Intelligent identification method, device and equipment for phase separation in cell fluorescence image

Abstract

The invention provides an intelligent identification method, device and equipment for phase separation in a cell fluorescence image. Aiming at the common problems of strong image noise, more background interference, high manual dependence, difficulty in dynamic process tracking and the like in the subcellular scale liquid-liquid phase separation process in living cells, spatial structure characteristics of suspected phase change points in an image are fully mined through image processing, noise interference is removed, and the dynamic process tracking is difficult. And accurate extraction of the dense region of the biological calibration object is realized. Besides, geometric coincidence analysis is carried out on the identified phase change areas in the image before and after illumination, the accurate corresponding relation of the area levels is constructed, and a reliable basis is provided for subsequent phase change degree quantitative trend analysis. In addition, the phase change degree of the biological calibration object in the image before and after illumination can be accurately calculated through a phase change degree index in combination with a region matching result. Therefore, automatic, accurate and efficient quantitative analysis of the phase separation area in the fluorescence microscopic image can be finally realized.

Description

Technical Field

[0001] The present application relates to an intelligent identification method, device and equipment for phase separation in cell fluorescence images, belonging to the field of computer vision and image processing technology. Background Art

[0002] Liquid-liquid phase separation is a widespread phenomenon in various cellular processes, particularly those involving membraneless organelles, and has become a frontier of research in the life sciences. Under normal conditions, cells can autonomously control the onset and disappearance of phase transitions, carrying out functions such as gene transcription, immune responses, and synaptic signaling. However, in some disease states (such as neurodegenerative diseases), phase separation can shift from a reversible dynamic process to an irreversible aggregation state, affecting the realization of physiological functions. Observing the phase separation process and outcomes of cells can help researchers observe the phase separation process under natural physiological conditions, the influence and regulation of cellular phase separation by environmental factors such as drugs and light, and explore the mechanisms of interaction between phase separation and biological functions.

[0003] Intrinsically disordered domains are generally considered the primary cause of liquid-liquid phase separation in cells. By combining light-sensitive proteins with intrinsically disordered domains, controlling phase separation through light, and thereby controlling cellular processes, cell state, and function, has become an important approach to studying the mechanisms and effects of phase separation. Observing the light-controlled phase separation process is crucial for precisely controlling the effects of light-controlled technologies.

[0004] Research on liquid-liquid phase separation is divided into in vitro and in vivo studies. In both types of studies, obtaining the state and dynamic process of phase transitions through fluorescent labeling and microscopy is the main means of studying phase transitions. In in vitro phase transition observations, the background noise is small, the environment is simple, and the phase transition droplets are obvious, making it easy to track and observe phase transitions. In vivo phase transition observations, on the other hand, face problems such as large differences between cells, large differences between different biological molecules, strong background noise, and small-scale movement of the phase change droplets themselves. The tracking and observation of phase transitions mainly rely on manual assisted identification. Currently, the identification and quantification of phase transitions mainly rely on manual threshold division and phase transition region identification. The process is cumbersome, time-consuming and labor-intensive. In addition, the differences between different cell types and proteins make manual identification extremely demanding on the judgment experience of researchers.

[0005] At the same time, the inherent variability of biological samples places high demands on sample quantity and processing efficiency. Therefore, there is an urgent need to develop efficient, automated phase separation image processing algorithms to replace traditional manual methods. Existing image processing technologies are limited by image quality and imaging conditions when processing intracellular phase separation images. First, phase separation is a rapidly changing dynamic process that typically requires time-series observation. However, the membraneless organelles in living cell samples inherently exhibit a certain degree of motion, which can hinder the identification of phase transition points. Existing image processing technologies can achieve basic recognition through edge detection and feature extraction, offering high computational efficiency, but struggle to eliminate dynamic noise beyond the phase separation point. Second, living biological samples may exhibit a certain amount of autofluorescence, while dead cells and impurities can generate strong nonspecific noise, which also hinders the observation of intracellular phase separation. The significant variability in nonspecific noise between samples reduces the generalization ability of existing recognition algorithms. Furthermore, in studies using short-wavelength light stimulation of cells, such as photoinduced phase separation, the light source itself can cause some quenching of fluorescence, resulting in either uniform or uneven decreases in fluorescence intensity in the image. Uneven fluorescence intensity reduction can affect manual thresholding, thereby interfering with phase transition identification under the final thresholding. These issues not only affect the accuracy of automated live cell phase separation observations but also increase the complexity of subsequent data processing, hindering the application and promotion of automated analysis technology. Summary of the Invention

[0006] To solve the problem that the phase transition in the existing subcellular scale liquid-liquid phase transition research in living cells relies on manual labor and is easily affected by complex background interferences such as autofluorescence, nonspecific noise, fluorescence quenching, etc., which makes it difficult to accurately segment the phase separation area. In addition, the dynamic changes such as the fusion and disappearance of phase transition droplets are rapid, making it difficult for traditional image processing methods to accurately track and quantify their timing processes. The present invention provides an intelligent recognition method, device and equipment for phase separation in cell fluorescence images. This solution fully considers the dynamics, background noise interference and other characteristics of liquid-liquid phase transition research, and realizes automatic, accurate and efficient quantification of phase separation areas in fluorescence microscopy images. Compared with traditional methods, it significantly improves recognition accuracy and processing efficiency, meeting the urgent needs of modern life sciences and drug research for automated monitoring and control of liquid-liquid phase transition processes.

