Cell image registration method and system, electronic equipment and storage medium

Abstract

The invention provides a cell image registration method and system, electronic equipment and a storage medium. The cell image registration method comprises the following steps: acquiring cell contours in a cell image map and a space transcriptome data map; respectively determining a part of target cells based on the definition of the cell contour; extracting geometric centers of a part of target cells, and respectively forming a cell image two-dimensional point cloud set and a space transcriptome data graph two-dimensional point cloud set; and finally obtaining registration parameters. And registering the cell image map and the spatial transcriptome data map according to registration parameters. According to the method, the cell contours and the definition of the cell contours in the cell image map and the space transcriptome data map are obtained, then the two-dimensional point cloud sets are formed based on the geometric centers of part of target cells, point cloud registration algorithm processing is carried out on the two-dimensional point cloud sets, and finally the registration parameters are obtained to carry out registration on the cell image map and the space transcriptome data map. The defect that in-situ analysis cannot be carried out is overcome, and the accuracy of a space transcriptome sequencing technology in a space-time omics technology is improved.

Description

Cell image registration method, system, electronic device and storage medium Technical Field

[0001] The present disclosure relates to the field of cell analysis, and in particular to a cell image registration method, system, electronic device, and storage medium. Background Art

[0002] In the Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology, the cell image is obtained by photographing with optical instruments and other equipment; the spatial transcriptome data map is quite different from the cell image map. The single-stranded DNA in the cell is first produced by RCA (rolling circle amplification technology) to produce DNB (DNA nano ball), and then the DNB sequence is obtained by DNB sequencing technology. Using CID (spatial barcode) technology, the DNB signal is mapped to the corresponding spatial position to form a spatial transcriptome data map. Therefore, the spatial transcriptome data map is a discrete image. Before analysis, the cell image map and the spatial transcriptome data image must first be aligned for in situ analysis. However, in the spatial transcriptome data map, it is difficult to obtain the position information of the cell tissue outline like the cell image map. Among existing registration technologies, most are based on feature similarity matching based on the features of the cell image itself, thereby obtaining a positional mapping relationship between the two images to be registered. However, the distribution of pixel values ​​of spatial transcriptome data maps is discrete, and there is no better method to align cell images with spatial transcriptome data maps, making it impossible to accurately perform in situ analysis of cell images and spatial transcriptome data maps.

[0003] Summary of the Invention

[0004] The technical problem to be solved by the present disclosure is to overcome the defect in the prior art that digital cell images and spatial transcription data images cannot be accurately analyzed in situ, and to provide a cell image registration method, system, electronic device and storage medium.

[0005] The present disclosure solves the above technical problems through the following technical solutions:

[0006] A cell image registration method, comprising:

[0007] Obtaining cell outlines in a cell image and a spatial transcriptome data map, wherein the cell image and the spatial transcriptome data map have a corresponding relationship;

[0008] Based on the clarity of the cell outlines in the cell image and the spatial transcriptome data map, respectively, determining some target cells;

[0009] Extracting the geometric centers of the target cells to form a two-dimensional point cloud set of a cell image and a two-dimensional point cloud set of a spatial transcriptome data map;

[0010] Processing the two-dimensional point cloud set of the cell image map and the two-dimensional point cloud set of the spatial transcriptome data map based on a point cloud registration algorithm to obtain registration parameters;

[0011] The cell image map and the spatial transcriptome data map are registered according to the registration parameters.

[0012] Preferably, the step of processing the two-dimensional point cloud set of the cell image map and the two-dimensional point cloud set of the spatial transcriptome data map based on a point cloud registration algorithm to obtain registration parameters includes:

[0013] Get the initial values ​​of rotation parameters and translation parameters;

[0014] Performing transformation processing on the two-dimensional point cloud set of the cell image based on the initial value of the rotation parameter and the initial value of the translation parameter;

[0015] Obtaining corresponding points in the two-dimensional point cloud set of the cell image and the two-dimensional point cloud set of the spatial transcriptome data image;

[0016] Updating the rotation parameter and the translation parameter by using an objective function, wherein the objective function is used to process the acquired corresponding points;

[0017] Determine whether a preset condition is met, and if so, obtain the rotation parameter and the translation parameter as a final value of the rotation parameter and a final value of the translation parameter;

[0018] If not, the rotation parameter and the translation parameter are updated to the initial rotation parameter value and the initial translation parameter value, and the process returns to the step of transforming the two-dimensional point cloud set of the cell image based on the initial rotation parameter value and the initial translation parameter value.

