Flow cytometry sperm integrity automatic classification method based on machine learning

Abstract

The invention relates to the technical field of medical diagnosis, in particular to a flow cytometry sperm integrity automatic classification method based on machine learning, which comprises the following steps: collecting cell characteristics of sperm samples by a flow cytometry, and drawing two-dimensional scatter diagrams of the cell characteristics on different fluorescence channels; normalizing the cell data points in the scatter diagram to obtain correction coordinate data corresponding to the cell data points; analyzing the scatter diagram by using the joint distribution diagram, and determining the number of cell populations and the initial center-of-mass position coordinates; preliminarily dividing cell data points in the scatter diagram to obtain initial input data; clustering the cell data points in the scatter diagram to obtain a clustering result; according to the method, a two-dimensional point diagram is drawn on the basis of a clustering result, cell clusters of different types are delineated by using different polygon doors, automatic classification of the cell clusters is completed, the sperm DNA fragmentation rate (DFI) and the high DNA colorability index (HDS) can be automatically calculated, and the accuracy and efficiency of sperm quality evaluation are improved.

Description

Technical Field

[0001] The present invention relates to the field of medical diagnosis technology, and in particular to a method for automatic sperm integrity classification using a flow cytometer based on machine learning. Background Art

[0002] Sperm DNA fragmentation is an important indicator of the integrity or damage of sperm genetic material. It is widely believed to play an important role in evaluating semen quality and is closely related to infertility, recurrent miscarriage and low success rate of assisted pregnancy. Sperm DNA fragmentation rate (DFI) is an important parameter for measuring sperm quality and predicting fertility.

[0003] Flow cytometer is an instrument that can quickly and quantitatively analyze single cells or tiny particles, and plays an important role in the detection of sperm nuclear integrity. By passing the sperm suspension through a laser beam, the flow cytometer can measure the light scattering and fluorescence properties of each sperm cell. These data can reflect the integrity of the sperm nucleus, including DNA fragmentation, chromatin structure, and other indicators related to sperm quality. Flow cytometer provides a rapid, high-throughput and quantitative detection method, which greatly improves the efficiency and accuracy of sperm quality assessment and is of great significance in the diagnosis and treatment of infertility.

[0004] Although flow cytometry can generate a large amount of cell data, the analysis of this data has traditionally relied on manual operations. Researchers usually need to classify different cell populations on cell scatter plots based on experience and intuition. Manual analysis methods have the problems of high subjectivity, low efficiency, and lack of consistency. Summary of the invention

[0005] The purpose of the present invention is to provide a flow cytometer sperm integrity automatic classification method based on machine learning, aiming to assist researchers in accurately and efficiently dividing cell populations in sperm cell samples during flow cytometry sperm nuclear integrity detection.

[0006] To achieve the above object, the present invention provides a method for automatic classification of sperm integrity by flow cytometry based on machine learning, comprising the following steps:

[0007] Collecting cell characteristics of sperm samples by flow cytometry, and drawing a two-dimensional scatter plot of the cell characteristics in different fluorescence channels;

[0008] Normalizing the cell data points in the scatter plot to obtain corrected coordinate data corresponding to the cell data points;

[0009] Analyzing the scatter plot using a joint distribution map to determine the number of cell populations and initial centroid position coordinates in the scatter plot;

[0010] Based on the Canopy method, the cell data points in the scatter plot are preliminarily divided to obtain initial input data;

[0011] Clustering the cell data points in the scatter plot based on the k-medoids clustering method to obtain a clustering result;

[0012] A two-dimensional dot map is drawn based on the clustering results, and different types of cell clusters are demarcated with different polygonal gates to complete the automatic classification of cell populations.

[0013] The flow cytometer data includes four parameters, namely pulse area FSC-A, pulse area SSC-A, pulse area FITC-A and pulse area PerCP-Cy5.5-A.

[0014] Wherein, the fluorescence channel includes forward scattered light, side scattered light and multiple other fluorescence signal channels.

[0015] The specific method of normalizing the cell data points in the scatter plot to obtain the corrected coordinate data corresponding to the cell data points is as follows:

[0016] Based on the cell data points in the scatter plot, find the minimum and maximum values ​​of the data on the X-axis and the Y-axis;

[0017] Based on the minimum value and the maximum value, the normalized value of each cell scatter point coordinate is calculated by the minimum-maximum normalization method to obtain the corresponding corrected coordinate data.

