is M = f2 / f1. The diameter of the expanded beam is precisely extended to 12 millimeters, which is matched with the liquid crystal panel of the spatial light modulator and transmitted to the depth adaptive control device 3.

[0048] The depth-adaptive control device 3 includes a half-wave plate 3-1, a prism 3-2, and a kilohertz-level pure phase silicon-based liquid crystal spatial light modulator 3-3, which together achieve precise control of the light field. The half-wave plate modulates the polarization state of the incident excitation light. By precisely rotating the angle of the half-wave plate, the linearly polarized light is adjusted to a polarization state that matches the optimal working polarization direction of the spatial light modulator. A prism with a 96° apex angle is used to deflect the light path, so that the incident beam forms a 6° angle with the modulator surface. The prism is used to cooperate with 1000Hz spatial light. The modulator performs optical field control on the Gaussian light after beam expansion, collimation, and modulation of polarization. The Gaussian light is reflected by one side of the prism and incident on the liquid crystal screen of the spatial light modulator. The spatial light modulator changes the axial focusing position of the excitation light behind the objective lens 10 by loading a phase hologram, thereby achieving optical field control and completing high-speed axial optical scanning. The phase of the excitation light is controlled by loading a phase hologram calculated by a feedback aberration-free axial scanning algorithm and adjusted in real time by an adaptive control method driven by deep learning, so as to achieve aberration-free axial focus movement behind the objective lens 10.

[0049] After passing through the dichroic mirror 4, the light passes through the 4f relay system 5, which includes two coaxially arranged input and output lenses, and relays the plane modulated by the spatial light modulator to the two-dimensional scanning system 6.

[0050] The adjustable aperture 5-1 is located at the back focal plane of the input lens and the front focal plane of the output lens, and is used to filter the first-order diffracted light with the highest diffraction efficiency that carries complete phase modulation information as the excitation light to the two-dimensional scanning system 6.

[0051] The two-dimensional scanning system 6 includes a resonant galvanometer and a high-precision galvanometer galvanometer; among which, such as Figure 2As shown, a resonant mirror is used in the x-direction to achieve high-speed scanning. A constant voltage is applied, the magnitude of which determines the scanning angle. The operating frequency is 7910 Hz, and high-efficiency cosine trajectory scanning is achieved through mechanical resonance. A high-precision galvanometer mirror is used in the y-direction. The galvanometer mirror's drive signal is a triangular wave with a 90% duty cycle, and the operating frequency is set to 1000 Hz. An analog voltage signal is output to the resonant mirror via a data acquisition card, controlling the mirror to oscillate continuously at a frequency of 8 kHz along a set scanning angle using a cosine waveform, completing a high-speed scan within a fixed range in the x-direction. A triangular wave analog voltage signal is also output to the galvanometer mirror via the data acquisition card to control the galvanometer... The galvanometer completes a fixed-range scan in the y-direction. While the resonant galvanometer oscillates at a kilohertz frequency, it outputs a square wave signal synchronized with the mechanical oscillation in real time, with its rising and falling edges corresponding to the two endpoints of the scan trajectory. These synchronization signals are captured by the data acquisition card and used as row trigger signals to control the card to send a specified voltage signal to the galvanometer in a triangular waveform, precisely triggering the stepping motion of the galvanometer in the y-direction, thus achieving strict timing coordination to complete the row-by-row x-scan in the y-direction. Taking a typical 128×128 sampling point scan as an example, the system executes an "S"-shaped scan trajectory: when a voltage signal is output to the resonant galvanometer... After the voltage signal is applied, the resonant mirror in the x-direction quickly completes the first row of forward scanning from (1,1) to (1,128). The edge signal generated at its endpoint instructs the data acquisition card to send a specified voltage signal to the galvanometer mirror, causing the galvanometer mirror to swing at an angle, triggering the scanning position to move one step along the y-direction, switching to the next row for scanning. Then, the resonant mirror in the x-direction immediately performs a reverse scan from (2,128) to (2,1), generating an edge signal at its endpoint, completing the second row of scanning sampling; and so on. Under the control of the data acquisition card, the two-dimensional scanning system 6 completes a two-dimensional plane with 128×128 sampling points. High-speed "S"-shaped scanning; after the two-dimensional scanning system 6 completes the two-dimensional scan, the data acquisition card will output a frame trigger signal to the spatial light modulator. After receiving the signal, the spatial light modulator will load and refresh a phase hologram to change the axial focusing position of the laser along the z-direction, that is, switch to the next layer to complete the z-direction scan; in this embodiment, a typical 128×128×100 sampling point three-dimensional scan can be completed in only 800ms, and the system scanning frame rate can reach 125Hz. The intelligent control system optimizes the bidirectional parallel optical scanning trajectory of the two-dimensional scanning system 6 in real time, which achieves an order of magnitude improvement compared with the traditional confocal microscope, and requires no mechanical movement;

