Multi-channel fluorescence imaging system based on time sequence pulse excitation and asynchronous acquisition

Abstract

The invention discloses a multichannel fluorescence imaging system based on time sequence pulse excitation and asynchronous acquisition, and particularly relates to the technical field of biomedical engineering and optical microscopic imaging. Comprising an inverted microscope main body, a multi-wavelength laser excitation module, a high-speed Scmos imaging device supporting hardware triggering, a lower computer time sequence control unit based on a microcontroller, and an upper computer calculation processing platform. The lower computer utilizes a hardware timer to generate a synchronous control signal with microsecond-level precision, adopts a time division multiplexing strategy to control opening of lasers with different wavelengths, and is matched with hardware handshake logic to force the camera to perform global shutter synchronous exposure. The upper computer builds a memory asynchronous buffer architecture, decouples high-bandwidth image acquisition and a low-speed hard disk storage process by using a producer-consumer model, and temporarily stores original data by using a memory buffer pool.

Description

Technical Field

[0001] This invention relates to the fields of biomedical engineering and optical microscopy imaging technology, specifically to a multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition. Background Technology

[0002] Fluorescence microscopy is a core tool in life science research for observing cellular microstructures (such as mitochondria and lysosomes) and dynamic physiological processes (such as synaptic transmission and autophagy). To analyze complex biological processes, researchers typically need to simultaneously label different biomolecules using multiple fluorescent probes of different colors, i.e., multicolor / multichannel imaging.

[0003] Existing multi-channel fluorescence imaging technology mainly faces the following technical bottlenecks: Spectral crosstalk: Due to the broadband characteristics of the emission spectrum of fluorophores, the emission light from short-wavelength channels often "leaks" into long-wavelength channels (e.g., GFP signal crosstalks into the YFP channel), resulting in a decrease in the imaging signal-to-noise ratio and seriously affecting the accuracy of colocalization analysis.

[0004] Insufficient temporal resolution: Traditional wide-field microscopes rely on mechanical filter wheels driven by stepper motors to switch channels. The switching time is usually between tens of milliseconds and hundreds of milliseconds, and sometimes even several seconds. It is also accompanied by physical vibration, making it difficult to capture rapid biological dynamics at the millisecond level.

[0005] Data throughput bottleneck: With the increase in pixel count of scientific-grade CMOS (Scmos) cameras, the data flow at full resolution is extremely large (e.g., 12 megapixels at 100fps generates approximately 1.2GB / s of data). Traditional host computer software uses a "synchronous acquisition + direct disk writing" architecture, which is limited by the write speed of SATA or USB interfaces, making it prone to data congestion and frame drops.

[0006] While acousto-optic tunable filters (AOTFs) address some of the issues, their high cost limits their widespread adoption in ordinary laboratories. Therefore, there is an urgent need for a low-cost, high-performance solution based on general-purpose hardware. Summary of the Invention

[0007] To address this, the present invention provides a multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition. It utilizes an embedded microcontroller to achieve microsecond-level photoelectric synchronization and combines upper computer memory buffering technology to achieve high-speed, low-crosstalk, multi-channel fluorescence imaging, thereby solving the problems mentioned in the background art.

[0008] The following technical solution is provided: a multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition, including: The inverted microscope body is used to support the target sample and provide an optical imaging path. Multi-wavelength laser excitation module, used to provide excitation light of different wavelengths; The high-speed imaging device includes a Scmos camera that supports hardware triggering and global shutter mode; a static multi-band emission filter array is set at the front end of the high-speed imaging device. The multi-wavelength laser excitation module outputs multiple excitation beams of different wavelengths, which are combined into a coaxial optical path and incident on the inverted microscope body to excite the target sample using multiple wavelengths. The fluorescence / signal light emitted by the target sample is filtered by a static multi-band emission filter array at the front end of the high-speed imaging device to achieve the separation of different emission signals. The high-speed imaging device uses a Scmos camera, which is triggered synchronously with the laser excitation sequence and exposed synchronously with the global shutter to complete high-speed, low-distortion imaging acquisition. The lower-level timing control unit, based on the microcontroller MCU, is configured with a time-division multiplexing strategy. It controls the multi-wavelength laser excitation module and the high-speed imaging device by concurrently outputting hardware pulse signals with microsecond-level precision. It isolates excitation light of different wavelengths on the physical time axis, realizes time-division multiplexing of excitation light of different wavelengths in millisecond-level time intervals, divides the imaging period into non-overlapping time windows, and sends a global trigger level to the optically isolated input terminal of the high-speed imaging device to force photosensitive reset, thereby aligning the physical timing of laser excitation with the global exposure window of the camera. The host computer computing and processing platform is used for system configuration, asynchronous data stream processing, and image storage.