[0018] The specific method of analyzing the scatter plot using the joint distribution map to determine the number of cell populations and the initial centroid position coordinates in the scatter plot is:

[0019] Draw a joint distribution diagram of the scatter plot, and find the histogram curves on the X-axis and the Y-axis;

[0020] Based on the histogram curve, peaks and troughs are identified, their number and position are analyzed, and the number of cell populations and initial centroid position coordinates in the scatter plot are determined.

[0021] The invention discloses a method for automatic classification of sperm integrity by flow cytometry based on machine learning. The method comprises the following steps: collecting cell features of a sperm sample by flow cytometry, and drawing a two-dimensional scatter plot of the cell features on different fluorescence channels; normalizing the cell data points in the scatter plot to obtain the corrected coordinate data corresponding to the cell data points; analyzing the scatter plot by using a joint distribution map to determine the number of cell populations and the initial centroid position coordinates in the scatter plot; performing a preliminary division of the cell data points in the scatter plot based on a Canopy method to obtain initial input data; and clustering the cell data points in the scatter plot based on a k-medoids clustering method. Clustering is performed to obtain cell populations with similar characteristics and obtain clustering results; a two-dimensional dot map is drawn based on the clustering results, different types of cell clusters are demarcated with different polygonal gates, and automatic classification of cell populations is completed. This method can automatically calculate the sperm DNA fragmentation rate (DFI) and high DNA staining index (HDS), thereby improving the accuracy and efficiency of sperm quality assessment. This method avoids the problems of high subjectivity, low efficiency and lack of consistency caused by manual classification, significantly improves the accuracy and efficiency of sperm cell population classification, enhances the reliability and repeatability of classification results, and provides strong support for the efficient detection of sperm nuclear integrity. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 This is a flowchart of a method for automatic classification of sperm integrity by flow cytometry based on machine learning provided by the present invention.

[0024] Figure 2 It is a two-dimensional cell scatter plot.

[0025] Figure 3 This is a schematic diagram of automatic cell classification.

[0026] Figure 4 The present invention provides a flow chart of a method for automatic classification of sperm integrity by flow cytometry based on machine learning.

[0027] Figure 5 It is a flowchart of a specific method for normalizing the cell data points in the scatter plot and obtaining the correction coordinate data corresponding to the cell data points.

[0028] Figure 6It is a flowchart of a specific method of analyzing the scatter plot using a joint distribution map to determine the number of cell populations and the initial centroid position coordinates in the scatter plot. DETAILED DESCRIPTION

[0029] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0030] See also Figures 1 to 6 The present invention provides a method for automatic classification of sperm integrity by flow cytometry based on machine learning, comprising the following steps:

[0031] S1 collects cell characteristics of sperm samples by flow cytometry, and draws a two-dimensional scatter plot of the cell characteristics in different fluorescence channels;

[0032] In an embodiment of the present invention, the fluorescence channel includes forward scattered light, side scattered light and multiple other fluorescence signal channels.

[0033] The flow cytometer data contains 4 parameters, namely pulse area (FSC-A, SSC-A, FITC-A, PerCP-Cy5.5-A). In each sample, the following two two-dimensional scatter plots are drawn: 1) the X-axis channel is the FSC-A channel, and the Y-axis is the SSC-A channel; 2) the X-axis is the PerCP-Cy5.5-A channel, and the Y-axis is the FITC-A channel. Through these scatter plots, the characteristic distribution of each cell in the sample can be reflected. The cell scatter plot data in all samples are collected for subsequent data processing and analysis. Figure 2 Shown is an example of a two-dimensional cell scatter plot drawn by flow cytometry software.

[0034] S2 normalizes the cell data points in the scatter plot to obtain corrected coordinate data corresponding to the cell data points;

[0035] Specific method:

[0036] S21, based on the cell data points in the scatter plot, find the minimum and maximum values ​​of the data on the X-axis and the Y-axis;

[0037] In the embodiment of the present invention, in each scatter plot, the coordinates of all cell scatter points are normalized to the minimum and maximum values ​​on the x-axis and the y-axis respectively.

[0038] S22 calculates the normalized value of each cell scatter point coordinate based on the minimum value and the maximum value by a minimum-maximum normalization method to obtain corresponding corrected coordinate data.

[0039] In the embodiment of the present invention, the normalized value calculation formula is as follows:

[0040]

[0041] Among them, x represents the x-axis or y-axis coordinate value of the current coordinate point, min(x) represents the minimum x-axis or y-axis coordinate value of all scatter points in the current scatter plot, and max(x) represents the maximum x-axis or y-axis coordinate value of all scatter points in the current scatter plot. Through normalization processing, the influence of different measurement conditions between different samples can be eliminated, making subsequent analysis more accurate.

