A flow cytometer laser delay method based on asymmetric dichotomy

Abstract

The application provides a flow cytometer laser delay method based on asymmetric dichotomy, and belongs to the technical field of cell analysis instruments.The method comprises the following steps: detecting preset standard microspheres by using a flow cytometer laser, and obtaining an average laser delay by using a sequential fitting method and asymmetric dichotomy; normally detecting a cell sample by using the flow cytometer, obtaining a current signal highest point position by using Gaussian filtering and dynamic baseline updating, obtaining a current cell multi-channel detection result by finding an actual signal position of the cell according to the average laser delay, and obtaining multi-channel detection results of each cell by cyclic detection, so as to complete flow cytometer laser delay.The application solves the accuracy problem of each laser delay of a multi-laser flow cytometer, thereby avoiding the occurrence of false signals and improving the accuracy of detection results.

Description

Technical Field

[0001] This invention belongs to the field of cell analysis instrument technology, and particularly relates to a laser delay method for flow cytometry based on the asymmetric dichotomy method. Background Technology

[0002] Flow cytometry (FCM) is an instrument for the rapid quantitative determination and analysis of cells or biological particles flowing at high speeds in a linear fashion. It primarily includes sample flow technology, cell counting and sorting techniques, and computer-aided data acquisition and analysis. Based on the theory of flow cytometry, flow cytometry is a culmination of knowledge from fluid mechanics, laser technology, electronic engineering, molecular immunology, cellular fluorescence chemistry, and computer science. Flow cytometry is an automated technique for analyzing and sorting cells or subcellular structures. Its characteristics include: high measurement speed, large sample size, and the ability to perform multi-parameter measurements—that is, to perform multiple measurements of the physical and biochemical properties of the same cell, with statistical validity.

[0003] With increasing demands in clinical and research settings, the requirements for flow cytometer detection channels are growing, leading to the development of multi-laser flow cytometers. Currently, most multi-laser flow cytometers employ non-collinear excitation, making the delay parameters between each laser crucial. Incorrect laser delays can result in signal errors and ultimately inaccurate detection results. Therefore, there is an urgent need for an algorithm that can automatically calculate laser delays to ensure the accuracy of the results. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, this invention provides a flow cytometer laser delay method based on asymmetric dichotomy, which solves the problems of low accuracy of laser delay detection results and high probability of erroneous signals in multi-laser flow cytometers.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a flow cytometry laser time-delay method based on the asymmetric dichotomy method, comprising the following steps: S1. Using flow cytometry laser, the preset standard microspheres are detected, and the average laser delay is obtained using the sequential fitting method and the asymmetric dichotomy method. S2. Using a flow cytometer, the cell sample is normally detected, and the position of the current highest signal point is obtained through Gaussian filtering and dynamic baseline updating. Based on the average laser delay, the actual signal position of the cell is found to obtain the multi-channel detection result of the current cell. Through cyclic detection, the multi-channel detection results of each cell are obtained, thus completing the flow cytometer laser delay.

[0006] The beneficial effects of this invention are as follows: This invention uses a combination of sequential adaptation and asymmetric bisection methods by pre-setting standard microspheres to quickly and accurately calculate laser delay, and greatly improves execution efficiency, meeting the operating conditions of embedded devices. Furthermore, based on the average laser delay and the currently identified signal, it performs further real-time calculations to ensure accurate identification of each signal and guarantee accurate results when detecting minute residual items.

[0007] Further, S1 includes the following steps: S101. Using flow cytometer laser, the preset standard microspheres are detected, and the peak value and area of ​​the preset standard microspheres are calculated using the sequential adaptation method to obtain a rough optimal laser delay value. S102. Based on the rough optimal laser delay value, the average laser delay is obtained by using the asymmetric bisection method, obtaining the search interval and performing asymmetric bisection iteration.

[0008] Furthermore, step S101 includes the following steps: S1011. The flow cytometer is initialized by initializing the candidate range and the interval of the laser delay to obtain the initialized flow cytometer. S1012. Using an initialized flow cytometer, standard microspheres of uniform size with fluorescence signals in each channel were measured to obtain measurement data. S1013. Based on the measurement data, using the sequential adaptation method and a fixed step size, the candidate range of laser delay is traversed. The peak value and area of ​​the preset standard microsphere are maximized in response to the calculation, and a rough optimal laser delay value is obtained.

[0009] Furthermore, step S1013 includes the following steps: A1. The laser delay is set based on the number of iterations and the laser delay interval; A2. Using a preset clustering algorithm, cluster the measurement data, calculate the peak value and area of ​​the preset standard microspheres for the current cycle number, and record the correspondence between the peak value and area of ​​the preset standard microspheres for the current cycle number and the laser delay. A3. Cluster the calibration signal by cyclic execution. In response to the current cycle number, the preset standard microsphere has the largest peak value and area. The laser delay with the largest peak value and area of ​​the preset standard microsphere fluorescence signal is taken as the coarse optimal laser delay value.