[0007] In a first aspect, embodiments of the present application provide an intelligent method for identifying phase separation in cell fluorescence images, comprising the following steps: Collect image a of the biological cell before light stimulation and image b after light stimulation; Performing grayscale processing on the image a and the image b respectively to obtain corresponding single-channel grayscale images; The Otsu adaptive threshold segmentation method is used to perform binarization processing on the grayscale images corresponding to the image a and the image b, and an optimal grayscale segmentation threshold is determined to obtain a set of suspected phase change points P; Using the Delaunay triangulation algorithm to geometrically divide the suspected phase change point set P, construct a triangular mesh structure and generate a triangle index matrix, and calculate the circumscribed circle parameters of each triangle; According to the relationship between the area of ​​the circumscribed circle and the adjacency of the triangular structure, the phase transition points are merged to obtain the dense areas of biological calibration objects in image a and image b; Based on the Weiler-Atherton polygon clipping algorithm, the effective overlapping area of ​​the biological calibration object dense area of ​​image a and image b is determined to obtain the target biological calibration object dense area; The density of phase transition points per unit area in the target biological calibrant dense area is calculated as an index of the phase transition degree.

[0008] Based on the above method, optionally, the Otsu adaptive threshold segmentation method is used to perform binarization processing on the grayscale images corresponding to the image a and the image b, and an optimal grayscale segmentation threshold is determined to obtain a set of suspected phase change points P, including: By analyzing the grayscale histogram of the image, the pixels are divided into two categories: foreground and background. When traversing all possible thresholds, the pixel probability of foreground and background under each threshold and their grayscale mean are used to evaluate the difference between classes, and the optimal grayscale segmentation threshold is determined based on the difference between classes. Perform connected domain analysis on the binary image, set the connection mode to 8-neighborhood, identify all connected regions consisting of foreground pixels, and record the indexes of all pixels in each region; calculate the centroid coordinates of each connected region based on the connected domain label structure for positioning or subsequent analysis; For suspected phase transition points in small dot domains, their pixel coordinates are used directly; for suspected phase transition points with larger areas, the centroid coordinates obtained in the previous step are taken, and finally a suspected phase transition point set P containing all suspected phase transition points is constructed.

[0009] Based on the above method, optionally, the method of merging phase transition points based on the relationship between the circumscribed circle area and the triangular structure adjacency to obtain the biological calibration object dense area in image a and image b includes: Constructing a set of circumscribed circle areas , where each Indicates the The area of ​​the circumscribed circle corresponding to the triangle; Discretize the area value into multiple intervals, construct an area histogram, and count the area in each interval as the basis for calculating the optimal threshold value based on the maximum inter-class variance; Calculate the optimal classification threshold by using the maximum inter-class variance method , based on the optimal classification threshold Each triangle is divided into a dense structure with a smaller area and a sparse structure with a larger area. The formula of the maximum inter-class variance method is as follows:

[0010] in, is the between-class variance, are the area ratios of the smaller circumscribed circle and the larger circumscribed circle respectively; are the means of the two areas respectively; the maximum inter-class variance method searches in the interval from the smallest area to the largest area to maximize the inter-class variance. The area that finally maximizes the inter-class variance is the optimal classification threshold. ; According to the best classification threshold , retain satisfaction The triangular area corresponding to the circumscribed circle of is merged as the dense area for preliminary screening; For a triangle that is still in an unmerged state, if all three of its adjacent triangles have been determined to be dense areas, the triangle will be included in the dense area, and finally a complete dense area of ​​the biological calibration object will be obtained. ;in, is the set of all triangles obtained by Delaunay triangulation; Represents the triangle set that is determined to be a dense area after preliminary screening; Represents the set of adjacent triangles of a triangle.

[0011] Based on the above method, optionally, the method of determining the overlapping area of ​​the biological calibration object dense areas of image a and image b based on the Weiler-Atherton polygon clipping algorithm to obtain the target biological calibration object dense area includes: Extract the boundary edges of the dense area triangle set: Let the triangle set , where each triangle For each triangle, normalize its three edges into undirected edges in ascending order of vertex index. , and count the number of times each edge appears in all triangles. The edge that appears only once is the "boundary edge". After counting all edges, extract the edge set with a number of occurrences of 1 , which is the boundary edge set; Constructing a closed polygon outline: Obtaining a set of boundary edges Finally, these edges are connected in sequence to form one or more closed contour paths; Based on the closed polygonal contours corresponding to the image a and the image b, the Weiler-Atherton algorithm is used to determine the overlapping area of ​​the biological calibration object dense areas of the image a and the image b to obtain the target biological calibration object dense area.

[0012] Based on the above method, optionally, determining the effective overlapping area of ​​the biological calibration object dense areas of the image a and the image b using the Weiler-Atherton algorithm based on the closed polygonal contours corresponding to the image a and the image b to obtain the target biological calibration object dense area includes: Detection and All the intersection points between the boundaries are inserted into the original polygon vertex sequence in order. The updated polygon is recorded as and ;in, and are closed polygons of corresponding dense areas in image a and image b respectively; Mark each intersection point and record whether each intersection point "enters" or "leaves" another polygon; Traverse all closed polygons that have overlapping areas, starting from an "entry" intersection point, moving along the current polygon boundary, and switching to another polygon when encountering an intersection point to continue traversing until returning to the starting point; use this method to construct all overlapping closed areas; According to each pair of regions and , get the polygon area and , the intersection area of ​​the dense biological calibration areas in image a and image b is obtained by the Weiler-Atherton algorithm ; Calculate the overlap based on the following formula ,when , then determine the area in image a With the area in image b is the response area of ​​the same biomarker:

[0013] All response areas are determined as valid overlapping areas of the biological calibration object dense areas of the image a and the image b, to obtain a target biological calibration object dense area.