[0052] The telecentric scanning correction system includes a scanning lens 7 and a first tube mirror 9. Light from the two-dimensional scanning system 6 first passes through the scanning lens 7 and then through the first tube mirror 9, ensuring that the light is relayed from the two-dimensional scanning system 6 to the back focal plane of the objective lens 10. The telecentric scanning correction system enables the system to maintain a highly consistent spot size throughout the entire scanning field of view, thereby ensuring the uniformity of imaging resolution. This design not only guarantees high-quality scanning imaging but also minimizes the need for post-processing images, improving the overall efficiency of the system. Without the telecentric scanning correction system, the scanning beam will produce significant trapezoidal distortion and field curvature, leading to positional deviations in the edge regions. Furthermore, it cannot maintain a constant magnification, and magnification fluctuations will occur as the scanning angle increases, causing measurement errors. Off-axis aberrations will degrade spot quality, increase spot size, and cause uneven energy distribution, reducing imaging resolution.

[0053] The excitation light from the telecentric scanning correction system is focused onto the sample by objective lens 10;

[0054] The white light source 12 emits white illumination light, which is uniformly illuminating the sample surface. The light signal is collected by the objective lens 10 and accurately transmitted to the first tube mirror 9, and then reaches the detection surface of the camera 13 through the beam splitter 8. The camera 13 collects the reflected light signal from the sample surface to obtain the two-dimensional planar information of the sample surface.

[0055] The sample is placed on the sample holder 11. Fluorescent signals generated on or inside the sample are collected by the same objective lens 10 and transmitted back to the two-dimensional scanning system 6 in the opposite direction of the excitation light. The two-dimensional scanning system 6 performs descanning and splits the excitation light beam through the dichroic mirror 4. The intelligent control system adjusts the focal length according to the change in the axial position of the excitation light through the electrically adjustable focusing lens 15. The focal length of the electrically adjustable focusing lens is controlled synchronously with high precision. In conjunction with the spatial light modulator, the focal fluorescence signals at different axial depths of the sample are relayed from the front focal plane of the objective lens to the rear focal plane of the second tube lens 17, i.e., within the sample. Fluorescent signals at different axial positions can be focused onto the detection surface of the photodetector 20. The adjustable lens 15 is filled with liquid. If placed vertically, the uneven distribution of liquid due to gravity will cause unexpected changes in the curvature of different areas of the lens, which cannot guarantee the imaging effect. The adjustable lens 15 is located between the first reflecting mirror 14 and the second reflecting mirror 16, which are parallel to each other, so that the motorized adjustable lens 15 can be placed flat. The stray light is removed by the filter 18 and only the signal light is retained. The defocused signal is filtered out by the pinhole 19 located in front of the photodetector 20 and then received by the photodetector 20.

[0056] The photodetector 20 includes a focusing lens, a pinhole 19, and a photomultiplier tube. The pinhole 19 allows only the in-focus light rays at the focal plane of the objective lens 10 to pass through the pinhole 19 and enter the photomultiplier tube to participate in imaging, while the fluorescence signals in the out-of-focus area are blocked outside the detection pinhole 19. The photomultiplier tube receives the weak fluorescence signal from the focal point of the objective lens 10, and finally reconstructs the three-dimensional structure of the sample through three-dimensional point-by-point scanning.