[0019] Preferably, the step of respectively determining some target cells based on the clarity of cell outlines in the cell image and the spatial transcriptome data map comprises:

[0020] Obtaining a first score of the cell outline in the cell image, and obtaining a second score of the cell outline in the spatial transcriptome data graph, wherein the first score and the second score are positively correlated with the clarity;

[0021] determining a portion of target cells according to the first score and the second score;

[0022] The first score is calculated based on at least one of the eccentricity, pixel gray value, and pixel area of ​​the cell in the cell image;

[0023] The second score is calculated based on at least one of the amount of RNA in the cell and the pixel area in the spatial transcriptome data map.

[0024] Preferably, the target cells are: cells of a preset proportion or a preset number of cells selected according to the first score in the cell image;

[0025] A preset proportion of cells or a preset number of cells are selected according to the second score in the spatial transcriptome data graph.

[0026] Preferably, the first score is determined by the following formula:

[0027] S 第一分值 =a / S x +b*S y +c*S z ;

[0028] Among them, S x is the cell eccentricity in the cell image, S y is the average gray value of the pixels in the cell image, S z is the pixel area of ​​the cell in the cell image; the sum of a, b, and c is 1;

[0029] The second score is determined by the following formula:

[0030] S 第二分值 =m*S p +n*S q ;

[0031] Among them, S p is the pixel area of ​​the cell in the spatial transcriptome data map, S q is the number of RNA in the cell in the spatial transcriptome data map; the sum of m and n is 1.

[0032] Preferably, the step of registering the cell image map and the spatial transcriptome data map according to the registration parameters comprises:

[0033] transforming the cell image based on the registration parameters;

[0034] Complete the registration of the cell image map and the spatial transcriptome data map.

[0035] Preferably, the step of obtaining corresponding points in the two-dimensional point cloud set of the cell image and the two-dimensional point cloud set of the spatial transcriptome data image is:

[0036] Calculating the Euclidean distance between each point in the two-dimensional point cloud set of the cell image graph and each point in the two-dimensional point cloud set of the spatial transcriptome data graph;

[0037] The two points with the smallest Euclidean distance are taken as the corresponding points.

[0038] Preferably, the preset conditions are:

[0039] The objective function is less than a convergence distance threshold; and / or,

[0040] The number of updates reaches an update number threshold.

[0041] Preferably, the convergence distance threshold is positively correlated with the intercellular distance in the cell image map and the spatial transcriptome data map.

[0042] Preferably, the objective function is:

[0043]

[0044] Where D is the objective function, i is the number of updates, N is the number of point pairs of the corresponding points, a i is a point in the two-dimensional point cloud of the cell image, b i The spatial transcriptome data is a two-dimensional point cloud with a i , R is the rotation parameter, and T is the translation parameter.

[0045] A cell image registration system, comprising:

[0046] An acquisition module, configured to acquire cell outlines in a cell image map and a spatial transcriptome data map, wherein the cell image map and the spatial transcriptome data map have a corresponding relationship;

[0047] a target cell determination module, configured to determine a portion of target cells based on the clarity of cell outlines in the cell image and the spatial transcriptome data map;

[0048] An extraction module extracts the geometric centers of the target cells to form a two-dimensional point cloud set of a cell image and a two-dimensional point cloud set of a spatial transcriptome data map;

[0049] A registration parameter acquisition module is used to process the two-dimensional point cloud set of the cell image map and the two-dimensional point cloud set of the spatial transcriptome data map based on a point cloud registration algorithm to obtain registration parameters;

[0050] A registration module is used to register the cell image and the spatial transcriptome data map according to the registration parameters.

[0051] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the above-mentioned cell image registration methods when executing the computer program.

[0052] A computer-readable storage medium stores a computer program, wherein the computer program implements any one of the above-mentioned cell image registration methods when executed by a processor.

[0053] The positive progress of the present disclosure is that: the present disclosure obtains the cell contours and clarity in the cell image and spatial transcriptome data map, respectively determines some target cells, and then forms a two-dimensional point cloud set of the cell image and a two-dimensional point cloud set of the spatial transcriptome data map based on the geometric centers of some target cells, and performs point cloud alignment algorithm processing on them. Finally, the alignment parameters are obtained to align the cell image and the spatial transcriptome data map, thereby overcoming the defect of being unable to accurately perform in situ analysis of cell digital images and spatial transcription data images, and improving the accuracy of Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] FIG1 is a flow chart of a cell image registration method according to Example 1 of the present disclosure.

[0055] FIG2 is a schematic diagram of the outline of a circled cell in the cell image of Example 1 of the present disclosure.

[0056] FIG3 is a schematic diagram of the outlines of the circled cells in the spatial transcriptome data graph of Example 1 of the present disclosure.

[0057] FIG4 is a diagram of cell images to be registered in Example 1 of the present disclosure.

[0058] FIG5 is a diagram of spatial transcriptome data to be aligned in Example 1 of the present disclosure.