[0014] Based on the above method, optionally, calculating the density of phase transition points per unit area in the target biological calibrant dense region as a phase transition degree index includes: Extract the number of phase transition points in the region: Assume the region Represents a polygon in image b, consisting of a series of vertex coordinates Composition, for all phase transition points , to determine whether it falls on Internal; then get the area Internal phase transition point set ; Finally, calculate the number of phase transition points in the region ; Calculate the area of ​​the region: According to the following formula Calculate the area of ​​a region, where the points are arranged clockwise and the last point loops back to the starting point:

[0015] Where, For the region area; The phase transition point density per unit area is calculated using the following formula:

[0016] Using the same method, the phase transition point density per unit area in the corresponding region of image a is calculated:

[0017] If the corresponding areas in image a and image b satisfy , it is believed that this region has undergone a significant phase change response after light activation and belongs to the high phase change region, which characterizes the sensitivity or activation effect of the biological calibrant to external stimuli.

[0018] In a second aspect, an embodiment of the present application further provides an intelligent recognition device for phase separation in a cell fluorescence image, comprising: An acquisition module, used for acquiring an image a of the biological cell before light stimulation and an image b after light stimulation; A grayscale processing module, configured to perform grayscale processing on the image a and the image b respectively to obtain corresponding single-channel grayscale images; A binarization processing module is used to perform binarization processing on the grayscale images corresponding to the image a and the image b using the Otsu adaptive threshold segmentation method, determine the optimal grayscale segmentation threshold, and obtain a suspected phase change point set P; A region partitioning module is used to geometrically partition the suspected phase change point set P using a Delaunay triangulation algorithm, construct a triangular mesh structure and generate a triangle index matrix, and calculate the circumscribed circle parameters of each triangle; The dense area determination module is used to merge the phase transition points based on the circumscribed circle area and the adjacency relationship of the triangular structure to obtain the dense area of ​​the biological calibration object in image a and image b; An overlap region determination module is used to determine the effective overlap region of the biological calibration object dense regions of image a and image b based on the Weiler-Atherton polygon clipping algorithm, and obtain the target biological calibration object dense region; The index calculation module is used to calculate the phase change point density per unit area in the target biological calibrant dense area as the phase change degree index.

[0019] In a third aspect, an embodiment of the present application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor calls and executes the computer program, it implements the intelligent identification method for phase separation in cell fluorescence images as described in any one of the first aspects.

[0020] The technical solution of this application has the following beneficial effects: The intelligent identification method, device, and apparatus for phase separation in cell fluorescence images provided in this application address the common issues of high image noise, extensive background interference, high manual intervention, and difficulty in tracking dynamic processes during subcellular-scale liquid-liquid phase separation in living cells. This approach proposes an efficient and automated image processing and analysis process. Adaptive threshold segmentation based on the Otsu method is combined with Delaunay triangulation to explore the spatial structural features of suspected phase transition points in the image. Geometric merging rules are used to remove noise interference such as autofluorescence, nonspecific signals, and fluorescence quenching, enabling accurate extraction of densely populated areas of biological markers and significantly improving the stability and robustness of identification. Furthermore, the Weiler-Atherton polygon clipping algorithm is used to perform geometric coincidence analysis on the identified phase transition regions in the pre- and post-illumination images, establishing precise region-level correspondences and providing a reliable basis for subsequent quantitative trend analysis of the phase transition degree. Furthermore, the phase change point density per unit area (level change) metric, combined with region matching results, enables precise calculation of the phase change degree of biological markers in pre- and post-illumination images. By counting the changes in the number and distribution density of phase transition points, we can objectively evaluate the response efficiency of cells to light stimulation, and then determine the phase transition activity in different cells or under different conditions.

[0021] In addition, the present application supports the analysis of phase separation responses under different gradient illumination intensities. By acquiring the phase separation response area through image recognition and spatially comparing it with the original illumination area, it is possible to determine the optimal stimulation intensity required to induce 10%, 50%, and 90% of cells to produce phase separation under different light intensities, thereby achieving a high proportion of response under low energy conditions and improving the efficiency of light-controlled phase separation. In addition, the present application has good adaptability and is suitable for different types of cells, biomarkers, imaging methods, and light stimulation schemes. It can be widely used in many fields such as liquid-liquid phase separation mechanism research, drug screening, and neurodegenerative disease model establishment. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the description, serve to explain the principles of the present application. In addition, these drawings and the description are not intended to limit the scope of the concept of the present application in any way, but rather to illustrate the concept of the present application for those skilled in the art by reference to specific embodiments.