[0042] S3 analyzes the scatter plot using a joint distribution map to determine the number of cell populations and initial centroid position coordinates in the scatter plot;

[0043] Specific method:

[0044] S31 draws a joint distribution diagram of the scatter plot, and finds the histogram curves on the X-axis and the Y-axis;

[0045] In an embodiment of the present invention, a histogram is drawn for each scatter plot in the X-axis and Y-axis directions. These histograms show the distribution frequency of cells in each direction. The histograms in the X-axis and Y-axis directions are combined to form a joint distribution graph. This joint distribution graph shows the joint probability density distribution of cells in two-dimensional space.

[0046] S32 identifies peaks and troughs based on the histogram curve, analyzes their number and position, and determines the number of cell populations in the scatter plot and the coordinates of the initial centroid position.

[0047] In an embodiment of the present invention, by observing the density distribution in the joint distribution map, medium and high density areas are identified, which usually correspond to cell populations. On the joint distribution map, these areas appear as concentrated point clouds or peaks. Further analyzing the peaks and troughs in the histogram, the peaks usually represent the concentrated positions of the cell populations, while the troughs may represent the boundaries between the populations. With this information, the boundary position of each cell population can be preliminarily determined. In addition, the joint distribution map can also help identify potential overlapping areas of cell populations, where cells from multiple populations may be mixed. This recognition provides basic support for subsequent clustering analysis or grouping, and helps to divide cell populations more accurately.

[0048] S4 performs preliminary division of the cell data points in the scatter plot based on the Canopy method to obtain initial input data;

[0049] In the embodiment of the present invention, based on the Canopy method, the cell population in the scatter plot is preliminarily divided to provide initial input data for the subsequent clustering method. The steps based on the Canopy method are as follows:

[0050] Step 1: Set two distance thresholds T1 and T2, where T1>T2. These two thresholds are used to define the coverage of Canopy, T1 is used to preliminarily determine the group, and T2 is used to further accurately divide the data points;

[0051] Step 2: The normalized data points in step 2 are used as the unassigned data set;

[0052] Step 3: Use the initial centroid position coordinates determined in step 3 as the center point of each cell population. Select one of them as the starting Canopy center point;

[0053] Step 4: Add all data points whose distance from the current Canopy center is less than T1 to the current group. These data points temporarily belong to the group corresponding to the current Canopy center. Remove all data points whose distance from the current Canopy center is less than T2 from the unassigned data set to avoid these data points being assigned to multiple Canopy centers, thereby reducing overlap.

[0054] Step 5: Select the center point of the next cell population as the new Canopy center point and repeat step 4 until all data points are assigned to a population.

[0055] Through these steps, the Canopy method can effectively divide the cell data points in the scatter plot and provide high-quality initial input data for the subsequent clustering method. This method helps to reduce the amount of calculation in the clustering process and improve the accuracy of the clustering results.

[0056] S5 clusters the cell data points in the scatter plot based on a k-medoids clustering method to obtain a clustering result;

[0057] In the embodiment of the present invention, the specific steps based on the k-medoids clustering method are as follows:

[0058] Step 1: Based on the number of cell clusters determined in step 3, determine the number of clusters K. This means dividing the data points into K different cell populations;

[0059] Step 2: According to the initial centroid position coordinates determined in step 3, set K initial centroids in the scatter plot. These centroids will serve as initial cluster centers to guide the initial allocation of clusters;

[0060] Step 3: Use the cell population data initially divided by the Canopy method in step 4 as the initial input data for each population. This can provide a good starting point for clustering and improve clustering efficiency;

[0061] Step 4: For each population, calculate the total dissimilarity of each data point in the population relative to the population centroid. Here, Euclidean distance is used as the metric. i , assuming its center is k i , there are n i Data points x j , where j = 1, 2, ..., n i , the total dissimilarity D i The calculation formula is as follows:

[0062]

[0063] Among them, x jl and k il The data points x j and the center point k i Coordinates in the lth dimension. The maximum data dimension of a two-dimensional scatter plot is 2.

[0064] Step 5: In each iteration, select the data point with the smallest total dissimilarity as the new centroid and update the center position of the group.