[0059] FIG6 is a schematic diagram of the result of the two-dimensional point cloud set of the cell image in Example 1 of the present disclosure.

[0060] FIG7 is a schematic diagram of the result of the two-dimensional point cloud set of the spatial transcriptome data graph of Example 1 of the present disclosure.

[0061] FIG8 is a partial flow chart of the cell image registration method of Example 1 of the present disclosure.

[0062] FIG9 is a partial flow chart of the cell image registration method of Example 1 of the present disclosure.

[0063] FIG10 is a module schematic diagram of the cell image registration system of Example 2 of the present disclosure.

[0064] FIG11 is a schematic structural diagram of an electronic device according to Embodiment 3 of the present disclosure. DETAILED DESCRIPTION

[0065] The present disclosure is further illustrated below by way of examples, but the present disclosure is not limited to the scope of the examples.

[0066] Example 1

[0067] The present disclosure provides a cell image registration method. Referring to FIG1 , the cell image registration method includes:

[0068] S1. Obtain cell outlines in a cell image and a spatial transcriptome data map, wherein the cell image and the spatial transcriptome data map have a corresponding relationship.

[0069] Among them, the cell image is a digital image taken of the cell by an optical instrument (such as a microscope). First, the cell outline in the cell image is obtained. The cell outline in the cell image can be obtained by circling, and the main basis for circling the cell is the grayscale value information of the image. Referring to Figure 2, the image of the circled cell outline can be clearly seen, and the highlighted area in the figure is the cell area. The cell outline can be circled using the deep learning model deepcell (a high-precision cell recognition engine based on deep learning technology) or the traditional watershed algorithm. The irregular circle is the cell outline. The cell outline in the spatial transcriptome data map can also be obtained by circling. The main basis for circling the cell is the density of the RNA space. Referring to Figure 3, the image of the circled cell outline can be clearly seen. The area with white highlight (local aggregation area) in the figure is the cell area, and the irregular circle is the cell outline. In addition, the method of circling cells can also use the cell bin image processing process supporting stereo-seq (spatiotemporal omics) technology, or other algorithms and manual tools can be used. Furthermore, the cell image and the spatial transcriptome data map have a corresponding relationship, both being images obtained after processing cells in the same region. Taking the cell circle results in the image map as an example, the results of cell selection can be stored as a binary image, commonly called a cell mask, where a pixel value of 0 represents the background (the area outside the cell outline) and a pixel value of 1 represents the cell (the area inside the cell outline).

[0070] S2. Based on the clarity of the cell outlines in the cell image and the spatial transcriptome data map, respectively determine some target cells.

[0071] Among them, based on the clarity of the circled cell outlines in the cell image and the spatial transcriptome data map, some cells are determined as target cells for the subsequent calculation of alignment parameters.

[0072] S3. Extracting the geometric centers of the target cells to form a two-dimensional point cloud set of a cell image and a two-dimensional point cloud set of a spatial transcriptome data map.

[0073] Among them, the role of extracting the geometric center of the cells in the cell image and the spatial transcriptome data map is to roughly locate the position of the cells in order to obtain the relative positions between some target cells. According to the geometric centers of some target cells, a two-dimensional point cloud set of the cell image and a two-dimensional point cloud set of the spatial transcriptome data map are formed respectively, and the point cloud set contains many point cloud data. The method for obtaining the two-dimensional point cloud set of the cell image and the two-dimensional point cloud set of the spatial transcriptome data map can be, for each target cell selected (each irregular circular connected domain), obtaining the geometric center point coordinate value of the target cell pixel in the x direction and y direction as the geometric center of the target cell currently selected (if the geometric center is selected as the center of gravity here, then obtain the mean value of the target cell pixel in the x direction and y direction). The geometric center is extracted for all the target cells selected in the cell image and the spatial transcriptome data map, and these center coordinate points (x, y) are stored (for example, in a .txt file) as a two-dimensional point cloud set of the cell image and a two-dimensional point cloud set of the spatial transcriptome data map.

[0074] See Figures 4 and 5. Figure 4 shows the cell image to be registered, and the numbers in Figure 4 are the numbers of the circled cells, where the numbers are used to index the circled cells. Figure 5 shows the spatial transcriptome data, and the numbers in Figure 5 are the numbers of the circled cells. Also see Figures 6 and 7. Figure 6 is a schematic diagram of the result of forming a two-dimensional point cloud set for the cell image, and Figure 7 is a schematic diagram of the result of forming a two-dimensional point cloud set for the spatial transcriptome data.

[0075] S4. Processing the two-dimensional point cloud set of the cell image map and the two-dimensional point cloud set of the spatial transcriptome data map based on a point cloud registration algorithm to obtain registration parameters.