[0010] The beneficial effects of the above-mentioned further solution are as follows: The present invention uses the sequential adaptation method to take the laser delay with the maximum peak value and area of ​​the preset standard microsphere fluorescence signal as the coarse optimal laser delay value, thereby realizing the rapid and robust positioning of the coarse optimal value of the laser delay within the entire candidate range. This provides a reliable initial center for the subsequent fine search of the asymmetric bisection method, avoids the fine search from getting trapped in local optima, and at the same time narrows the subsequent search range and reduces the amount of computation.

[0011] Furthermore, step S102 includes the following steps: S1021. Based on the coarse optimal laser delay value, take the coarse optimal laser delay value as the center and use the initial search radius of the asymmetric bisection method. By subtracting the coarse optimal laser delay value from the initial search radius, the left boundary of the search interval is obtained. By adding the coarse optimal laser delay value to the initial search radius, the right boundary of the search interval is obtained. S1022. Obtain the search interval based on the left and right boundaries of the search interval; S1023. Based on the search interval, obtain the search range of the asymmetric binary search algorithm, and gradually narrow the search interval through asymmetric binary search iteration to obtain the average laser delay value.

[0012] Furthermore, step S1023 includes the following steps: B1. Based on the search interval, subtract the right boundary of the search interval from the left boundary of the search interval to obtain the search range of the asymmetric binary search algorithm. B2. Taking the left boundary of the search interval as the left-end laser delay, calculate the peak value and area of ​​the left-end laser delay; and taking the right boundary of the search interval as the right-end laser delay, calculate the peak value and area of ​​the right-end laser delay. B3. Compare the peak value and area of ​​the laser delay at the left end with the peak value and area of ​​the laser delay at the right end to obtain the comparison results; B4. In response to the comparison result that the peak value and area of ​​the laser delay at the left end are greater than the peak value and area of ​​the laser delay at the right end, the left boundary of the search interval is subtracted by half of the search range of the asymmetric binary search algorithm to obtain the updated left boundary of the search interval. The right boundary of the search interval is subtracted by the search range of the asymmetric binary search algorithm to obtain the updated right boundary of the search interval. The search range of the asymmetric binary search algorithm is then updated to half of the search range of the asymmetric binary search algorithm. B5. In response to the comparison result that the peak value and area of ​​the laser delay at the left end are less than the peak value and area of ​​the laser delay at the right end, the left boundary of the search interval is added to the search range of the asymmetric binary search algorithm to obtain the updated left boundary of the search interval. The right boundary of the search interval is added to half of the search range of the asymmetric binary search algorithm to obtain the updated right boundary of the search interval. The search range of the asymmetric binary search algorithm is then updated to half of the search range of the asymmetric binary search algorithm. B6. Determine whether the updated asymmetric binary search range is equal to one. If not, return to B2 based on the updated left boundary of the search interval, the updated right boundary of the search interval, and the updated asymmetric binary search range. If so, the loop ends, and the laser delay corresponding to the larger of the peak and area values ​​of the left-end laser delay and the peak and area values ​​of the right-end laser delay calculated in the last time is taken as the average laser delay.

[0013] Furthermore, the expression for the asymmetric binary iteration is as follows: ; in, This indicates the peak value and area of ​​the laser delay at the left end. This indicates the peak value and area of ​​the laser delay at the right end. Indicates the left boundary of the updated search interval. Indicates the left boundary of the search interval. This indicates the search range of the asymmetric binary search algorithm. Indicates the right boundary of the updated search interval. This indicates the right boundary of the search interval.

[0014] The beneficial effects of the above-mentioned further solution are as follows: By taking the coarse optimal laser delay value obtained by the sequential adaptation method as the center, the present invention uses the asymmetric bisection method to obtain the average laser delay, thereby achieving rapid convergence of the search interval with fewer iterations. While ensuring the accuracy of laser delay calculation, it greatly improves the computational efficiency, enabling it to meet the requirements of real-time operation on embedded devices.

[0015] Furthermore, S2 includes the following steps: S201. Using a flow cytometer, the cell sample is subjected to normal testing to obtain the original sampling signal; S202. Perform Gaussian filtering on the original sampled signal to obtain the filtered sampled signal; S203. Using an analog-to-digital converter, sample the filtered sampling signal and perform dynamic baseline updates to obtain the position of the highest point of the current sampling signal. S204. Based on the average laser delay and the position of the highest point of the current sampling signal, the multi-channel detection result of the current cell is obtained by finding the actual signal position of the current cell sample. S205. By cyclically detecting the actual signal position of each cell sample, the multi-channel detection results of each cell are obtained, and the laser delay of the flow cytometer is completed.