[0009] Preferably, the time-division multiplexing strategy is accomplished through two sets of complementary interrupt service logic and hardware handshake logic.

[0010] Preferably, the first timer control parameters, timing constraints, and second timer control parameters corresponding to each laser control channel are configured within the MCU, and a hardware handshake link is constructed between the MCU input / output interface GPIO and the opto-isolation circuit, which is then connected to the opto-isolation input terminal of the high-speed imaging device.

[0011] Preferably, a dual-timer cooperative control model based on a discrete-time state machine is built inside the MCU to implement the interrupt service logic, as follows: Set the MCU system clock frequency to If the timer prescaler is PSC, then the minimum time granularity of the hardware counter is... satisfy: ; The MCU internally maintains a laser channel index variable k, k ∈{0, 1, ..., M-1}, where M is the total number of laser channels; the control model's working logic is as follows: (1) Excited state: Start the first timer TIM_ON and set its auto-reload register value to... At this time, the single-channel excitation duration Defined by the following formula: ; The MCU retrieves the light intensity mapping from the preset light intensity mapping table based on the current index variable k. Read the corresponding pulse width modulation value Write to the Comparator Register (CCR) k This results in the output of duty cycle D at the corresponding GPIO port. k satisfy: ; (2) Dead zone switching state_DeadTime: When the first timer overflows and triggers the first interrupt, the MCU immediately disables the first timer output and starts the second timer TIM_OFF, with a reload value of [value to be filled in]. Dead zone interval duration satisfy: ; (3) Looping logic: When the second timer overflows and triggers the second interrupt, the MCU executes the channel index update algorithm: ; Where mod represents the modulo operation. It is the total number of channels, which the MCU determines based on k. next Update the CCR parameters for the next cycle and restart the first timer to achieve hardware closed-loop control without software intervention.

[0012] Preferably, the lower-level timing control unit is configured with hardware handshake logic based on general-purpose input / output interface (GPIO) and opto-isolation circuit. The GPIO pins of the MCU are connected to the external trigger input of the imaging device to execute the following microsecond-level timing actions: (1) Trigger Response: At the start of a single imaging cycle, the MCU generates a TTL synchronization trigger signal, i.e., pulse width The effective edge of this signal directly drives the imaging device that supports global shutter mode, controlling all pixels of the sensor array to start photoelectric integration at the same time, forming a common exposure window with an absolute time reference. (2) The MCU synchronously outputs the laser PWM modulation signal, i.e., the effective light emission pulse width. The system satisfies the temporal envelope constraint to ensure the complete energy of laser excitation. Completely enclosed within the camera's global exposure window [ , Within this range, lossless integration of photon energy is achieved, and the time-gating characteristics of the global shutter are utilized to effectively suppress ambient stray light noise outside the exposure window; among which, It is the start time of exposure. It is the end time of the exposure; The temporal envelope constraint inequality is as follows: ; in, Due to the inherent turn-on delay of the laser driver, This provides a timing safety margin for the system.