[0076] Among them, the point cloud registration algorithm can use ICP (Iterative Closest Point) to obtain registration parameters.

[0077] S5. Registering the cell image and the spatial transcriptome data graph according to the registration parameters.

[0078] Among them, the cell image is transformed by the obtained registration parameters to complete the registration with the spatial transcriptome data map.

[0079] In this embodiment, by obtaining the cell contours and clarity in the cell image and the spatial transcriptome data map, some target cells are determined respectively, and then based on the geometric centers of some target cells, a two-dimensional point cloud set of the cell image and a two-dimensional point cloud set of the spatial transcriptome data map are formed respectively and processed by a point cloud registration algorithm. Finally, the registration parameters are obtained to align the cell image and the spatial transcriptome data map, thereby overcoming the defect of being unable to accurately perform in situ analysis of cell digital images and spatial transcription data images, and improving the accuracy of Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology.

[0080] In an optional embodiment, the geometric center includes the center of gravity, and step S3 includes:

[0081] The centroid method is used to extract the centroids of cells in the cell image map and the spatial transcriptome data map to form a two-dimensional point cloud set of the cell image map and a two-dimensional point cloud set of the spatial transcriptome data map, respectively.

[0082] In an optional embodiment, the registration parameters include rotation parameters and translation parameters. Referring to FIG8 , step S4 includes:

[0083] S41. Obtain initial values ​​of rotation parameters and initial values ​​of translation parameters.

[0084] S42: performing transformation processing on the two-dimensional point cloud set of the cell image based on the initial value of the rotation parameter and the initial value of the translation parameter.

[0085] Assume that the 2D point cloud set of the digital cell image is A and the 2D point cloud set of the spatial transcription data image is B. The rotation parameter R and translation parameter T are randomly initialized as the initial values ​​of the rotation parameter and the initial values ​​of the translation parameter. The 2D point cloud set A of the cell image image is processed based on the initial values ​​of the rotation parameter and the initial values ​​of the translation parameter.

[0086] S43, obtaining corresponding points in the two-dimensional point cloud set of the cell image and the two-dimensional point cloud set of the spatial transcriptome data image.

[0087] The distance between each point in the two-dimensional point cloud set A of the cell image and each point in the two-dimensional point cloud set B of the spatial transcriptome data is calculated, and the two points with the closest distance are taken as corresponding points to obtain the corresponding point pair a. i and b i , a i+1 and b i+1 ,…a N- 1 and b N-1 , a N and b N .

[0088] S44. Updating the rotation parameter and the translation parameter by using an objective function, wherein the objective function is used to process the acquired corresponding points;

[0089] The least squares method is used to optimize the objective function to obtain the updated rotation parameter R and translation parameter T.

[0090] S45, determining whether a preset condition is met, and if so, obtaining the rotation parameter and the translation parameter as final rotation parameter values ​​and final translation parameter values, and using them as registration parameters;

[0091] If not, the rotation parameter and the translation parameter are updated to the initial rotation parameter value and the initial translation parameter value, and the step of transforming the two-dimensional point cloud set of the cell image based on the initial rotation parameter value and the initial translation parameter value is returned to.

[0092] The preset conditions are: the objective function is less than the convergence distance threshold; and / or the number of updates reaches the update threshold. When the number of updates equals the update threshold or the objective function is less than the convergence distance threshold, the update iteration is stopped and the final values ​​of the rotation parameters and the translation parameters are obtained. When both conditions are used simultaneously, the update results are more accurate. The objective function can be:

[0093]

[0094] D is the objective function, i is the number of updates, N is the number of point pairs of the corresponding points, a i is a point in the two-dimensional point cloud of the cell image, b i The spatial transcriptome data is a two-dimensional point cloud with a i , R is the rotation parameter, and T is the translation parameter.

[0095] Here, it is determined whether the iteration of the process of updating the rotation parameter R and the translation parameter T meets the preset conditions: if so, the last updated rotation parameter and the translation parameter are obtained as the final rotation parameter value and the final translation parameter value. If not, the rotation parameter and the translation parameter updated in step S44 are used as the initial rotation parameter value and the initial translation parameter value, and the process returns to step S42 to continue updating the rotation parameter R and the translation parameter T.

[0096] In this embodiment, by optimizing and updating the rotation and translation parameters, more accurate rotation and translation parameters are ultimately obtained. This overcomes the drawback of being unable to accurately perform in situ analysis of digital cell images and spatial transcriptome data images, and improves the accuracy of Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology.

[0097] In an optional embodiment, referring to FIG9 , step S2 includes:

[0098] S21. Obtain a first score of the cell contour in the cell image, and obtain a second score of the cell contour in the spatial transcriptome data map, wherein the first score and the second score are positively correlated with the clarity.