[0016] Furthermore, S203 specifically includes: S2031. Using an analog-to-digital converter, sample the filtered sampling signal and determine whether the sampled value of the analog-to-digital converter is lower than a preset threshold. If yes, proceed to step S2032; otherwise, proceed to step S2033. S2032. Add the current sampled value of the analog-to-digital converter to the baseline queue. In response to the baseline queue sampled points exceeding the preset number, calculate the average value of the preset number of sampled points as the baseline value, and delete the first sampled point from the baseline queue. Using the first-in-first-out principle, keep the number of sampled points in the baseline queue constant at the preset number. S2033. Determine if a calculated baseline value already exists. If so, subtract the baseline value from the sampled value of the analog-to-digital converter and place it in the signal queue. Otherwise, return to S2031. S2034. Determine if the number in the signal queue is less than five. If yes, return to S2031. If no, determine if the signal queue is on an upward trend at the five points. If the sample value at the fifth point is greater than the sample value at the third point and the sample value at the third point is greater than the sample value at the first point, then a valid signal queue is obtained. S2035. In response to the end of the filtered sampling signal, based on all sampling points in the signal queue, the maximum value is calculated by accumulating all sampling points as the signal area and recording the number of signal queues as the signal width. The maximum value is recorded as the height of the filtered sampling signal, and the position of the highest point of the current sampling signal is obtained by recording the position of the highest point.

[0017] The beneficial effects of the above-mentioned further solution are as follows: This invention performs Gaussian filtering for noise reduction, dynamic baseline updating, and effective signal identification and extraction on the original sampling signal obtained from cell sample detection. By real-time particle-by-particle delay compensation based on the average laser delay and the filtered sampling signal of the cell sample, the multi-channel detection results of each cell are obtained. This achieves accurate identification and matching of the laser channel signals of each cell, avoids signal misalignment caused by inaccurate laser delay, improves the accuracy of multi-channel detection results, and ensures the reliability of output results in scenarios such as the detection of minute residues. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method of the present invention.

[0019] Figure 2 This is a schematic diagram of the asymmetric bisection method in this embodiment. Detailed Implementation

[0020] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0021] Before describing this embodiment, the following terms will be explained: K-means: The K-means algorithm; DSP: Digital Signal Processing; FPGA: Field Programmable Gate Array.

[0022] Example like Figure 1 As shown, this invention provides a flow cytometry laser time-delay method based on the asymmetric dichotomy method, the implementation of which is as follows: S1. Using flow cytometry laser, the preset standard microspheres are detected, and the average laser delay is obtained using the sequential fitting method and the asymmetric dichotomy method. The specific steps are as follows: S101. Using flow cytometry laser, the preset standard microspheres are detected, and the peak value and area of ​​the preset standard microspheres are maximized by calculating the response using the sequential adaptation method, thus obtaining a rough optimal laser delay value. The specific steps are as follows: S1011. The flow cytometer is initialized by initializing the candidate range and the interval of the laser delay to obtain the initialized flow cytometer. S1012. Using an initialized flow cytometer, standard microspheres of uniform size with fluorescence signals in each channel were measured to obtain measurement data.

[0023] In this embodiment, to calculate the average laser delay of the flow cytometer, based on the flow cytometer, a preset standard microsphere of uniform size with fluorescence signals in each channel is used for calibration before the instrument detects the actual cell sample to obtain the average laser delay. Specifically: The flow cytometer was initialized with a laser delay range of 10µs to 1000µs, with an interval of 10µs, to obtain the initialized flow cytometer. Using an initialized flow cytometer, standard microspheres of uniform size with fluorescence signals in each channel were measured to obtain measurement data.

[0024] S1013. Based on the measurement data, using the sequential adaptation method with a fixed step size, the candidate range of laser delay is traversed. The calculation results in the maximum peak value and area of ​​the preset standard microsphere, yielding a roughly optimal laser delay value. The specific steps are as follows: A1. The laser delay is set based on the number of iterations and the laser delay interval; A2. Using a preset clustering algorithm, cluster the measurement data, calculate the peak value and area of ​​the preset standard microspheres for the current cycle number, and record the correspondence between the peak value and area of ​​the preset standard microspheres for the current cycle number and the laser delay. A3. Cluster the calibration signal by cyclic execution. In response to the current cycle number, the preset standard microsphere has the largest peak value and area. The laser delay with the largest peak value and area of ​​the preset standard microsphere fluorescence signal is taken as the coarse optimal laser delay value.

[0025] In this embodiment, the sequential adaptation method is used to traverse the candidate range of laser delay based on a fixed step size, and find a coarse delay value that maximizes the peak value and area. M The peak value and area mentioned refer to the peak value and area of ​​the fluorescence signal, specifically: Set laser delay N , ,in, i Indicates the number of loop iterations; Using a preset clustering algorithm, the measurement data is clustered to calculate the peak value and area of ​​the preset standard microspheres for the current cycle number. Specifically, in this embodiment, the K-means algorithm is used to cluster each standard microsphere event in the measurement data according to its fluorescence signal characteristics. After removing outliers, the target standard microsphere population is obtained. For each microsphere event in the population, the maximum amplitude value of its pulse signal at each sampling point is taken as the peak value of the event, and the sum of the amplitude values ​​of its pulse signal at each sampling point is taken as the area of ​​the event. Then, the average value of the peak value and area of ​​all events in the population is taken to obtain the peak value and area of ​​the preset standard microspheres for the current cycle number. The calculation expression is as follows: ; ; in, Indicates the peak value of microsphere events. This represents the area of ​​the microsphere event. This represents the average of all events within a group. This indicates that the maximum amplitude of the event sampling point is taken. The first pulse signal representing a certain microsphere event. k Amplitude at each sampling point, This represents the summation of the amplitudes over the sampled points of the event; It also records the correspondence between the peak value and area of ​​the preset standard microspheres and the laser delay for the current cycle number; By clustering the measurement data through iterative execution, and maximizing the peak value and area of ​​the preset standard microspheres in response to the current iteration, a rough optimal laser delay value is obtained. M ,in , This represents the laser delay at which the fluorescence signal peak and area are at their maximum.