[0057] The deep learning-driven adaptive control method is based on a neural network architecture constrained by a physical model. Through end-to-end training, it achieves autonomous optimization and real-time adjustment of the optical scanning system. The neural network architecture employs a multi-scale feature fusion convolutional neural network (CNN). The CNN processes spatially distributed data, optimizing the laser intensity and the phase hologram calculated by a feedback-based aberration-free axial scanning algorithm loaded onto the spatial light modulator. A multi-input multi-output (MIMO) CNN is used, with three types of spatial data as inputs. The CNN extracts local distortion features, sidelobe patterns of the intensity distribution, and spatial distribution patterns of phase errors from the phase hologram through convolutional layers. After dimensionality compression via pooling layers, the outputs are three types of static correction values. The first type of input is a phase hologram loaded by a spatial light modulator (SLM), and the corresponding first type of output is a phase hologram optimized for spatial residuals. This is used to correct the nonlinear response of SLM pixels caused by residuals due to objective conditions of devices or environment, and to obtain an accurate relationship between the phase hologram and the axial position. The second type of input is a focal region intensity distribution image, used to improve the consistency of focal intensity at different axial positions. The corresponding second type of output is an intensity distribution equalization parameter, which optimizes the laser intensity according to the axial position of the focal point. The third type of input is a phase residual map, i.e., the distribution of the difference between the actual phase and the theoretical phase, to eliminate the inherent aberrations of the optical system. The corresponding third type of output is a static phase compensation matrix, which performs corresponding phase compensation for the inherent phase aberrations of different optical systems. The first type of dataset for convolutional neural networks loads theoretical phase holograms covering the entire phase range and different spatial frequency modes into the SLM. The phase distribution is solved inversely by the light field intensity distribution detected by camera 13. This phase distribution is considered to be the actual phase. The difference between the actual phase and the theoretical phase is calculated, which is the nonlinear residual of the SLM pixel. The output residual inverse compensation amount is the phase hologram that optimizes the spatial residual. The dataset covers the entire pixel area of ​​the SLM and different operating temperatures.

[0058] The second type of dataset of convolutional neural networks achieves axial scanning by controlling the loading of different phase holograms by SLM. Camera 13 records the light intensity distribution of the focal region under different axial positions and different sample scattering characteristics. The light intensity deviation is calculated based on the light field intensity of the focal plane of the objective lens 10 without axial displacement, and the mapping relationship between the light intensity deviation and the laser power adjustment is calibrated through experiments. The light intensity distribution equalization parameters are output. The dataset contains different sample types and scanning depths.

[0059] Different optical systems have different inherent phase aberrations. The third type of dataset of the convolutional neural network is input to the phase residual map into the SLM. The camera 13 detects the actual three-dimensional intensity distribution at different axial depths and field positions, calculates the difference between it and the ideal light field distribution, and further reverse-engineers the corresponding phase distribution from the difference. The output is a static phase compensation matrix. The dataset covers the entire field of view and scanning depth of the objective lens 10. It is tested in actual optical systems to perform corresponding phase compensation for the inherent phase aberrations of different optical systems.

[0060] The convolutional neural network was trained using the first to third types of training datasets to obtain the trained convolutional neural network.

[0061] Furthermore, the neural network architecture also includes a Long Short-Term Memory (LSTM) network and a Temporal Convolutional Network (TCN) to obtain temporal dependencies. This integrates control of the laser source, adjustable aperture, electrically adjustable focusing lens 15, and photodetector, enabling efficient synchronization and automated operation of multiple devices and high-speed axial focusing adjustment. The LSTM network is used to model temporal dependencies. The three types of static correction values ​​output by the Convolutional Neural Network (CNN) are arranged in time series and used as input to the LSTM. A historical position error sequence reflecting dynamic drift trends is also included. The LSTM captures temporal dependencies through a gating mechanism, such as the linear evolution of phase drift caused by temperature changes over time and the cumulative pattern of position errors during continuous scanning. It outputs dynamic correction parameters: a temporally smoothed phase adjustment sequence corrects dynamic phase deviations to suppress abrupt changes in light intensity at adjacent positions; real-time phase compensation increments ensure temporal consistency of light intensity, i.e., tracking dynamic errors; and position prediction correction values ​​offset accumulated temporal deviations.

[0062] The Long Short-Term Memory (LSTM) network dataset is obtained by collecting two types of temporal data as input according to the system scan frame rate: First, three types of static correction sequences output by the CNN, including a phase hologram for optimizing spatial residuals, light intensity distribution equalization parameters, and a static phase compensation matrix; each sequence covers a complete scan cycle. Second, the focal position error temporal data recorded by camera 13, with simultaneous acquisition of ambient temperature and optical path vibration parameters to ensure temporal alignment with the static correction sequence. The focal position error temporal data is used to temporally sort and correct the static correction sequence output by the convolutional neural network, resulting in three types of dynamic correction parameters: a temporally smoothed phase adjustment sequence obtained by taking the inverse compensation amount of phase abrupt changes in adjacent frames; a real-time phase compensation increment calculated from the measured phase residuals; and a position prediction correction value obtained by predicting the position error and taking its inverse value. The above input and output data are time-aligned and normalized, and the dataset is expanded by adding dynamic interference to form training samples for LSTM network training.