[0099] S22. Determine a portion of target cells according to the first score and the second score.

[0100] Wherein, the first score is calculated based on at least one of the eccentricity, pixel grayscale value and pixel area of ​​the cell in the cell image. It should be understood by those skilled in the art that the above three are used as reference factors at the same time, and the first score calculated is more accurate, and the higher the first score, the clearer the cells in the cell image. The second score is calculated based on at least one of the amount of RNA in the cell and the pixel area in the spatial transcriptome data map. The above three are used as reference factors at the same time, and the second score calculated is more accurate, and the higher the second score, the clearer the cells in the spatial transcriptome data map. By using the first score of the cell in the cell image map and the second score of the cell in the spatial transcriptome data map, some target cells are selected respectively, and the target cells are used as the basis for subsequent calculation of the registration parameters.

[0101] In this embodiment, by using the first and second scores to select a portion of target cells, the target cells can be more accurately acquired, thereby obtaining more accurate registration parameters. This overcomes the limitation of being unable to accurately analyze digital cell images and spatial transcriptome data images in situ, and improves the accuracy of Stereo-seq spatial transcriptome sequencing technology.

[0102] In an optional embodiment, the portion of target cells represents: a preset proportion of cells or a preset number of cells selected according to a first score in the cell image; a preset proportion of cells or a preset number of cells selected according to a second score in the spatial transcriptome data map.

[0103] For example, in a single image registration, target cells representing 15% of the total cell population in the cell image are selected based on the first score, and target cells representing 15% of the total cell population in the spatially transcribed data image are selected based on the second score. These target cells serve as the basis for subsequent calculation of registration parameters. The preset ratio and number are selected based on actual conditions.

[0104] In this embodiment, by selecting a preset ratio of cells as target cells, or selecting a preset number of cells as target cells, more accurate registration parameters can be obtained. This overcomes the limitation of being unable to accurately analyze digital cell images and spatial transcriptome data images in situ, and improves the accuracy of Stereo-seq spatial transcriptome sequencing technology.

[0105] In an optional embodiment, the first score and the second score are determined by the following formula:

[0106] S 第一分值 =a / S x +b*S y +c*S z ;

[0107] Among them, S x is the cell eccentricity in the cell image, S y is the average gray value of the pixels in the cell image, S z is the pixel area of ​​the cell in the cell image; the sum of a, b, and c is 1. Those skilled in the art will appreciate that the values ​​of a, b, and c are set based on actual conditions.

[0108] In a preferred solution, when a=0.3, b=0.4, and c=0.3, the obtained first score is more accurate and can more accurately represent the clarity.

[0109] S 第二分值 =m*S p +n*S q ;

[0110] Among them, S p is the pixel area of ​​the cell in the spatial transcriptome data map, S q is the number of RNA in the cell in the spatial transcriptome data graph; the sum of m and n is 1. It should be understood by those skilled in the art that the values ​​of m and n are set according to actual conditions.

[0111] In a preferred solution, when m=0.4 and n=0.6, the obtained first score is more accurate and can more accurately represent the clarity.

[0112] In this embodiment, the calculation formula of the first and second scores can more accurately select a portion of target cells and obtain more accurate registration parameters. This overcomes the limitation of inaccurate in situ analysis of digital cell images and spatial transcriptome data images, and improves the accuracy of Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology.

[0113] In an optional embodiment, step S5 includes:

[0114] Based on the registration parameters, the cell image is transformed to complete the registration of the cell image and the spatial transcriptome data map.

[0115] After obtaining the registration parameters, only the cell image is transformed based on the registration parameters, while the spatial transcriptome data remains unchanged. The rotation angle is determined by the rotation parameters in the registration parameters, and the translation length and direction of the x- and y-axis are determined by the translation parameters. Finally, the transformation is completed simultaneously, so that the cell image and spatial transcriptome data are registered.

[0116] In this embodiment, the cell image is transformed using the acquired registration parameters to align the cell image with the spatial transcriptome data. This overcomes the drawback of being unable to accurately perform in situ analysis of digital cell images and spatial transcriptome data images, and improves the accuracy of Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology.

[0117] In an optional embodiment, step S43 includes:

[0118] Calculate the Euclidean distance between each point in the two-dimensional point cloud set of the cell image map and each point in the two-dimensional point cloud set of the spatial transcriptome data map, and take the two points with the smallest Euclidean distance as corresponding points.