[0013] Preferably, the host computer computing platform is configured with an underlying asynchronous memory buffer architecture based on a producer-consumer concurrent model to decouple image acquisition from hard disk writing; the specific execution logic is as follows: (1) Physical memory pre-allocation and addressing topology: Define the pixel width of a single frame image as The height is The depth of a single pixel byte is The amount of image data per frame ; The host computer pre-allocates a contiguous memory space in the system's physical memory as a circular data buffer, setting its total number of data block nodes to [number missing]. Then pre-allocate the total memory size The starting address of this memory block is defined as ; The system initializes and maintains independent and atomic write pointers. With read pointer All initial values ​​are set to 0; (2) Zero-copy ultra-fast enqueue based on pointer offset: When the raw image data of the i-th frame triggers a data transmission interruption, the producer thread performs the following status check and absolute address lookup:

[0014] First, determine the status of idle nodes. If the condition is met... If the buffer has sufficient space, then the physical absolute target address of the current node in memory is calculated: ; Then, the system's underlying memory block shifting instruction is invoked to move the block of length to... The binary data is from the sensor's source address. Copy exactly to This avoids heap object allocation in high-level languages; after copying is complete, the producer obtains a high-precision system timestamp. The queue tuple containing pointers and time parameters ( Push the code into the non-blocking queue Q and update the write pointer: ; (3) Asynchronous consumption of matrix mapping demixing and pointer dequeueing: A standalone consumer thread listens to the non-blocking queue Q and extracts the queued tuples. ), and update the read pointer synchronously. To release the corresponding memory node; the consumer based on the extracted Pointers directly access the original image matrix And execute them sequentially: a) Color space conversion: Perform matrix mapping using interpolation algorithms. Reconstruct the Bayer array into an RGB matrix; b) Spectral unmixing: Constructing a linear mixture mathematical model based on prior spectral observations. Where Y is the current observed signal matrix. A represents the known endmember spectral matrix extracted by the system calibration, containing the autofluorescence background spectrum and the target probe spectrum; E represents the system noise term; the consumer backend solves for the feature abundance matrix C using a non-negative matrix factorization algorithm, removes the autofluorescence background interference signal, and reconstructs a clean multi-channel target image. ; c) Timing Encapsulation and Disk Writing: Dequeuing the timestamp Encoding is included in the image file header identifier, which will then be used to demix the target image. Asynchronous I / O writes to the hard disk are performed according to a preset compression algorithm.

[0015] The present invention has the following advantages: 1. This invention completely eliminates crosstalk caused by the overlap of excitation light spectra through temporal physical isolation, significantly improves the signal-to-noise ratio of multicolor imaging, and achieves zero crosstalk imaging.

[0016] 2. This invention combines hardware PWM triggering with software memory buffering to achieve extreme frame rate acquisition at full resolution, which is suitable for observing fast processes such as calcium ion scintillation and achieves high-speed lossless recording.

[0017] 3. This invention is based on a general-purpose STM32 chip and a consumer-grade microscope camera. It does not require an expensive AOTF or a dedicated acquisition card. It is easy to modify and upgrade ordinary inverted microscopes, and is low-cost and easily expandable. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the system initialization and interaction process of the system of the present invention.

[0019] Figure 2 This is a schematic diagram simulating the excitation and emission optical paths of the overall architecture of this invention.

[0020] Figure 3 This is the logic diagram of the dual timer autonomous loop of the lower-level machine in this invention.

[0021] Figure 4 This is a schematic diagram of the asynchronous acquisition "producer-consumer" thread model of the host computer in this invention.

[0022] Figure 5 This is an oscilloscope waveform diagram of the three-channel laser and camera exposure signals in the timing mode of this invention. Detailed Implementation

[0023] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1: Hardware Circuit and Optical Path System Construction: Optical path section: Combination Figure 2 The overall architecture, including the excitation and emission optical paths, is illustrated in the schematic diagram. This system is based on an inverted fluorescence microscope. The excitation end uses three solid-state lasers with wavelengths of 488nm, 532nm, and 635nm. The three laser beams are spatially combined using dichroic mirrors, collimated, and then coaxially introduced into the microscope's optical path. At the imaging end, a multi-band pass filter is installed in front of the camera. This filter has high transmittance for the fluorescence emission peaks corresponding to the three wavelengths (e.g., 525nm, 580nm, and 680nm) and cuts off the excitation wavelength. This design ensures that there are no moving mechanical parts in the imaging optical path.