[0063] The Temporal Convolutional Network (TCN) models the temporal dependencies between devices, optimizes the control command sequence, and ensures precise synchronization of each component. The dynamic correction parameters output from the LSTM, along with device control sequences such as the power adjustment command of the laser source, the aperture control signal of the adjustable aperture, the focal length parameter of the motorized adjustable lens, and the sampling frequency of the photodetector, are used as inputs to the TCN. The TCN models the temporal dependencies between multiple devices (such as the hysteresis characteristics of laser power changes and light intensity response, and the synchronization accuracy of the focusing lens movement and focal position) through causal convolution and dilation convolution. It optimizes and generates a cooperative control command sequence, outputting a multi-device cooperative control command sequence to ensure precise synchronization of the laser source power, the aperture of the adjustable aperture, the scanning trajectory of the 2D scanning system, the focal length of the motorized adjustable lens, and the phase modulation of the photodetector and SLM. This achieves efficient cooperative control of the laser source, adjustable aperture, 2D scanning system, motorized adjustable lens, and photodetector. The timing control diagram is shown below. Figure 2 As shown.

[0064] The dataset for the temporal convolutional network is obtained by synchronously collecting two types of time-series data at the frame rate as input: one is the original control sequence of laser source power command, adjustable aperture signal, motorized adjustable focus lens focal length parameter and photodetector sampling frequency; the other is the dynamic correction parameters output by LSTM, which outputs the optimal control command sequence that each device can execute, and these are used to form training samples for training the temporal convolutional network.

[0065] The network training employs a physically constrained loss function to ensure that the prediction results conform to the laws of optical transmission: L = αL MSE +βL Physics +γL Smoothness , where L MSE To predict error loss, L PhysicsL is a constraint term in the Helmholtz equation. Smoothness This is the phase smoothing regularization term. The adaptive control employs a reinforcement learning framework, using scan quality (such as focus sharpness, aberration compensation, and scan efficiency) as reward signals. Control parameters are dynamically adjusted through a policy gradient method to adapt to different sample characteristics and environmental disturbances.

[0066] like Figure 3 As shown, the field of view and three-dimensional resolution of this invention can be calibrated by imaging fluorescent microspheres. (a) is the result of imaging a 5μm fluorescent microsphere using a high-speed spatial light field-controlled all-optical three-dimensional scanning confocal fluorescence microscopy imaging device, with an imaging field of view of 186.18μm × 186.18μm in the XY direction; (b) is the full width at half maximum (FWHM) plot of light intensity obtained by imaging a 100nm microsphere using the method of this invention, with 202nm, 221nm, and 554nm in the X, Y, and Z directions, respectively. Experimental results show that, while maintaining sufficient field of view coverage and near-diffraction-limited resolution, the technical solution of this invention achieves a significant improvement in three-dimensional imaging speed compared to traditional confocal microscopy systems, expanding the volume imaging range of all-optical scanning confocal microscopy. This breakthrough is mainly due to the innovative non-mechanical axial scanning scheme and the collaborative working mode of the high-speed resonant galvanometer, enabling the system to meet the temporal requirements of dynamic observation of live samples while maintaining subcellular spatial resolution.

[0067] Finally, it should be noted that the purpose of disclosing the embodiments is to help further understand the present invention. However, those skilled in the art will understand that various substitutions and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the embodiments, and the scope of protection of the present invention is defined by the claims.