[0119] Among them, the Euclidean distance between each point in the two-dimensional point cloud set A of the cell image and each point in the two-dimensional point cloud set B of the spatial transcriptome data is calculated, and the two points with the closest Euclidean distance are taken as corresponding points to obtain the point pair a of the corresponding points. i and b i , a i+1 and b i+1 ,…a N-1 and b N-1 , a N and b N Those skilled in the art will appreciate that other distances besides the Euclidean distance may be selected, but the method of determining corresponding points by the Euclidean distance is more accurate in the present disclosure.

[0120] In this embodiment, by selecting Euclidean distance to determine corresponding points in the two-dimensional point cloud of the cell image and the two-dimensional point cloud of the spatial transcriptome data, more accurate registration parameters can be obtained. This overcomes the limitation of inaccurate in situ analysis of digital cell images and spatial transcriptome data images, and improves the accuracy of Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology.

[0121] In an optional embodiment, the convergence distance threshold is positively correlated with the intercellular distance in the cell image map and the spatial transcriptome data map.

[0122] The convergence distance threshold should be selected based on the actual intercellular distances in the cell image and the spatial transcriptome data map to be registered, as well as the distribution of different cell types. Preferably, the convergence distance threshold is positively correlated with the intercellular distances in the cell image and the spatial transcriptome data map.

[0123] In this embodiment, a convergence distance threshold is selected based on the actual intercellular distances between the cell image and the spatial transcriptome data map to be registered, which allows for more accurate registration parameters to be updated using the objective function. This overcomes the limitation of being unable to accurately perform in situ analysis of digital cell images and spatial transcriptome data images, and improves the accuracy of Stereo-seq spatial transcriptome sequencing technology.

[0124] Example 2

[0125] This embodiment provides a cell image registration system, and Figure 10 shows a schematic diagram of the modules of this embodiment. Referring to Figure 10, the cell image registration system of this embodiment includes:

[0126] The acquisition module 11 is used to acquire cell contours in the cell image map and the spatial transcriptome data map, wherein the cell image map and the spatial transcriptome data map have a corresponding relationship.

[0127] Among them, the cell image is a digital image taken of the cell by an optical instrument (such as a microscope). First, the cell outline in the cell image is obtained. The cell outline in the cell image can be obtained by circling, and the main basis for circling the cell is the grayscale value information of the image. Referring to Figure 2, the image of the circled cell outline can be clearly seen, and the highlighted area in the figure is the cell area. The cell outline can be circled using the deep learning model deepcell (a high-precision cell recognition engine based on deep learning technology) or the traditional watershed algorithm. The irregular circle is the cell outline. The cell outline in the spatial transcriptome data map can also be obtained by circling. The main basis for circling the cell is the density of the RNA space. Referring to Figure 3, the image of the circled cell outline can be clearly seen. The area with white highlight (local aggregation area) in the figure is the cell area, and the irregular circle is the cell outline. In addition, the method of circling cells can also use the cell bin image processing process supporting stereo-seq (spatiotemporal omics) technology, or other algorithms and manual tools can be used. Furthermore, the cell image and the spatial transcriptome data map have a corresponding relationship, both being images obtained after processing cells in the same region. Taking the cell circle results in the image map as an example, the cell selection results can be stored as a binary image, commonly called a cell mask, where a pixel value of 0 represents the background (the area outside the cell outline) and a pixel value of 1 represents the cell (the area inside the cell outline).

[0128] The target cell determination module 21 is configured to determine a portion of target cells based on the clarity of the cell contours in the cell image and the spatial transcriptome data map.

[0129] Among them, based on the clarity of the circled cell outlines in the cell image and the spatial transcriptome data map, some cells are determined as target cells for the subsequent calculation of alignment parameters.

[0130] The extraction module 31 extracts the geometric centers of the target cells to form a two-dimensional point cloud set of a cell image and a two-dimensional point cloud set of a spatial transcriptome data map.

[0131] Among them, the role of extracting the geometric center of the cells in the cell image and the spatial transcriptome data map is to roughly locate the position of the cells in order to obtain the relative positions between some target cells. According to the geometric centers of some target cells, a two-dimensional point cloud set of the cell image and a two-dimensional point cloud set of the spatial transcriptome data map are formed respectively, and the point cloud set contains many point cloud data. The method for obtaining the two-dimensional point cloud set of the cell image and the two-dimensional point cloud set of the spatial transcriptome data map can be, for each target cell selected (each irregular circular connected domain), obtaining the geometric center coordinate value of the target cell pixel in the x direction and the y direction as the geometric center of the target cell currently selected. The geometric center is extracted for all the target cells selected in the cell image and the spatial transcriptome data map, and these geometric center coordinate points (x, y) are stored (for example, in a .txt file) as a two-dimensional point cloud set of the cell image and a two-dimensional point cloud set of the spatial transcriptome data map.