[0026] S102. Based on the rough optimal laser delay value, the average laser delay is obtained by using the asymmetric bisection method, which involves acquiring the search interval and performing asymmetric bisection iterations. The specific steps are as follows: S1021. Based on the coarse optimal laser delay value, take the coarse optimal laser delay value as the center and use the initial search radius of the asymmetric bisection method. By subtracting the coarse optimal laser delay value from the initial search radius, the left boundary of the search interval is obtained. By adding the coarse optimal laser delay value to the initial search radius, the right boundary of the search interval is obtained. S1022. Based on the left and right boundaries of the search interval, the search interval is obtained.

[0027] In this embodiment, as Figure 2 As shown, asymmetric bisection method is used for fine positioning to obtain a rough optimal laser delay value. M As the center, and using the initial search radius of the asymmetric bisection method. X Construct the search interval Obtain the search range of the asymmetric binary search algorithm. D The interval is gradually narrowed through asymmetric binary search iteration, and finally the accurate average laser delay value is output. Among them, the initial search radius of the asymmetric bisection method X Used to roughly optimize laser delay value M Based on this, the left and right boundaries of the fine search are determined, and the value is 1 to 2 times the step size of the sequential adaptation method; L This represents the left boundary of the current search interval in the asymmetric bisection method, i.e., the end with the smaller delay. The initial value is... ; R This represents the right boundary of the current search interval in the asymmetric bisection method, i.e., the end with the larger delay. The initial value is... ; D Let represent the search range of the asymmetric binary search algorithm, and .

[0028] S1023. Based on the search interval, obtain the search range of the asymmetric binary search algorithm, and gradually narrow the search interval through asymmetric binary search iteration to obtain the average laser delay value. The specific steps are as follows: B1. Based on the search interval, subtract the right boundary of the search interval from the left boundary of the search interval to obtain the search range of the asymmetric binary search algorithm. B2. Taking the left boundary of the search interval as the left-end laser delay, calculate the peak value and area of ​​the left-end laser delay; and taking the right boundary of the search interval as the right-end laser delay, calculate the peak value and area of ​​the right-end laser delay. B3. Compare the peak value and area of ​​the laser delay at the left end with the peak value and area of ​​the laser delay at the right end to obtain the comparison results; B4. In response to the comparison result that the peak value and area of ​​the laser delay at the left end are greater than the peak value and area of ​​the laser delay at the right end, the left boundary of the search interval is subtracted by half of the search range of the asymmetric binary search algorithm to obtain the updated left boundary of the search interval. The right boundary of the search interval is subtracted by the search range of the asymmetric binary search algorithm to obtain the updated right boundary of the search interval. The search range of the asymmetric binary search algorithm is then updated to half of the search range of the asymmetric binary search algorithm. B5. In response to the comparison result that the peak value and area of ​​the laser delay at the left end are less than the peak value and area of ​​the laser delay at the right end, the left boundary of the search interval is added to the search range of the asymmetric binary search algorithm to obtain the updated left boundary of the search interval. The right boundary of the search interval is added to half of the search range of the asymmetric binary search algorithm to obtain the updated right boundary of the search interval. The search range of the asymmetric binary search algorithm is then updated to half of the search range of the asymmetric binary search algorithm. B6. Determine whether the updated asymmetric binary search range is equal to one. If not, return to B2 based on the updated left boundary of the search interval, the updated right boundary of the search interval, and the updated asymmetric binary search range. If so, the loop ends, and the laser delay corresponding to the larger of the peak and area values ​​of the left-end laser delay and the peak and area values ​​of the right-end laser delay calculated in the last time is taken as the average laser delay.