[0025] Circuit section: The core controller uses the STM32F103C8T6 microcontroller.

[0026] Laser control interface: GPIO pins PA0, PA1, and PA2 are connected to the TTL modulation terminals of the three lasers, configured in push-pull output mode. In the circuit configuration, these three pins are mapped to channels 1, 2, and 3 (CH1-CH3) of the general-purpose timer TIM2. The MCU is configured in "multiplexed push-pull output (AF_PP)" mode to allow the timer hardware to directly generate PWM waveforms, thereby achieving stepless adjustment of laser intensity.

[0027] Camera trigger interface: GPIO pin PA3 is connected to the camera's optocoupler-isolated trigger input. PA3 is mapped to channel 4 (CH4) of TIM2 and is also configured as a multiplexed push-pull output. Since both the camera trigger signal and the laser drive signal are driven by the same 16-bit counter of TIM2, they achieve strict synchronization at the microsecond level in hardware, eliminating timing jitter caused by software control.

[0028] System initialization and communication interaction: combined Figure 1 As shown in the system initialization and interaction flowchart, after the host computer user sets the exposure time, laser intensity, and working mode through the interface, they send a start command (such as 'S' or 'F') to the slave computer via USART1 (PA9 / PA10, baud rate 115200bps). Figure 1 As shown in the STM32 response logic on the right, after the MCU triggers the USART interrupt to receive the instruction, it parses the intensity variable and performs hardware configuration (configuration pins, PWM and dead timer), and then starts the first frame TIM2, officially entering the timing loop.

[0029] Example 2: Lower-level firmware timing logic: This embodiment details the control logic of the STM32 firmware, which achieves microsecond-level synchronization based on a hardware timer.

[0030] Timer configuration: TIM2 (ON_TIMER): Responsible for defining the "excitation window". Configured as PWM output mode, with a period of 10ms (corresponding to 100fps). Within this window, TIM2 simultaneously sends a 1ms trigger pulse to the camera and outputs a PWM waveform with an adjustable duty cycle to the currently selected laser channel (e.g., outputting the corresponding pulse width according to 0-100% intensity), achieving synchronization between excitation and exposure.

[0031] TIM3 (OFF_TIMER): Controls the "channel switching interval". Defines the "dead interval". Configured in pure counting mode with a period of 1ms, it is used for fluorescence attenuation and data readout during channel switching to prevent residual fluorescence interference.

[0032] Time-sequence polling state machine: State S1 (excitation): such as Figure 3 As shown at the top, the MCU activates TIM2, pulling high the current channel to illuminate the current laser (e.g., PA0 / L1) and the camera trigger pin (PA3) according to a preset intensity. This process is a hardware parallel output without CPU intervention. The camera detects a TTL trigger high level (rising edge) and initiates global exposure.

[0033] Timing Envelope Limitation Design: To address the pulse swallowing and synchronization errors caused by the physical turn-on delay of the laser, the camera trigger pulse width (T1) set in TIM2 in this embodiment is configured to strictly enclose the laser PWM modulation pulse width (T2). That is, T1 ≥ T2 + Δt. After the PWM signal is emitted, the laser reaches the effective emission threshold only after experiencing a driver ramp-up delay of Δt (typically 10 to 50 microseconds). This envelope design ensures that the actual physical emission process occurs 100% within the camera's global exposure window, avoiding focal plane blur (smear effect) caused by exposure during the readout process.

[0034] State S2 (Interrupt Switching): such as Figure 3 As shown on the right, a TIM2 counter overflow generates an interrupt (Update IRQ). The interrupt service routine shuts down TIM2, pulls all laser pins and camera trigger pins low, and starts TIM3.

[0035] State S3 (Dead Zone and Index Update): For example Figure 3 As shown at the bottom and left, an interrupt is generated when the TIM3 counter overflows. The interrupt service routine shuts down TIM3, executes the algorithm current_laser_index = (current_laser_index + 1) %3, and reads the preset light intensity parameters for the next channel (such as L2) from the memory array, loading them into the TIM2's Comparison Register (CCR). TIM2 is then restarted, seamlessly transitioning to the next frame acquisition.