[0132] See Figures 4 and 5. Figure 4 shows the cell image to be registered, and the numbers in Figure 4 are the numbers of the circled cells; Figure 5 shows the spatial transcriptome data, and the numbers in Figure 5 are the numbers of the circled cells. Also see Figures 6 and 7. Figure 6 shows the result of forming a two-dimensional point cloud set for the cell image, and Figure 7 shows the result of forming a two-dimensional point cloud set for the spatial transcriptome data.

[0133] The registration parameter acquisition module 41 is used to process the two-dimensional point cloud set of the cell image image and the two-dimensional point cloud set of the spatial transcriptome data image based on a point cloud registration algorithm to obtain registration parameters.

[0134] Among them, the point cloud registration algorithm can use ICP (Iterative Closest Point) to obtain registration parameters.

[0135] The registration module 51 registers the cell image and the spatial transcriptome data graph according to the registration parameters.

[0136] Among them, the cell image is transformed by the obtained registration parameters to complete the registration with the spatial transcriptome data map.

[0137] In this embodiment, by obtaining the cell contours and clarity in the cell image and the spatial transcriptome data map, some target cells are determined respectively, and then based on the geometric centers of some target cells, a two-dimensional point cloud set of the cell image and a two-dimensional point cloud set of the spatial transcriptome data map are formed respectively and processed by a point cloud registration algorithm. Finally, the registration parameters are obtained to align the cell image and the spatial transcriptome data map, thereby overcoming the defect of being unable to accurately perform in situ analysis of cell digital images and spatial transcription data images, and improving the accuracy of Stereo-seq (spatiotemporal omics) spatial transcriptome sequencing technology.

[0138] Example 3

[0139] This embodiment provides an electronic device, which can be expressed in the form of a computing device (for example, a server device), including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the cell image registration method provided in Example 1 when executing the computer program.

[0140] FIG11 shows a schematic diagram of the hardware structure of this embodiment. As shown in FIG11 , the electronic device 9 specifically includes:

[0141] At least one processor 91, at least one memory 92, and a bus 93 for connecting different system components (including the processor 91 and the memory 92), wherein:

[0142] The bus 93 includes a data bus, an address bus, and a control bus.

[0143] The memory 92 includes a volatile memory, such as a random access memory (RAM) 921 and / or a cache memory 922 , and may further include a read-only memory (ROM) 923 .

[0144] Memory 92 also includes a program / utility 925 having a set (at least one) of program modules 924, such program modules 924 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0145] The processor 91 executes various functional applications and data processing by running the computer programs stored in the memory 92, such as the cell image registration method provided in Example 1 of the present disclosure.

[0146] The electronic device 9 can further communicate with one or more external devices 94. Such communication can be performed via an input / output (I / O) interface 95. Furthermore, the electronic device 9 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 96. The network adapter 96 communicates with other modules of the electronic device 9 via a bus 93. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the electronic device 9, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0147] It should be noted that while the above detailed description refers to several units / modules or sub-units / modules of the heating device, this division is merely exemplary and not mandatory. In practice, depending on the embodiment of the present application, the features and functions of two or more units / modules described above may be embodied in a single unit / module. Conversely, the features and functions of a single unit / module described above may be further divided and embodied by multiple units / modules.

[0148] Example 4

[0149] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the cell image registration method provided in Example 1 is implemented.

[0150] Specifically, the readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0151] In a possible implementation, the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the cell image registration method provided in Example 1.

[0152] The program code for executing the present disclosure may be written in any combination of one or more programming languages, and the program code may be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0153] Although the above describes specific embodiments of the present invention, it should be understood by those skilled in the art that these are merely illustrative and that various changes or modifications may be made to these embodiments without departing from the principles and essence of the present invention. Therefore, the scope of protection of the present invention is defined by the appended claims.

Claims

1. A cell image registration method, It is characterized in that The method comprises: Acquire cell contours in a cell image and a spatial transcriptome data map, wherein the cell image and the spatial transcriptome data map have a corresponding relationship; Based on the clarity of the cell outlines in the cell image and the spatial transcriptome data map, respectively determine some target cells; Extracting the geometric centers of the target cells to form a two-dimensional point cloud set of a cell image and a two-dimensional point cloud set of a spatial transcriptome data map; Processing the two-dimensional point cloud set of the cell image map and the two-dimensional point cloud set of the spatial transcriptome data map based on a point cloud registration algorithm to obtain registration parameters; The cell image map and the spatial transcriptome data map are registered according to the registration parameters.