[0029] In this embodiment, the asymmetric binary search iteration gradually narrows the interval specifically as follows: The left boundary of the search interval is used as the left laser delay, and the peak value and area of ​​the left laser delay are calculated. Similarly, the right boundary of the search interval is used as the right laser delay, and the peak value and area of ​​the right laser delay are calculated. Compare the peak value and area of ​​the laser delay at the left end with the peak value and area of ​​the laser delay at the right end; Since the peak value and area of ​​the laser delay at the left end are greater than those at the right end, half of the search range of the asymmetric binary search algorithm is subtracted from the left boundary of the search interval. The updated left boundary of the search interval is obtained. Subtract the search range of the asymmetric binary search algorithm from the left boundary of the search interval. The updated right boundary of the search interval is obtained. The search range of the asymmetric binary search algorithm is updated to half of the search range of the asymmetric binary search algorithm. ; Since the peak value and area of ​​the laser delay at the left end are smaller than those at the right end, the search range of the asymmetric binary search algorithm is added to the left boundary of the search interval. The updated left boundary of the search interval is obtained. Add half the search range of the asymmetric binary search algorithm to the right boundary of the search interval. The updated right boundary of the search interval is obtained. The search range of the asymmetric binary search algorithm is updated to half of the search range of the asymmetric binary search algorithm. ; Determine the search range of the updated asymmetric binary search algorithm. If the value is 1, and if not, recalculate the peak value and area based on the updated left boundary of the search interval, the updated right boundary of the search interval, and the updated search range of the asymmetric binary search algorithm, and compare them; if yes, end the loop and take the laser delay at the left or right end of the search interval corresponding to the larger value of the last calculated peak value and area value as the average laser delay.

[0030] S2. Using a flow cytometer, the cell sample is normally detected. Gaussian filtering and dynamic baseline updates are used to obtain the current signal peak position. Based on the average laser delay, the actual signal position of the cell is located to obtain the multi-channel detection result of the current cell. Through cyclic detection, the multi-channel detection results of each cell are obtained, completing the flow cytometer laser delay. The specific steps are as follows: S201. Using a flow cytometer, the cell sample is subjected to normal testing to obtain the original sampling signal; S202. Perform Gaussian filtering on the original sampled signal to obtain the filtered sampled signal; S203. Using an analog-to-digital converter, sample the filtered signal and perform dynamic baseline updates to obtain the position of the highest point of the current sampled signal. The specific steps are as follows: S2031. Using an analog-to-digital converter, sample the filtered signal and determine whether the sampled value is lower than a preset threshold. If yes, proceed to step S2032; otherwise, proceed to step S2033. S2032. Add the current sampled value of the analog-to-digital converter to the baseline queue. In response to the baseline queue sampled points exceeding the preset number, calculate the average value of the preset number of sampled points as the baseline value, and delete the first sampled point from the baseline queue. Using the first-in-first-out principle, keep the number of sampled points in the baseline queue constant at the preset number. S2033. Determine if a calculated baseline value already exists. If so, subtract the baseline value from the sampled value of the analog-to-digital converter and place it in the signal queue. Otherwise, return to S2031. S2034. Determine if the number in the signal queue is less than five. If yes, return to S2031. If no, determine if the signal queue is on an upward trend at the five points. If the sample value at the fifth point is greater than the sample value at the third point and the sample value at the third point is greater than the sample value at the first point, then a valid signal queue is obtained. S2035. In response to the end of the filtered sampling signal, based on all sampling points in the signal queue, the maximum value is calculated by accumulating all sampling points as the signal area and recording the number of signal queues as the signal width. The maximum value is recorded as the height of the filtered sampling signal, and the position of the highest point of the current sampling signal is obtained by recording the position of the highest point.

[0031] In this embodiment, the cell sample is normally detected using a flow cytometer to obtain the original sampling signal; The original sampled signal is subjected to Gaussian filtering, and then the baseline is calculated based on the filtered signal. The useful signal is then calculated from the baseline. Based on the calculated average delay, the real signal is found near this delay position to obtain the multi-channel detection result of the current cell. The original sampled signal is subjected to Gaussian filtering to obtain the filtered sampled signal; Using an analog-to-digital converter (ADC), the filtered sampling signal is sampled. It is determined whether the sampled value of the ADC is lower than a preset threshold. If so, the current sampled value of the ADC is added to the baseline queue. When the number of sampling points in the baseline queue exceeds a preset number C, the average value of these preset number C points is calculated as the baseline value, and the first sampling point is deleted from the baseline queue. Using the first-in-first-out principle, the number of points in the baseline queue is kept constant at the preset number C. Otherwise, determine whether a baseline value has already been calculated. If so, subtract the baseline value from the sampled value of the analog-to-digital converter and put it into the signal queue. Otherwise, use the analog-to-digital converter to calculate the baseline again. Determine if the number in the signal queue is less than five. If so, return to use the analog-to-digital converter to sample the filtered signal. If not, determine if the signal queue shows an upward trend at the five points. If the sample value at the fifth point is greater than the sample value at the third point, and the sample value at the third point is greater than the sample value at the first point, then a valid signal is considered to have been found, and the signal flag is set to 1. The sampled value is the amplitude of the current sampled point in the signal queue output by the analog-to-digital converter after Gaussian filtering and subtraction of the dynamic baseline; When a sampled value is less than a preset threshold, the signal terminates. Based on all sampled points in the signal queue, the maximum amplitude of the sampled point is calculated and recorded as the signal height, and the location of the highest point is also recorded. P The area of ​​the signal is summed up by all the sampling points, and the width is recorded by the number of signal queues. This gives the position of the highest point of the current sampled signal. The preset threshold is specifically a noise threshold for determining the start and end of a signal, and the formula for calculating the maximum amplitude of the sampling point is: ,in, Indicates the height of the signal. Indicates the first [number] in the signal queue j The amplitude of each sampling point after subtracting the baseline.