Claims

1. A full-light three-dimensional scanning confocal fluorescence microscopy imaging device, characterized in that, The fluorescence microscopy imaging device includes: an excitation source, a depth adaptive control device, a dichroic mirror, an adjustable aperture, a two-dimensional scanning system, a telecentric scanning correction system, an objective lens, a motorized adjustable focusing lens, a pinhole camera, a photodetector, and an intelligent control system; wherein, the two-dimensional scanning system is connected to a data acquisition card, and the photodetector is connected to an image acquisition card; the excitation source, the depth adaptive control device, the adjustable aperture, the data acquisition card, the motorized adjustable focusing lens, and the image acquisition card are respectively connected to the intelligent control system; The excitation source emits a polarized Gaussian beam as excitation light, which enters the depth adaptive control device. The phase of the excitation light is controlled by loading a phase hologram. After passing through a dichroic mirror, the first-order diffracted light is filtered out by an adjustable aperture and sent to the two-dimensional scanning system for two-dimensional planar optical scanning. The light is then focused onto the sample by the objective lens. The two-dimensional scanning system uses a resonant mirror in the x-direction, which achieves high-efficiency cosine trajectory scanning through mechanical resonance characteristics. The sample generates a fluorescence signal, which is collected by the same objective lens and transmitted back to the two-dimensional scanning system in the opposite direction of the excitation light. The two-dimensional scanning system performs a descan and splits the excitation light beam by a dichroic mirror. The electric adjustable focusing lens adjusts the focal length according to the change of the excitation light along the axial position, so that the fluorescence signals at different axial positions in the sample can be focused on the detection surface of the photodetector. After passing through the pinhole, the fluorescence is received by the photodetector. The intelligent control system employs a deep learning-driven adaptive control method to optimize the phase hologram calculated by the feedback aberration-free axial scanning algorithm, obtaining an accurate relationship between the phase hologram and the axial position. Through a multi-input multi-output convolutional neural network, it extracts local distortion features, light intensity distribution, and phase error of the phase hologram, generating a static correction amount to correct the pixel nonlinear response of the spatial light modulator. Furthermore, it optimizes the laser's light intensity based on the axial position of the focal point and performs phase compensation. Combining a long short-term memory network and a temporal convolutional network, it obtains a temporal dependency relationship, dynamically eliminating interference from ambient temperature and optical path vibrations. This ensures aberration-free axial displacement of the excitation light at the focal point behind the objective lens. In addition, it optimizes the bidirectional parallel optical scanning trajectory of the two-dimensional scanning system in real time, collaboratively controlling the laser source, adjustable aperture, two-dimensional scanning system, electrically adjustable focusing lens, and photodetector to achieve all-optical three-dimensional scanning.

2. The fluorescence microscopy imaging device as described in claim 1, characterized in that, It also includes a beam expander and collimator, which uses a dual-lens system for beam expansion and collimation. The diameter of the expanded beam is matched with the spatial light modulator of the depth adaptive control device.

3. The fluorescence microscopy imaging device as described in claim 1, characterized in that, The depth adaptive control device includes a half-wave plate, a prism, and a spatial light modulator. The half-wave plate modulates the polarization state of the incident excitation light. By precisely rotating the angle of the half-wave plate, the linearly polarized light is adjusted to a polarization state that matches the optimal working polarization direction of the spatial light modulator. The prism, in conjunction with the spatial light modulator, completes the optical field control of the Gaussian light. The Gaussian light is reflected by one side of the prism and incident on the spatial light modulator. The spatial light modulator changes the axial position of the excitation light focused behind the objective lens by loading a phase hologram.

4. The fluorescence microscopy imaging device as described in claim 1, characterized in that, It also includes a telecentric scanning correction system, which is disposed between the two-dimensional scanning system and the objective lens; it includes a scanning lens and a first tube mirror, and the light passing through the two-dimensional scanning system first passes through the scanning lens and then through the first tube mirror, ensuring that the light is relayed from the two-dimensional scanning system to the back focal plane of the objective lens.

5. The fluorescence microscopy imaging device as described in claim 1, characterized in that, The two-dimensional scanning system includes a resonant galvanometer and a high-precision galvanometer galvanometer. The resonant galvanometer is used in the x-direction to achieve high-speed scanning with a stable operating frequency in the kilohertz range. It achieves high-efficiency cosine trajectory scanning through mechanical resonance characteristics. The high-precision galvanometer galvanometer is used in the y-direction to achieve programmable positioning scanning.

6. A method for implementing the all-optical three-dimensional scanning confocal fluorescence microscopy imaging device as described in claim 1, characterized in that, The implementation method includes the following steps: 1) The excitation source emits a polarized Gaussian beam as the excitation light, which enters the depth adaptive control device; 2) The spatial light modulator of the depth adaptive control device loads a phase hologram calculated by a feedback aberration-free axial scanning algorithm and adjusted in real time by an adaptive control method driven by deep learning to control the phase of the excitation light, thereby achieving aberration-free axial focus movement behind the objective lens. 3) After passing through a dichroic mirror, the first-order diffracted light is filtered through an adjustable aperture to a two-dimensional scanning system for two-dimensional planar optical scanning. The objective lens focuses the light onto the surface or internal area of ​​the sample to be observed. 4) Fluorescent signals are generated on or inside the sample surface. They are collected by the same objective lens and transmitted back to the two-dimensional scanning system in the opposite direction of the excitation light. The two-dimensional scanning system performs the scanning and splits the excitation light beam by a dichroic mirror. 5) The focal length is adjusted according to the change of the position of the excitation light along the axis by the electrically adjustable focusing lens, so that the fluorescence signals at different axial positions in the sample can be focused on the detection surface of the photodetector. 6) After passing through the pinhole located in front of the photodetector, the out-of-focus signal is filtered out and then received by the photodetector; 7) Through the intelligent control system, the relationship between the accurate phase hologram and the axial position is obtained by adopting the deep learning-driven adaptive control method. The laser intensity is optimized according to the axial position of the focal point, and the inherent aberrations of the optical system are eliminated. The phase distribution of the depth adaptive control device, the laser intensity and phase compensation are dynamically controlled, so that the excitation light has aberration-free axial displacement at the focal point of the high numerical aperture objective lens. The bidirectional parallel optical scanning trajectory of the two-dimensional scanning system is optimized in real time to realize all-optical three-dimensional scanning.