[0036] Constant-on debugging mode: When a single-channel constant-on command is received, the system keeps TIM2 on but turns off TIM3, and forces the duty cycle of the camera trigger channel (CH4) to be set to 0, outputting only the laser PWM waveform for optical path calibration or power testing.

[0037] Based on the above mathematical model, using the STM32F103 main frequency... Taking 72 MHz as an example. If the prescaler PSC is set to 71, then the counter time granularity... = 1μs.

[0038] Excitation window configuration: If the exposure time needs to be set to 10 ms, then the first timer reload value is... Set to 9999, the calculation is as follows: = (9999+1) × 1μs = 10 ms.

[0039] Dead window configuration: If you need to set the dead time to 2ms, then the second timer reload value... Set to 1999, the calculation is as follows: = 2 ms.

[0040] Polling logic: Set the total number of channels. = 3. When the current index k=2 (corresponding to the third wavelength) ends and the dead timer is completed, the system executes k = (2+1) mod 3 = 0, automatically resets to the first wavelength channel, and completes the closed loop.

[0041] Timing waveform characteristics description: Combination Figure 5 The oscilloscope waveforms of the three-channel laser and camera exposure signals in timing mode, along with the description of the level change trends and physical meanings during stable operation, are as follows: Figure 5 The horizontal axis represents time (unit: ms), and the vertical axis, from top to bottom, represents the modulation level of the first to third laser channels (CH1 to CH3) and the TTL level of the camera trigger channel (CH4). The entire waveform exhibits a strictly periodic, step-like alternation.

[0042] (1) In the 0-7 ms interval: CH1 and CH4 generate rising edge transitions simultaneously. CH1 continuously outputs a valid high level of a preset width to drive laser L1 to light up. At the same time, the high level of CH4 forces the camera to maintain exposure (exposure / lighting time is set to 7.0ms). CH2 and CH3 strictly maintain a low level off state in this interval.

[0043] (2) Within the 7-10 ms interval: CH1 and CH4 simultaneously drop to low level, and the system enters the dead interval (set to 3.0ms). During this period, all lasers are in the off state, and the camera uses this safe window to complete the sensor charge readout, avoiding focal plane blur.

[0044] (3) Subsequent timing evolution: At 10 ms, CH2 and CH4 are simultaneously pulled up, exciting the second wavelength (L2); at 17 ms, the dead zone is entered; at 20 ms, CH3 and CH4 are simultaneously pulled up, exciting the third wavelength (L3). By 30 ms, a complete multi-channel polling cycle ends.

[0045] The waveform diagram visually demonstrates the hardware-level concurrent output capability of this system: the camera trigger signal (CH4) achieves a perfect time envelope for each single effective laser pulse (CH1 / CH2 / CH3). Simultaneously, each laser channel achieves non-overlapping physical absolute isolation on the time axis, eliminating spectral crosstalk in multi-wavelength imaging at its source.

[0046] Example 3: Design of Asynchronous Data Acquisition Software for Host Computer Combination Figure 4The schematic diagram of the asynchronous acquisition "producer-consumer" thread model of the host computer is shown. This embodiment is based on the host computer software architecture of Python and adopts a multi-threaded producer-consumer model to solve the problem of high data throughput of 100fps.

[0047] Memory buffer data structures: such as Figure 4 As shown in the intermediate steps, the software defines a global list `self.memory_buffer` as a circular buffer. Each element is a dictionary containing `data` (raw binary data), `width`, `height`, and `timestamp`.

[0048] Producer thread (data acquisition phase): Zero-copy and pointer passing mechanism: such as Figure 4 As shown on the left, the camera's hardware trigger signal drives data transmission. To achieve zero-copy and pointer-passing mechanisms and overcome data congestion at 100fps, the producer thread is strictly prohibited from performing image matrix copying or format conversion operations in its callback function. After data is written to pre-allocated circular memory via `ctypes.memmove`, the producer only places the memory address pointer and timestamp of that data block into a thread-safe queue. The consumer thread accesses the original data by retrieving the pointer. This design compresses the producer thread's single interrupt response time to less than 1 millisecond, completely eliminating reverse blocking and frame drops caused by disk I / O fluctuations.