[0023] Figure 1 A schematic flow chart of an intelligent identification method for phase separation in cell fluorescence images provided in one embodiment of the present application; Figure 2 This is a schematic diagram of a cell fluorescence image collected in an embodiment of the present application before and after illumination, wherein: Figure 2 Figure (a) is a schematic diagram before illumination. Figure 2 Figure (b) is a schematic diagram after illumination; Figure 3 This is a schematic diagram of suspected phase transition points in a cell fluorescence image before and after illumination that has been binarized after threshold segmentation processing in one embodiment of the present application, wherein: Figure 3 Figure (a) is a schematic diagram before illumination. Figure 3 Figure (b) is a schematic diagram after illumination; Figure 4 This is a schematic diagram of a biological marker-dense area obtained through processing in one embodiment of the present application; Figure 5 This is a schematic diagram of the binarization and colorization of the high phase change region of the cell fluorescence image finally identified after light stimulation in one embodiment of the present application, wherein: Figure 5 Figure (a) is a binary diagram. Figure 5 Figure (b) is a color schematic diagram; Figure 6 This is a schematic structural diagram of an intelligent identification device for phase separation in cell fluorescence images according to one embodiment of the present application; Figure 7 This is a schematic structural diagram of an electronic device in one embodiment of the present application. DETAILED DESCRIPTION

[0024] To make the purpose, technical solutions and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0025] To address the common issues of high image noise, extensive background interference, high manual effort, and difficulty tracking dynamic processes during subcellular liquid-liquid phase separation within living cells, this application proposes an efficient and automated image processing and analysis process. The following non-limiting examples illustrate this implementation process through specific embodiments and examples.

[0026] Reference Figure 1 , Figure 1 This is a flow chart of an intelligent identification method for phase separation in cell fluorescence images provided in one embodiment of the present application. Figure 1 As shown, the method comprises at least the following steps: Step S101: collecting image a of the biological cell before light stimulation and image b after light stimulation.

[0027] Specifically, a photosensitive protein, a fluorescent marker protein, and an intrinsically disordered domain capable of undergoing phase transitions can be combined through gene sequence editing to create a photoinduced phase separation protein (OptoFUS). A plasmid containing the target gene is packaged via lentivirus and transfected into target cells, allowing the cells to express the fluorescently labeled photoinduced phase separation protein. Next, live cell images of the fluorescently labeled phase separation protein are captured using a laser confocal microscope, capturing one image before illumination and another after illumination, showing the phase transition. Figure 2 This is a schematic diagram of a cell fluorescence image collected in an embodiment of the present application before and after illumination, wherein: Figure 2 Figure (a) is a schematic diagram before illumination. Figure 2 Figure (b) is a schematic diagram after illumination.

[0028] In order to facilitate the distinction between images before and after light stimulation, all images of cells that have not been stimulated by light are marked as image a; and all images of cells that have undergone phase change after light stimulation are marked as image b.

[0029] Step S102: performing grayscale processing on image a and image b respectively to obtain corresponding single-channel grayscale images.

[0030] Specifically, since the original captured images (Image a and Image b) are RGB three-channel color images, the biomarker region during liquid-liquid phase separation primarily depends on the intensity distribution of the fluorescence signal. Therefore, in this step, the color images are converted to grayscale images to highlight variations in fluorescence signal intensity. This conversion process is well-known and will not be further described here.

[0031] Step S103: using the Otsu adaptive threshold segmentation method to perform binarization processing on the grayscale images corresponding to image a and image b, determine the optimal grayscale segmentation threshold, and obtain a set of suspected phase change points P.

[0032] Specifically, by binarizing the grayscale image, the resulting binary image can better highlight the suspected phase transition area. Among them, the Otsu algorithm is an adaptive threshold segmentation algorithm that can automatically determine the optimal threshold based on the grayscale distribution of the image, thereby achieving effective image segmentation.

[0033] In this solution, the Otsu algorithm is used to automatically determine the optimal segmentation threshold, dividing the image into two parts: suspected phase transition points (foreground) and background. Connected domain analysis is then performed on the set of suspected phase transition points obtained by segmentation. For connected domains with large volumes (pixel areas), only their geometric centers (circle centers) are retained as representative points, thus obtaining the final set of suspected phase transition points and eliminating redundancy or noise. Once the suspected phase transition points are found, the suspected phase transition point set is obtained. (For suspected phase transition points with larger volumes, the circle centers are used.) Figure 3 This is a schematic diagram of suspected phase transition points in a cell fluorescence image before and after illumination that has been binarized after threshold segmentation processing in one embodiment of the present application, wherein: Figure 3 Figure (a) is a schematic diagram before illumination. Figure 3 Figure (b) is a schematic diagram after illumination.

[0034] In some embodiments, in step S103, the grayscale images corresponding to image a and image b are binarized using the Otsu adaptive threshold segmentation method to determine the optimal grayscale segmentation threshold and obtain the suspected phase change point set P, which specifically includes: 1) By analyzing the grayscale histogram of the image, pixels are divided into two categories: foreground (suspected phase transition points) and background. When traversing all possible thresholds, the pixel probabilities of foreground and background at each threshold and their grayscale means are used to evaluate the inter-class differences. The optimal grayscale segmentation threshold is determined based on the inter-class differences. This process can achieve reliable binarization processing without manual intervention.

[0035] 2) Perform connected domain analysis on the binary image, setting the connectivity mode to 8-neighborhood, identifying all connected regions consisting of foreground pixels, and recording the indices of all pixels in each region. Based on the connected domain marker structure, calculate the centroid coordinates of each connected region for positioning or subsequent analysis. The centroid refers to the geometric center of the pixels in the region and serves as a representative point for positioning or subsequent analysis.

[0036] 3) For suspected phase transition points in the small dot domain, their pixel coordinates are used directly; for suspected phase transition points with larger areas, the centroid coordinates obtained in the previous step are used. Finally, a set P of suspected phase transition points containing all suspected phase transition points is constructed. In practice, an area threshold can be pre-set. Suspected phase transition points with an area less than or equal to the area threshold are considered suspected phase transition points in the small dot domain. Suspected phase transition points with an area greater than the area threshold are considered suspected phase transition points with larger areas and are represented by their centroid coordinates.