[0065] Step 6: Iterate the process multiple times until the preset number of iterations is reached, or the centroid position of the group no longer changes, and the clustering is terminated. At this point, the cell data points in the scatter plot have been divided into multiple cell groups with similar characteristics.

[0066] Through the above steps, the k-medoid method can effectively perform cluster analysis on flow cytometry data and divide cell data points into multiple accurate cell populations. These populations can be further used for sperm quality assessment and research.

[0067] S6 draws a two-dimensional dot map based on the clustering results, demarcates different types of cell clusters with different polygonal gates, and completes automatic classification of cell populations.

[0068] In the embodiment of the present invention, a two-dimensional dot map is drawn based on the results of the clustering model, and different types of cell clusters are demarcated with different polygonal gates to complete the automatic classification of cell populations. The specific steps are as follows:

[0069] Based on the results of the k-medoids clustering in the previous step, a two-dimensional scatter plot is drawn. In the figure, each data point represents a cell and is identified according to the cluster group to which it belongs. In order to clearly display different types of cell populations, they are circled with different polygon gates for partitioning. Figure 3 The schematic diagram of the two-dimensional scatter plot after automatic classification of cell populations is shown. The scatter points encircled by the upper polygonal gate are the cell populations required to calculate the high DNA staining index (HDS), and the scatter points encircled by the lower polygonal gate are the cell populations required to calculate the sperm DNA fragmentation rate (DFI). This method of using different polygonal gates to delineate partitions helps to intuitively observe and analyze the distribution of each cell population. After the classification is completed, the values ​​of the high DNA staining index (HDS) and sperm DNA fragmentation rate (DFI) can be obtained by calculating the proportion of the HDS population and the DFI population in the total sample cells. These two indicators are of great significance for evaluating sperm quality and fertility. The distribution of different populations can be clearly seen through polygonal gate partitioning, which is convenient for researchers to further analyze and make decisions.

[0070] Through these steps, researchers can automatically and accurately classify cell populations, greatly improving the efficiency and accuracy of sperm quality assessment and making the calculation of HDS and DFI more reliable.

[0071] The above disclosure is only a preferred embodiment of the method for automatic classification of sperm integrity by flow cytometry based on machine learning of the present invention. Of course, it cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for automatic classification of sperm integrity by flow cytometry based on machine learning, characterized in that: The following steps are involved: Collecting cell characteristics of sperm samples by flow cytometry, and drawing a two-dimensional scatter plot of the cell characteristics in different fluorescence channels; Normalizing the cell data points in the scatter plot to obtain corrected coordinate data corresponding to the cell data points; Analyzing the scatter plot using a joint distribution map to determine the number of cell populations and initial centroid position coordinates in the scatter plot; Based on the Canopy method, the cell data points in the scatter plot are preliminarily divided to obtain initial input data; Clustering the cell data points in the scatter plot based on the k-medoids clustering method to obtain a clustering result; A two-dimensional dot map is drawn based on the clustering results, and different types of cell clusters are demarcated with different polygonal gates to complete the automatic classification of cell populations.

2. The automatic classification method for sperm integrity by flow cytometry based on machine learning as claimed in claim 1, characterized in that: The flow cytometer data includes four parameters, namely pulse area FSC-A, pulse area SSC-A, pulse area FITC-A and pulse area PerCP-Cy5.5-A.

3. The automatic classification method for sperm integrity by flow cytometry based on machine learning as claimed in claim 1, characterized in that: The fluorescence channel includes forward scattered light, side scattered light and multiple other fluorescence signal channels.

4. The method for automatic classification of sperm integrity by flow cytometry based on machine learning as claimed in claim 1, characterized in that: The specific method of normalizing the cell data points in the scatter plot to obtain the corrected coordinate data corresponding to the cell data points is: Based on the cell data points in the scatter plot, find the minimum and maximum values ​​of the data on the X-axis and the Y-axis; Based on the minimum value and the maximum value, the normalized value of each cell scatter point coordinate is calculated by the minimum-maximum normalization method to obtain the corresponding corrected coordinate data.

5. The automatic classification method for sperm integrity by flow cytometry based on machine learning as claimed in claim 1, characterized in that: The specific method of analyzing the scatter plot using the joint distribution map to determine the number of cell populations and the initial centroid position coordinates in the scatter plot is: Draw a joint distribution diagram of the scatter plot, and find the histogram curves on the X-axis and the Y-axis; Based on the histogram curve, peaks and troughs are identified, their number and position are analyzed, and the number of cell populations and initial centroid position coordinates in the scatter plot are determined.

Images