2. The cell image registration method according to claim 1, It is characterized in that The registration parameters include rotation parameters and translation parameters. The step of processing the two-dimensional point cloud set of the cell image map and the two-dimensional point cloud set of the spatial transcriptome data map based on the point cloud registration algorithm to obtain the registration parameters includes: Get the initial values ​​of rotation parameters and translation parameters; Performing transformation processing on the two-dimensional point cloud set of the cell image based on the initial value of the rotation parameter and the initial value of the translation parameter; Obtaining corresponding points in the two-dimensional point cloud set of the cell image and the two-dimensional point cloud set of the spatial transcriptome data image; Updating the rotation parameter and the translation parameter by using an objective function, wherein the objective function is used to process the acquired corresponding points; Determine whether the preset conditions are met, and if so, obtain the rotation parameter and the translation parameter as the final value of the rotation parameter and the final value of the translation parameter, and use them as the registration parameters; If not, the rotation parameter and the translation parameter are updated to the initial values ​​of the rotation parameter and the initial values ​​of the translation parameter, and the step of transforming the two-dimensional point cloud set of the cell image based on the initial values ​​of the rotation parameter and the initial values ​​of the translation parameter is returned to.

3. The cell image registration method according to claim 1, It is characterized in that The step of respectively determining some target cells based on the clarity of the cell contours in the cell image and the spatial transcriptome data map comprises: Obtain a first score of the cell contour in the cell image, and obtain a second score of the cell contour in the spatial transcriptome data map, wherein the first score and the second score are positively correlated with the clarity; determining a portion of target cells according to the first score and the second score; The first score is calculated based on at least one of the eccentricity, pixel gray value and pixel area of ​​the cell in the cell image; The second score is calculated based on at least one of the amount of RNA in the cell and the pixel area in the spatial transcriptome data map.

4. The cell image registration method according to claim 3, It is characterized in that The part of target cells is: a preset proportion of cells or a preset number of cells selected according to the first score in the cell image; A preset proportion of cells or a preset number of cells are selected according to the second score in the spatial transcriptome data map.

5. The cell image registration method according to claim 3, It is characterized in that The first score is determined by the following formula: S 第一分值 =a / S x +b*S y +c*S z ; Among them, S x S is the cell eccentricity in the cell image. y is the average gray value of the pixels in the cell image, S z is the pixel area of ​​the cell in the cell image; the sum of a, b, and c is 1; The second score is determined by the following formula: S 第二分值 =m*S p +n*S q ; Among them, S p is the pixel area of ​​the cell in the spatial transcriptome data map, S q is the number of RNA in the cell in the spatial transcriptome data map; the sum of m and n is 1.

6. The cell image registration method according to claim 2, It is characterized in that The step of registering the cell image map and the spatial transcriptome data map according to the registration parameters comprises: Based on the registration parameters, transform the cell image; The registration of the cell image map and the spatial transcriptome data map is completed.

7. The cell image registration method according to claim 2, It is characterized in that The step of obtaining corresponding points in the two-dimensional point cloud set of the cell image and the two-dimensional point cloud set of the spatial transcriptome data image is: Calculate the Euclidean distance between each point in the two-dimensional point cloud set of the cell image graph and each point in the two-dimensional point cloud set of the spatial transcriptome data graph; The two points with the smallest Euclidean distance are taken as the corresponding points.

8. The cell image registration method according to claim 2, It is characterized in that The preset conditions are: The objective function is less than a convergence distance threshold; and / or, The number of updates reaches an update number threshold.

9. The cell image registration method according to claim 8, It is characterized in that The convergence distance threshold is positively correlated with the intercellular distances in the cell image map and the spatial transcriptome data map.

10. The cell image registration method according to claim 2, It is characterized in that The objective function is: Where D is the objective function, i is the number of updates, N is the number of point pairs of the corresponding points, and a i is a point in the two-dimensional point cloud of the cell image, b i The spatial transcriptome data map is a two-dimensional point cloud with a i The corresponding point of , R is the rotation parameter, and T is the translation parameter.

11. A cell image registration system, It is characterized in that The system comprises: An acquisition module, used for acquiring cell contours in a cell image map and a spatial transcriptome data map, wherein the cell image map and the spatial transcriptome data map have a corresponding relationship; A target cell determination module, used to determine part of the target cells based on the clarity of the cell contours in the cell image and the spatial transcriptome data map; An extraction module extracts the geometric centers of the target cells to form a two-dimensional point cloud set of a cell image and a two-dimensional point cloud set of a spatial transcriptome data map; A registration parameter acquisition module, used for processing the two-dimensional point cloud set of the cell image map and the two-dimensional point cloud set of the spatial transcriptome data map based on a point cloud registration algorithm to obtain registration parameters; A registration module is used to register the cell image and the spatial transcriptome data map according to the registration parameters.

12. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, the cell image registration method according to any one of claims 1 to 10 is implemented.

13. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the cell image registration method according to any one of claims 1 to 10 is implemented.

Images