[0037] Through the above process, the grayscale images of image a and image b can be converted into binary images, and then the suspected phase transition point set P is determined based on the binary images, that is, the set of all regions that may be phase transition points.

[0038] Step S104: Use the Delaunay triangulation algorithm to geometrically divide the suspected phase change point set P, construct a triangular mesh structure and generate a triangle index matrix, and calculate the circumscribed circle parameters of each triangle.

[0039] The Delaunay triangulation algorithm is a computational geometry method used to divide a planar point set into a triangular mesh. Its core properties are satisfying the empty circle property (i.e., the circumcircle of any triangle contains no other points) and maximizing the minimum angle criterion (i.e., minimizing the appearance of "slender" triangles). Triangulated meshes constructed using the Delaunay triangulation algorithm have a more uniform structure and can better express the spatial distribution characteristics of a point set. From a geometric perspective, this algorithm is particularly suitable for describing the distribution density (distance description) and proximity relationships of biological markers.

[0040] In this embodiment, the Delaunay triangulation algorithm can automatically divide each suspected phase transition region in the suspected phase transition point set P into multiple triangle structures and generate a corresponding triangle index matrix. The triangle index matrix is ​​used to indicate which three points each triangle is composed of, facilitating search and processing in subsequent steps.

[0041] Specifically, the Delaunay triangulation algorithm is used to construct a triangular mesh for the suspected phase change point set P obtained in step 103. In the generated triangulated mesh, the length of each edge represents the spatial distance between adjacent points. The Delaunay triangulation algorithm divides the point set into a series of non-overlapping triangles based on the geometric relationship between the points, ensuring that the circumscribed circle of any triangle does not contain other points. This results in a triangle index matrix, in which each row indicates which three points constitute a triangle.

[0042] On this basis, the circumscribed circle parameters of each triangle are further calculated, including the center coordinates and radius.

[0043] Specifically, the corresponding circumcircle of the triangle obtained by Delaunay triangulation is obtained. , , , we can get the only point that makes the three points All are on the circle. This means that by using the circle center coordinate formula, the three vertices of the triangle can be used to obtain the center of its circumscribed circle. , which is expressed as follows:

[0044]

[0045]

[0046] By obtaining the center of the circle , choose any triangle vertex (such as point ) can be calculated to get the radius :

[0047] Step S105: According to the adjacency relationship between the circumscribed circle area and the triangular structure, the phase change points are merged to obtain the biological calibration object dense areas in image a and image b.

[0048] After completing the circumscribed circle construction in step 104, a set of Delaunay triangulated meshes with the phase transition points as vertices and their corresponding circumscribed circle parameters are obtained. To further extract areas where biological markers may be densely distributed, these triangles need to be analyzed and their structural properties determined to meet the "mergeability" criteria (based on the geometric characteristics of the densely distributed areas and the area of ​​the circumscribed circle).

[0049] To achieve automatic screening of circumscribed circles and avoid subjective bias caused by manually setting area thresholds, the circumscribed circle area of ​​each triangle is calculated, and then the large inter-class variance method is used to perform threshold segmentation on the circumscribed circle area. This distinguishes dense structures with smaller areas (less than or equal to the area threshold) from sparse structures with larger areas (greater than the area threshold). The sparse structures with larger areas are not dense areas of biological calibration objects and are therefore not merged.

[0050] Figure 4 This is a schematic diagram of a biological marker-dense area obtained after processing in one embodiment of the present application.

[0051] In some embodiments, step S105 combines phase transition points based on the area of ​​the circumscribed circle and the adjacency relationship between the triangular structures to obtain the biological calibration object dense areas in image a and image b, specifically including: 1) Construct a set of circumscribed circle areas , where each Indicates the The area of ​​the circumcircle corresponding to the triangle is Indicates the The radius of the circumcircle corresponding to the triangle; 2) Discretize the area values ​​into multiple intervals, construct an area histogram, and count the area in each interval as the basis for calculating the optimal threshold value based on the maximum inter-class variance; 3) Calculate the optimal classification threshold by the maximum inter-class variance method , based on the optimal classification threshold Each triangle is divided into a dense structure with a smaller area and a sparse structure with a larger area. The formula of the maximum inter-class variance method is as follows:

[0052] in, is the between-class variance, are the area ratios of the smaller circumscribed circle and the larger circumscribed circle respectively; are the means of the two types of areas respectively; the maximum inter-class variance method searches in the interval from the smallest area to the largest area so that the inter-class variance Reaching the maximum value, ultimately making the between-class variance The area that reaches the maximum value is the optimal classification threshold ; 4) According to the best classification threshold , retain satisfaction The triangular area corresponding to the circumscribed circle of is merged as the dense area for preliminary screening; 5) For a triangle that is still in an unmerged state, if all three of its adjacent triangles have been determined to be dense areas, the triangle will be included in the dense area, and finally a complete biological calibration dense area will be obtained. ;in, is the set of all triangles obtained by Delaunay triangulation; Represents the triangle set that is determined to be a dense area after preliminary screening; Represents the set of adjacent triangles of a triangle.

[0053] It should be noted that for a triangle that is still in an unmerged state, if all its three adjacent triangles have been determined to be dense areas, it is considered that the unmerged triangle may have been mistakenly excluded due to its slightly larger area or other local abnormal factors, but from the perspective of its spatial relationship, it should belong to the same dense area as the adjacent triangles. Therefore, the triangle is also included in the corresponding dense area.