7. The implementation method as described in claim 6, characterized in that, The deep learning-driven adaptive control method is based on a neural network architecture constrained by a physical model. The neural network architecture employs a multi-input multi-output (MIMO) convolutional neural network (CNN). The input consists of three types of spatial data. The CNN extracts local distortion features of the phase hologram, sidelobe patterns of the light intensity distribution, and the spatial distribution of phase error through convolutional layers. After dimensionality compression via pooling layers, the output consists of three types of static corrections. The first type of input is the phase hologram loaded by the spatial light modulator (SLM), and the corresponding first type of output is a phase hologram with optimized spatial residuals. This corrects the nonlinear response of SLM pixels caused by residuals due to objective conditions of the device or environment, obtaining an accurate relationship between the phase hologram and the axial position. The second type of input is the light intensity distribution image of the focal region, used to improve the consistency of focal light intensity at different axial positions. The corresponding second type of output is a light intensity distribution equalization parameter, optimizing the corresponding light intensity of the laser based on the axial position of the focal point. The third type of input is the phase residual map, i.e., the distribution of the difference between the actual phase and the theoretical phase, eliminating inherent aberrations of the optical system. The corresponding third type of output is a static phase compensation matrix, performing corresponding phase compensation for the inherent phase aberrations of different optical systems.

8. The implementation method as described in claim 7, characterized in that, The neural network architecture also includes a Long Short-Term Memory (LSTM) network and a Temporal Convolutional Network (TCN) to obtain temporal dependencies. It integrates control of the laser source, adjustable aperture, two-dimensional scanning system, motorized adjustable focus lens, and photodetector to control the efficient synchronization and automated operation of multiple devices and high-speed axial focusing adjustment.

9. The implementation method as described in claim 8, characterized in that, The dataset for the Long Short-Term Memory (LSTM) network is obtained by collecting two types of temporal data at the scanning frame rate of an all-optical 3D scanning confocal fluorescence microscopy device: one is the three types of static correction sequences output by the CNN, including phase holograms for optimizing spatial residuals, light intensity distribution equalization parameters, and static phase compensation matrices, with each sequence covering the entire scanning cycle; the other is the focal position error temporal data recorded by the camera, with ambient temperature and optical path vibration parameters collected simultaneously to ensure time alignment with the static correction sequence. The focal position error temporal data is used to temporally sort and correct the static correction sequence output by the convolutional neural network, and the output is three types of dynamic correction parameters: a temporally smoothed phase adjustment sequence obtained by taking the inverse compensation amount of phase abrupt changes in adjacent frames, a real-time phase compensation increment calculated by the measured phase residuals, and a position prediction correction value obtained by predicting the position error and taking the inverse value.

10. The implementation method as described in claim 9, characterized in that, The Temporal Convolutional Network (TCN) models temporal dependencies and optimizes control command sequences. It uses the dynamic correction parameters output from the LSTM, along with the power adjustment commands of the laser source, the aperture control signal of the adjustable aperture, the scanning trajectory commands of the 2D scanning system, the focal length parameters of the motorized adjustable lens, and the sampling frequency of the photodetector, as inputs to the TCN. Through causal convolution and dilated convolution, the TCN models the temporal dependencies between multiple devices, optimizes the generation of collaborative control command sequences, and outputs a multi-device collaborative control command sequence. This ensures precise synchronization of the laser source power, the aperture of the adjustable aperture, the scanning trajectory of the 2D scanning system, the focal length of the motorized adjustable lens, and the phase modulation of the photodetector and SLM, achieving efficient collaborative control of the laser source, adjustable aperture, 2D scanning system, motorized adjustable lens, and photodetector.

Images