[0032] S204. Based on the average laser delay and the position of the highest point of the current sampling signal, the multi-channel detection result of the current cell is obtained by finding the actual signal position of the current cell sample. S205. By cyclically detecting the actual signal position of each cell sample, the multi-channel detection results of each cell are obtained, and the laser delay of the flow cytometer is completed.

[0033] In this embodiment, based on the average laser delay, with the current highest signal point P as the benchmark, within the range of [P - average laser delay - search tolerance, P - average laser delay + search tolerance], the sampling point with the largest peak value is searched in the corresponding other laser channel sampling data. The sampling point with the largest peak value is taken as the actual signal position of the cell corresponding to the laser channel, and the peak value and area of ​​the actual signal position are extracted. The search tolerance is a pre-set search window radius in units of sampling points, used to compensate for calculation errors in laser delay and signal position shifts caused by factors such as fluid jitter. Its value can be empirically determined based on the maximum time jitter of the system, and should be as small as possible while still covering the actual signal position to avoid mismatching adjacent signals. The purpose of this range is to define a search interval near the predicted signal position, thereby robustly locating the true signal peak value of each laser channel corresponding to the cell. The peak value and area of ​​the first laser channel signal are combined with the peak values ​​and areas of the other laser channels in channel order to output the complete multi-channel detection result of the current cell. This completes the real-time calculation and signal matching of the flow cytometer laser delay for one cell. The flow cytometer is then used to detect the cell sample normally to continue processing the next cell until all cells are processed, and the multi-channel detection results of each cell are obtained, thus completing the flow cytometer laser delay.

[0034] In this embodiment, the hardware platform uses a TMS320C6678 DSP and is programmed using C language. First, after power-on, the DSP initializes the relevant peripherals and interfaces. After the system starts detection, the FPGA collects data from each channel and performs first-order filtering. The FPGA and DSP communicate to transmit the data to the DSP for signal processing. The data is clustered using the microsphere clustering function beads_clustering() and the Kmeans algorithm is used to cluster the data. Then, the average value of the population peak and area is calculated. The microsphere clustering function is used to perform cluster analysis on standard microsphere data during the laser delay calibration stage. Its core function is to classify the multi-event fluorescence signals acquired by the FPGA according to their features, thereby extracting the target microsphere population from the mixed signals. The function usually uses unsupervised clustering methods such as K-means to model the features of the fluorescence pulse signal of each event, and combines the peak value (max sample value) and area (sample accumulation sum) as the main criteria. After removing outliers, the microsphere events of the same category are statistically averaged, and finally the average peak value and average area of ​​the population are output. These are used for subsequent delay evaluation index calculation, as well as basic signal purification and feature extraction in the entire delay calibration link.

[0035] The laser delay is roughly calculated using the sequential adaptation search function sequential_daptation() and the sequential adaptation algorithm. The sequential adaptation search function is used to implement a coarse search process for laser delay parameters. Essentially, it performs a sequential traversal within a preset delay range based on a fixed step size, and evaluates the signal quality corresponding to each delay point in conjunction with cluster analysis results. Specifically, under each candidate laser delay setting, the function calls the clustering module to obtain the characteristics of the standard microsphere population and calculates the corresponding peak and area comprehensive index. By traversing all candidate delay values ​​and recording the position of the maximum value of the evaluation function, a coarsely optimal laser delay value M is determined. This value serves as the initial center point for the subsequent fine optimization algorithm, ensuring that the subsequent search is carried out within a smaller effective area, thereby significantly reducing computational complexity and improving convergence efficiency.

[0036] The laser delay is calculated using the asymmetric bisection optimization function asymmetric_bisection(). The asymmetric binary optimization function is used to perform coarse-time optimization. M Based on this, an asymmetric binary search is performed to obtain a high-precision average laser delay. Its core idea is to... M An initial search interval is constructed around the center, and the optimal solution is gradually approached by dynamically adjusting the asymmetric shrinkage strategy of the left and right boundaries. In each iteration, the signal peak value and area under the laser delay corresponding to the left and right endpoints of the interval are calculated respectively, and the search interval is shrunk to which side based on the comprehensive evaluation results. At the same time, the search step size (search range D) is adaptively adjusted to accelerate the convergence while maintaining the stability of the search process. Finally, when the step size converges to the minimum threshold, the final average laser delay is output, realizing high-precision calibration of the system time offset.

[0037] The original acquired signal is processed by Gaussian filtering using the Gaussian_filter() function. The Gaussian filtering function is used to perform Gaussian filtering on the raw ADC signal acquired by the FPGA. Its main function is to suppress high-frequency noise and smooth the signal waveform to improve the stability of subsequent peak detection and baseline estimation. The function performs convolution operation on the input discrete signal by constructing a Gaussian kernel, so that the signal presents a continuous and smooth change trend on the time axis, thereby reducing instantaneous abnormal fluctuations caused by electronic noise, fluid disturbance or sampling jitter, and providing more reliable input data for subsequent dynamic baseline update and peak extraction.