[0054] The logic behind the merge operation is that if a triangle is completely surrounded by three merged triangles, then based on the spatial continuity and the natural clustering characteristics of biomarkers, the interior of the triangle is likely to have similar density properties to the adjacent areas and should be considered as part of the same dense area. Furthermore, from an observational perspective, biomarkers do not form hollow structures, so omissions of intermediate structures are likely errors and should be corrected.

[0055] After completing the triangulation of image a and image b, and the merge verification based on the circumscribed circle area threshold and spatial structure relationship, the triangle sets corresponding to the densely distributed areas of biological calibration objects in the two images are obtained, which are respectively denoted as and .

[0056] The geometric merging rules can effectively remove noise interference such as autofluorescence, nonspecific signals and fluorescence quenching, achieve accurate extraction of dense areas of biological calibration objects, and greatly improve the stability and robustness of recognition.

[0057] Step S106: Based on the Weiler-Atherton polygon clipping algorithm, determine the effective overlapping area of ​​the biological calibration object dense areas of image a and image b to obtain the target biological calibration object dense area.

[0058] Specifically, in step 105, the dense area triangle set of each biological calibration object in image a and image b has been obtained. and Each triangle in a set consists of three vertices, and the set can be geometrically represented as multiple connected regions. To calculate the geometric overlap of corresponding regions in the two images, we need to further extract the outer boundaries of these triangle sets and construct closed polygonal regions. The Weiler-Atherton algorithm is then used to accurately determine the overlap.

[0059] The Weiler-Atherton algorithm is a general-purpose graphics clipping algorithm applicable to arbitrarily complex polygons (such as convex polygons, concave polygons, and polygons with holes). It achieves precise interactive clipping between complex shapes by processing the directed relationship between polygon vertex sequences.

[0060] In this embodiment, the Weiler-Atherton polygon clipping algorithm is used to perform geometric coincidence analysis on the phase change regions identified in the images before and after illumination, so as to establish an accurate correspondence relationship at the regional level, providing a reliable basis for the subsequent quantitative trend analysis of the phase change degree.

[0061] In some embodiments, in step S106, based on the Weiler-Atherton polygon clipping algorithm, the overlapping area of ​​the biological calibration object dense areas of image a and image b is determined to obtain the target biological calibration object dense area, which specifically includes: 1) Extract the boundary edge of the triangle set in the dense area: Let the triangle set , where each triangle For each triangle, normalize its three edges into undirected edges in ascending order of vertex index. , and count the number of times each edge appears in all triangles. The edge that appears only once is the "boundary edge". After counting all edges, extract the edge set with a number of occurrences of 1 , which is the set of boundary edges.

[0062] 2) Construct a closed polygon outline: obtain the boundary edge set Finally, these edges are connected in sequence to form one or more closed contour paths.

[0063] First, the edge set is converted into an undirected graph structure, where each point is a vertex and each edge is a connecting line.

[0064] Then, randomly select a starting point and traverse the entire boundary according to the connection relationship until a closed loop is formed. If there are multiple connected subgraphs in the graph, construct multiple independent closed contours for each.

[0065] Finally, the boundary path obtained is an ordered set of points , represents the closed polygonal outline corresponding to the dense area; the closed polygonal outline truly reflects the geometric coverage of the area in the image.

[0066] 3) Based on the closed polygonal contours corresponding to images a and b, the Weiler-Atherton algorithm is used to determine the overlapping area of ​​the biomarker-dense areas of images a and b to obtain the target biomarker-dense area.

[0067] Specifically, let the closed polygon of a corresponding dense area in image a and image b be and , then in the above steps, the judgment steps of "based on the closed polygon contours corresponding to images a and b, using the Weiler-Atherton algorithm to determine the overlapping area of ​​the biological calibration object dense areas of images a and b, and obtaining the target biological calibration object dense area" are as follows: First, detect and All the intersection points between the boundaries are inserted into the original polygon vertex sequence in order. The updated polygon is recorded as and ;in, and are closed polygons of corresponding dense areas in image a and image b respectively; Then, mark each intersection point and record whether each intersection point "enters" or "exits" another polygon; Then, traverse all closed polygons with overlapping areas, starting from an "entry" intersection point, move along the current polygon boundary, and switch to another polygon when encountering an intersection point and continue traversing until returning to the starting point; use this method to construct all overlapping closed areas; Then, according to each pair of regions and , get the polygon area and , the intersection area of ​​the dense biological calibration areas in image a and image b is obtained by the Weiler-Atherton algorithm ; Calculate the overlap based on the following formula ,when , then determine the area in image a With the area in image b are the response areas of the same biological calibrant, that is, the two are successfully matched (in practice, a and b are often highly matched):

[0068] in, represents the union area of ​​the dense biological calibration areas in image a and image b; Finally, all response areas are determined as the effective overlapping areas of the bio-calibration dense areas of image a and image b, and the target bio-calibration dense area is obtained, that is, the effective bio-calibration dense area finally identified after eliminating noise and interference.

[0069] Figure 5 This is a schematic diagram of the binarization and colorization of the high phase change region of the cell fluorescence image finally identified after light stimulation in one embodiment of the present application, wherein: Figure 5 Figure (a) is a binary diagram. Figure 5 Figure (b) is a color schematic diagram.

[0070] Step S107: Calculate the phase transition point density per unit area in the target biological marker dense area as a phase transition degree index.

[0071] After determining the target biological calibration object dense area, the phase change degree index is obtained by calculating the phase change point density per unit area, so as to evaluate the response intensity change of the biological calibration object in the area before and after light activation. , here you only need to select one representative (such as , that is, the area that may change after illumination) is calculated.