[0049] Consumer thread (image processing and storage stage): such as Figure 4 As shown on the right, the consumer thread, which runs independently in the background, is responsible for reading RAW data and finally writing it to the hard disk (storage).

[0050] Strategy A (pure memory buffer): Data is temporarily stored in RAM and written to the hard disk after the collection is completed. This strategy is suitable for ultra-high-speed short-term recording.

[0051] Strategy B (Pipeline-based On-Demand Data Acquisition): Data is retrieved from the queue in real time, color space conversion is performed using OpenCV (BayerRG -> RGB), and the data is saved as either a highly compressed JPG or a lossless TIFF file, depending on user requirements. This architecture completely decouples I / O operations from acquisition operations, ensuring that fluctuations in hard drive write speed do not affect the acquisition frame rate.

[0052] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition, characterized in that: include: The inverted microscope body is used to support the target sample and provide an optical imaging path. Multi-wavelength laser excitation module, used to provide excitation light of different wavelengths; The high-speed imaging device includes a Scmos camera that supports hardware triggering and global shutter mode; a static multi-band emission filter array is set at the front end of the high-speed imaging device. The multi-wavelength laser excitation module outputs multiple excitation beams of different wavelengths, which are combined into a coaxial optical path and incident on the inverted microscope body to excite the target sample using multiple wavelengths. The fluorescence / signal light emitted by the target sample is filtered by a static multi-band emission filter array at the front end of the high-speed imaging device to achieve the separation of different emission signals. The high-speed imaging device uses a Scmos camera, which is triggered synchronously with the laser excitation sequence and exposed synchronously with the global shutter to complete high-speed, low-distortion imaging acquisition. The lower-level timing control unit, based on the microcontroller MCU, is configured with a time-division multiplexing strategy. It controls the multi-wavelength laser excitation module and the high-speed imaging device by concurrently outputting hardware pulse signals with microsecond-level precision. It isolates excitation light of different wavelengths on the physical time axis, realizes time-division multiplexing of excitation light of different wavelengths in millisecond-level time intervals, divides the imaging period into non-overlapping time windows, and sends a global trigger level to the optically isolated input terminal of the high-speed imaging device to force photosensitive reset, thereby aligning the physical timing of laser excitation with the global exposure window of the camera. The host computer computing and processing platform is used for system configuration, asynchronous data stream processing, and image storage.

2. The multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition according to claim 1, characterized in that: The time-division multiplexing strategy is accomplished through two sets of complementary interrupt service logic and hardware handshake logic.

3. The multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition according to claim 2, characterized in that: Inside the MCU, the first timer control parameters, timing constraints, and second timer control parameters corresponding to each laser control channel are configured, and a hardware handshake link is built between the MCU input / output interface GPIO and the opto-isolation circuit, connecting to the opto-isolation input terminal of the high-speed imaging device.

4. The multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition according to claim 3, characterized in that: A dual-timer cooperative control model based on a discrete-time state machine is built inside the MCU to implement interrupt service logic, as follows: Set the MCU system clock frequency to If the timer prescaler is PSC, then the minimum time granularity of the hardware counter is... satisfy: ； The MCU internally maintains a laser channel index variable k, k ∈{0, 1, ..., M-1}, where M is the total number of laser channels; the control model's working logic is as follows: (1) Excited state: Start the first timer TIM_ON and set its auto-reload register value to... At this time, the single-channel excitation duration Defined by the following formula: ； The MCU retrieves the light intensity mapping from the preset light intensity mapping table based on the current index variable k. Read the corresponding pulse width modulation value Write to the Comparator Register (CCR) k This results in the output of duty cycle D at the corresponding GPIO port. k satisfy: ； (2) Dead zone switching state_DeadTime: When the first timer overflows and triggers the first interrupt, the MCU immediately disables the first timer output and starts the second timer TIM_OFF, with a reload value of [value to be filled in]. Dead zone interval duration satisfy: ； (3) Looping logic: When the second timer overflows and triggers the second interrupt, the MCU executes the channel index update algorithm: ； Where mod represents the modulo operation. This is the total number of laser channels, determined by the MCU based on k. next Update the CCR parameters for the next cycle and restart the first timer to achieve hardware closed-loop control without software intervention.