[0038] The dynamic baseline update function baseline_update() is used to perform real-time baseline calculation on the Gaussian filtered data. The dynamic baseline update function is used to implement a dynamic baseline estimation and update mechanism. Its core purpose is to eliminate the influence of system drift and background noise on signal amplitude, thereby improving the accuracy of effective signal detection. The function maintains a fixed-length FIFO sampling queue, continuously adding background sampling points below a preset threshold to the queue and calculating their moving average as the current baseline level. Simultaneously, the queue is continuously updated using a first-in-first-out (FIFO) mechanism, enabling the baseline to track the slow-changing characteristics of the system in real time. During signal detection, the baseline value is subtracted from the real-time sampled value to obtain the net signal, thus significantly improving the robustness of peak identification.

[0039] The peak detection function peak_detect() is used to detect useful data signals in the original signal and extract information such as the peak value and area of ​​the signal. The peak detection function is used to identify individual cell events and extract their key feature parameters from a continuous sampling stream. It is the core event detection module of the entire signal analysis chain. The function first judges the local upward trend based on the signal after baseline removal, identifies potential effective signal segments through the five-point trend rule, and includes them in the signal queue for accumulation. When the signal meets the termination condition, a complete statistical analysis is performed on the event, including calculating the peak value (maximum sampling amplitude), area (cumulative sampling value), and width (number of sampling points). At the same time, the time position corresponding to the peak value is recorded, thereby realizing the accurate separation of the complete pulse event of a single cell from the continuous data stream.

[0040] The real-time delay compensation function realtime_delay() is used to calculate the laser delay in real time, and the calculated results are applied to the detection of other channels to output all correct detection results. The real-time delay compensation function is used to perform multi-channel laser delay compensation and signal alignment during real-time detection. Its core function is to correct and match the signals of different laser channels on the time axis based on the calibrated average laser delay. The function uses the peak position detected by the main channel. P Based on the baseline, combined with delay parameters in P A search window is constructed nearby, and the maximum response peak within the corresponding time window is searched in other channel data to determine the true response position of the cell in each channel. Finally, the peak and area information of all channels are combined and output according to a unified time reference to achieve time synchronization and signal consistency reconstruction among multiple laser channels.

Claims

1. A flow cytometry laser time-delay method based on asymmetric dichotomy, characterized in that, Includes the following steps: S1. Using flow cytometry laser, the preset standard microspheres are detected, and the average laser delay is obtained using the sequential fitting method and the asymmetric dichotomy method. S2. Using a flow cytometer, the cell sample is normally detected, and the position of the current highest signal point is obtained through Gaussian filtering and dynamic baseline updating. Based on the average laser delay, the actual signal position of the cell is found to obtain the multi-channel detection result of the current cell. Through cyclic detection, the multi-channel detection results of each cell are obtained, thus completing the flow cytometer laser delay.

2. The flow cytometry laser time-delay method based on asymmetric dichotomy as described in claim 1, characterized in that, S1 includes the following steps: S101. Using flow cytometer laser, the preset standard microspheres are detected, and the peak value and area of ​​the preset standard microspheres are calculated using the sequential adaptation method to obtain a rough optimal laser delay value. S102. Based on the rough optimal laser delay value, the average laser delay is obtained by using the asymmetric bisection method, obtaining the search interval and performing asymmetric bisection iteration.

3. The flow cytometry laser time-delay method based on asymmetric dichotomy as described in claim 2, characterized in that, S101 includes the following steps: S1011. The flow cytometer is initialized by initializing the candidate range and the interval of the laser delay to obtain the initialized flow cytometer. S1012. Using an initialized flow cytometer, standard microspheres of uniform size with fluorescence signals in each channel were measured to obtain measurement data. S1013. Based on the measurement data, using the sequential adaptation method and a fixed step size, the candidate range of laser delay is traversed. The peak value and area of ​​the preset standard microsphere are maximized in response to the calculation, and a rough optimal laser delay value is obtained.

4. The flow cytometry laser time-delay method based on asymmetric dichotomy as described in claim 3, characterized in that, S1013 includes the following steps: A1. The laser delay is set based on the number of iterations and the laser delay interval; A2. Using a preset clustering algorithm, cluster the measurement data, calculate the peak value and area of ​​the preset standard microspheres for the current cycle number, and record the correspondence between the peak value and area of ​​the preset standard microspheres for the current cycle number and the laser delay. A3. Cluster the calibration signal by cyclic execution. In response to the current cycle number, the preset standard microsphere has the largest peak value and area. The laser delay with the largest peak value and area of ​​the preset standard microsphere fluorescence signal is taken as the coarse optimal laser delay value.

5. The flow cytometry laser time-delay method based on asymmetric dichotomy according to claim 2, characterized in that, S102 includes the following steps: S1021. Based on the coarse optimal laser delay value, take the coarse optimal laser delay value as the center and use the initial search radius of the asymmetric bisection method. By subtracting the coarse optimal laser delay value from the initial search radius, the left boundary of the search interval is obtained. By adding the coarse optimal laser delay value to the initial search radius, the right boundary of the search interval is obtained. S1022. Obtain the search interval based on the left and right boundaries of the search interval; S1023. Based on the search interval, obtain the search range of the asymmetric binary search algorithm, and gradually narrow the search interval through asymmetric binary search iteration to obtain the average laser delay value.