[0072] In some embodiments, the region in image b is For example, in step S107, the density of phase transition points per unit area in the target biological calibrant dense area is calculated as the phase transition degree index, specifically including: 1) Extract the number of phase transition points in the region: Assume the region Represents a polygon in image b, consisting of a series of vertex coordinates Composition, for all phase transition points , to determine whether it falls on Internal; then get the area Internal phase transition point set ; Finally, calculate the number of phase transition points in the region ; 2) Calculate the area of ​​the region: Use the following formula to calculate the area Calculate the area of ​​a region, where the points are arranged clockwise and the last point loops back to the starting point:

[0073] Where, For the region area; 3) Calculate the phase transition point density per unit area using the following formula:

[0074] 4) Using the same method, calculate the phase transition point density per unit area of ​​the corresponding region in image a:

[0075] Based on the above indicators, if the corresponding areas in image a and image b meet , it is considered that the region has undergone a significant phase change response after light activation (the phase change degree in image b is significantly increased or decreased compared with the region at the same position in image a), and belongs to the high phase change region, which represents the sensitivity or activation effect of the biological calibration object to external stimuli.

[0076] The technical solution of this application has the following beneficial effects: This application uses image processing technology to quickly calculate the cell areas that have phase separation within the microscope field of view, avoiding the time-consuming process of traditional manual phase separation identification. Combining computer vision and image processing algorithms can achieve automated analysis and reduce human error.

[0077] The recognition results of the present application may also include the phase separation reaction efficiency of the cells to light; the phase separation images of the cells before and after light are recognized to obtain recognition results, and the cells that produce phase separation are determined based on the phase separation recognition results; the phase separation reaction efficiency of the cells to light is determined based on the comparison between the phase-separated cells and the original light stimulation area.

[0078] This application can also determine reasonable conditions for photoinduced phase separation of cells. For example, the original light is subjected to a gradient intensity design of 1%, 5%, 10%, 20%, 50%, and 100%. The phase separation response area is obtained through image processing technology. By comparing the response area with the original light, it is possible to determine the light conditions that cause 10%, 50%, and 90% of the illuminated cells to produce phase separation. Then, under lower light stimulation, the effect of causing more than 90% of cells to produce phase separation is achieved, thereby improving the efficiency of light-controlled phase separation.

[0079] The present application can also determine the original illumination position. Cells are stimulated under the aforementioned illumination conditions that cause phase separation in more than 90% of cells. Image processing techniques are used to determine the phase separation response region and the original illumination position of the cell phase transition response. Cells can then be used as a negative for a photosensitive imaging device.

[0080] In addition, the present invention provides an intelligent recognition device for phase separation in cell fluorescence images, such as Figure 6 As shown, the device includes: An acquisition module 21 is used to acquire an image a of the biological cell before light stimulation and an image b after light stimulation; A grayscale processing module 22 is used to perform grayscale processing on the image a and the image b respectively to obtain corresponding single-channel grayscale images; A binarization processing module 23 is configured to perform binarization processing on the grayscale images corresponding to the image a and the image b using the Otsu adaptive threshold segmentation method, determine an optimal grayscale segmentation threshold, and obtain a set of suspected phase change points P; A region partitioning module 24 is configured to geometrically partition the suspected phase change point set P using a Delaunay triangulation algorithm, construct a triangular mesh structure, generate a triangle index matrix, and calculate the circumscribed circle parameters of each triangle; A dense area determination module 25 is used to merge phase transition points based on the area of ​​the circumscribed circle and the adjacency relationship of the triangular structure to obtain the dense area of ​​the biological calibration object in image a and image b; The overlap region determination module 26 is configured to determine the effective overlap region of the biological calibration object dense regions of the image a and the image b based on the Weiler-Atherton polygon clipping algorithm, and obtain the target biological calibration object dense region; The index calculation module 27 is used to calculate the phase change point density per unit area in the target biological calibrant dense area as a phase change degree index.

[0081] The specific implementation process of each functional module of the above-mentioned device can refer to the corresponding content in the above-mentioned method embodiment, and will not be repeated here.

[0082] In addition, an embodiment of the present application provides an electronic device, such as Figure 7 As shown, the electronic device includes a memory 31 and a processor 32; wherein the memory 31 stores a computer program, and when the processor 32 calls and executes the computer program, it implements the intelligent recognition method for phase separation in cell fluorescence images in any of the above embodiments.

[0083] The electronic device may be a server or a computer.

[0084] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0085] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" is at least two.

[0086] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0087] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0088] Those skilled in the art will understand that all or part of the steps carried out in the method for implementing the above-mentioned embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0089] Furthermore, the functional units in the various embodiments of the present invention may be integrated into a single processing module, each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in either hardware or software functional modules. If the integrated modules are implemented as software functional modules and sold or used as standalone products, they may also be stored in a computer-readable storage medium. The aforementioned storage medium may be a read-only memory, a magnetic disk, or an optical disk, etc.

[0090] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0091] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An intelligent identification method for phase separation in cell fluorescence images, characterized in that: The steps include: Collect image a of the biological cell before light stimulation and image b after light stimulation; Performing grayscale processing on the image a and the image b respectively to obtain corresponding single-channel grayscale images; The Otsu adaptive threshold segmentation method is used to perform binarization processing on the grayscale images corresponding to the image a and the image b, and an optimal grayscale segmentation threshold is determined to obtain a set of suspected phase change points P; Using the Delaunay triangulation algorithm to geometrically divide the suspected phase change point set P, construct a triangular mesh structure and generate a triangle index matrix, and calculate the circumscribed circle parameters of each triangle; According to the relationship between the area of ​​the circumscribed circle and the adjacency of the triangular structure, the phase transition points are merged to obtain the dense areas of biological calibration objects in image a and image b; Based on the Weiler-Atherton polygon clipping algorithm, the effective overlapping area of ​​the biological calibration object dense area of ​​image a and image b is determined to obtain the target biological calibration object dense area; The density of phase transition points per unit area in the target biological calibrant dense area is calculated as an index of the phase transition degree.