5. The multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition according to claim 3, characterized in that: The lower-level timing control unit is equipped with hardware handshake logic based on general-purpose input / output interface (GPIO) and opto-isolation circuit. The GPIO pins of the MCU are connected to the external trigger input of the imaging device to execute the following microsecond-level timing actions: (1) Trigger Response: At the start of a single imaging cycle, the MCU generates a TTL synchronization trigger signal, i.e., pulse width The effective edge of this signal directly drives the imaging device that supports global shutter mode, controlling all pixels of the sensor array to start photoelectric integration at the same time, forming a common exposure window with an absolute time reference. (2) The MCU synchronously outputs the laser PWM modulation signal, i.e., the effective light emission pulse width. The system satisfies the temporal envelope constraint to ensure the complete energy of laser excitation. Completely enclosed within the camera's global exposure window [ , Within this range, lossless integration of photon energy is achieved, and the time-gating characteristics of the global shutter are utilized to effectively suppress ambient stray light noise outside the exposure window; among which, It is the start time of exposure. It is the end time of the exposure; The temporal envelope constraint inequality is as follows: ； in, Due to the inherent turn-on delay of the laser driver, This provides a timing safety margin for the system.

6. The multi-channel fluorescence imaging system based on time-series pulse excitation and asynchronous acquisition according to claim 1, characterized in that: The host computer computing platform is configured with an underlying asynchronous memory buffer architecture based on a producer-consumer concurrent model, used to decouple image acquisition from hard disk writing; the specific execution logic is as follows: (1) Physical memory pre-allocation and addressing topology: Define the pixel width of a single frame image as The height is The depth of a single pixel byte is The amount of image data per frame ; The host computer pre-allocates a contiguous memory space in the system's physical memory as a circular data buffer, setting its total number of data block nodes to [number missing]. Then pre-allocate the total memory size The starting address of this memory block is defined as ; The system initializes and maintains independent and atomic write pointers. With read pointer All initial values ​​are set to 0; (2) Zero-copy ultra-fast enqueue based on pointer offset: When the raw image data of the i-th frame triggers a data transmission interruption, the producer thread performs the following status check and absolute address lookup: First, determine the status of the idle node. If the condition is met... If the buffer has sufficient space, then the physical absolute target address of the current node in memory is calculated: ； Then, the system's underlying memory block shifting instruction is invoked to move the block of length to... The binary data is from the sensor's source address. Copy exactly to This avoids heap object allocation in high-level languages; after copying is complete, the producer obtains a high-precision system timestamp. The queue tuple containing pointers and time parameters ( Push the code into the non-blocking queue Q and update the write pointer: ； (3) Asynchronous consumption of matrix mapping demixing and pointer dequeueing: A standalone consumer thread listens to the non-blocking queue Q and extracts the queued tuples. ), and update the read pointer synchronously. To release the corresponding memory node; the consumer based on the extracted Pointers directly access the original image matrix And execute them sequentially: a) Color space conversion: Perform matrix mapping using interpolation algorithms. Reconstruct the Bayer array into an RGB matrix; b) Spectral unmixing: Constructing a linear mixture mathematical model based on prior spectral observations. Where Y is the current observed signal matrix. A represents the known endmember spectral matrix extracted by the system calibration, containing the autofluorescence background spectrum and the target probe spectrum; E represents the system noise term; the consumer backend solves for the feature abundance matrix C using a non-negative matrix factorization algorithm, removes the autofluorescence background interference signal, and reconstructs a clean multi-channel target image. ; c) Timing Encapsulation and Disk Writing: Dequeuing the timestamp Encoding is included in the image file header identifier, which will then be used to demix the target image. Asynchronous I / O writes to the hard disk are performed according to a preset compression algorithm.

Images