6. The flow cytometry laser time-delay method based on asymmetric dichotomy according to claim 5, characterized in that, S1023 includes the following steps: B1. Based on the search interval, subtract the right boundary of the search interval from the left boundary of the search interval to obtain the search range of the asymmetric binary search algorithm. B2. Taking the left boundary of the search interval as the left-end laser delay, calculate the peak value and area of ​​the left-end laser delay; and taking the right boundary of the search interval as the right-end laser delay, calculate the peak value and area of ​​the right-end laser delay. B3. Compare the peak value and area of ​​the laser delay at the left end with the peak value and area of ​​the laser delay at the right end to obtain the comparison results; B4. In response to the comparison result that the peak value and area of ​​the laser delay at the left end are greater than the peak value and area of ​​the laser delay at the right end, the left boundary of the search interval is subtracted by half of the search range of the asymmetric binary search algorithm to obtain the updated left boundary of the search interval. The right boundary of the search interval is subtracted by the search range of the asymmetric binary search algorithm to obtain the updated right boundary of the search interval. The search range of the asymmetric binary search algorithm is then updated to half of the search range of the asymmetric binary search algorithm. B5. In response to the comparison result that the peak value and area of ​​the laser delay at the left end are less than the peak value and area of ​​the laser delay at the right end, the left boundary of the search interval is added to the search range of the asymmetric binary search algorithm to obtain the updated left boundary of the search interval. The right boundary of the search interval is added to half of the search range of the asymmetric binary search algorithm to obtain the updated right boundary of the search interval. The search range of the asymmetric binary search algorithm is then updated to half of the search range of the asymmetric binary search algorithm. B6. Determine whether the updated asymmetric binary search range is equal to one. If not, return to B2 based on the updated left boundary of the search interval, the updated right boundary of the search interval, and the updated asymmetric binary search range. If so, the loop ends, and the laser delay corresponding to the larger of the peak and area values ​​of the left-end laser delay and the peak and area values ​​of the right-end laser delay calculated in the last time is taken as the average laser delay.

7. The flow cytometry laser time-delay method based on asymmetric dichotomy as described in claim 5, characterized in that, The expression for the asymmetric binary iteration is as follows: ； in, This indicates the peak value and area of ​​the laser delay at the left end. This indicates the peak value and area of ​​the laser delay at the right end. Indicates the left boundary of the updated search interval. Indicates the left boundary of the search interval. This indicates the search range of the asymmetric binary search algorithm. Indicates the right boundary of the updated search interval. This indicates the right boundary of the search interval.

8. The flow cytometry laser time-delay method based on asymmetric dichotomy according to claim 1, characterized in that, S2 includes the following steps: S201. Using a flow cytometer, the cell sample is subjected to normal testing to obtain the original sampling signal; S202. Perform Gaussian filtering on the original sampled signal to obtain the filtered sampled signal; S203. Using an analog-to-digital converter, sample the filtered sampling signal and perform dynamic baseline updates to obtain the position of the highest point of the current sampling signal. S204. Based on the average laser delay and the position of the highest point of the current sampling signal, the multi-channel detection result of the current cell is obtained by finding the actual signal position of the current cell sample. S205. By cyclically detecting the actual signal position of each cell sample, the multi-channel detection results of each cell are obtained, and the laser delay of the flow cytometer is completed.

9. The flow cytometry laser time-delay method based on asymmetric dichotomy as described in claim 8, characterized in that, Specifically, S203 is: S2031. Using an analog-to-digital converter, sample the filtered sampling signal and determine whether the sampled value of the analog-to-digital converter is lower than a preset threshold. If yes, proceed to step S2032; otherwise, proceed to step S2033. S2032. Add the current sampled value of the analog-to-digital converter to the baseline queue. In response to the baseline queue sampled points exceeding the preset number, calculate the average value of the preset number of sampled points as the baseline value, and delete the first sampled point from the baseline queue. Using the first-in-first-out principle, keep the number of sampled points in the baseline queue constant at the preset number. S2033. Determine if a calculated baseline value already exists. If so, subtract the baseline value from the sampled value of the analog-to-digital converter and place it in the signal queue. Otherwise, return to S2031. S2034. Determine if the number in the signal queue is less than five. If yes, return to S2031. If no, determine if the signal queue is on an upward trend at the five points. If the sample value at the fifth point is greater than the sample value at the third point and the sample value at the third point is greater than the sample value at the first point, then a valid signal queue is obtained. S2035. In response to the end of the filtered sampling signal, based on all sampling points in the signal queue, the maximum value is calculated by accumulating all sampling points as the signal area and recording the number of signal queues as the signal width. The maximum value is recorded as the height of the filtered sampling signal, and the position of the highest point of the current sampling signal is obtained by recording the position of the highest point.

Images