2. The method according to claim 1, characterized in that The Otsu adaptive threshold segmentation method is used to perform binarization processing on the grayscale images corresponding to the image a and the image b, and an optimal grayscale segmentation threshold is determined to obtain a set of suspected phase change points P, including: By analyzing the grayscale histogram of the image, the pixels are divided into two categories: foreground and background. When traversing all possible thresholds, the pixel probability of foreground and background under each threshold and their grayscale mean are used to evaluate the difference between classes, and the optimal grayscale segmentation threshold is determined based on the difference between classes. Perform connected domain analysis on the binary image, set the connection mode to 8-neighborhood, identify all connected regions consisting of foreground pixels, and record the indexes of all pixels in each region; calculate the centroid coordinates of each connected region based on the connected domain label structure for positioning or subsequent analysis; For suspected phase transition points in small dot domains, their pixel coordinates are used directly; for suspected phase transition points with larger areas, the centroid coordinates obtained in the previous step are taken, and finally a suspected phase transition point set P containing all suspected phase transition points is constructed.

3. The method according to claim 1, characterized in that The method of merging phase transition points based on the circumscribed circle area and the adjacency relationship between the triangular structures to obtain the biological calibration object dense areas in image a and image b includes: Constructing a set of circumscribed circle areas , where each Indicates the The area of ​​the circumscribed circle corresponding to the triangle; Discretize the area value into multiple intervals, construct an area histogram, and count the area in each interval as the basis for calculating the optimal threshold value based on the maximum inter-class variance; Calculate the optimal classification threshold by using the maximum inter-class variance method , based on the optimal classification threshold Each triangle is divided into a dense structure with a smaller area and a sparse structure with a larger area. The formula of the maximum inter-class variance method is as follows: in, is the between-class variance, are the area ratios of the smaller circumscribed circle and the larger circumscribed circle respectively; are the means of the two areas respectively; the maximum inter-class variance method searches in the interval from the smallest area to the largest area to maximize the inter-class variance. The area that finally maximizes the inter-class variance is the optimal classification threshold. ; According to the best classification threshold , retain satisfaction The triangular area corresponding to the circumscribed circle of is merged as the dense area for preliminary screening; For a triangle that is still in an unmerged state, if all three of its adjacent triangles have been determined to be dense areas, the triangle will be included in the dense area, and finally a complete dense area of ​​the biological calibration object will be obtained. ;in, is the set of all triangles obtained by Delaunay triangulation; Represents the triangle set that is determined to be a dense area after preliminary screening; Represents the set of adjacent triangles of a triangle.

4. The method according to claim 1, wherein The method of determining the overlapping area of ​​the biological calibration object dense areas of image a and image b based on the Weiler-Atherton polygon clipping algorithm to obtain the target biological calibration object dense area includes: Extract the boundary edges of the dense area triangle set: Let the triangle set , where each triangle For each triangle, normalize its three edges into undirected edges in ascending order of vertex index. , and count the number of times each edge appears in all triangles. The edge that appears only once is the "boundary edge". After counting all edges, extract the set of edges with a number of appearances of 1 , which is the boundary edge set; Constructing a closed polygon outline: Obtaining a set of boundary edges Finally, these edges are connected in sequence to form one or more closed contour paths; Based on the closed polygonal contours corresponding to the image a and the image b, the Weiler-Atherton algorithm is used to determine the overlapping area of ​​the biological calibration object dense areas of the image a and the image b to obtain the target biological calibration object dense area.

5. The method according to claim 4, characterized in that The method of determining the effective overlapping area of ​​the biological calibration object dense areas of the image a and the image b using the Weiler-Atherton algorithm based on the closed polygonal contours corresponding to the image a and the image b to obtain the target biological calibration object dense area includes: Detection and All the intersection points between the boundaries are inserted into the original polygon vertex sequence in order. The updated polygon is recorded as and ;in, and are closed polygons of corresponding dense areas in image a and image b respectively; Mark each intersection point and record whether it "enters" or "leaves" another polygon; Traverse all closed polygons that have overlapping areas, starting from an "entry" intersection point, moving along the current polygon boundary, and switching to another polygon when encountering an intersection point to continue traversing until returning to the starting point; use this method to construct all overlapping closed areas; According to each pair of regions and , get the polygon area and , the intersection area of ​​the dense biological calibration areas in image a and image b is obtained by the Weiler-Atherton algorithm ; Calculate the overlap based on the following formula ,when , then determine the area in image a With the area in image b is the response area of ​​the same biomarker: All response areas are determined as valid overlapping areas of the biological calibration object dense areas of the image a and the image b, to obtain a target biological calibration object dense area.

6. The method according to claim 1, characterized in that Calculating the density of phase transition points per unit area in the target biological calibrant dense region as a phase transition degree indicator includes: Extract the number of phase transition points in the region: Assume the region Represents a polygon in image b, consisting of a series of vertex coordinates Composition, for all phase transition points , to determine whether it falls on Internal; then get the area Internal phase transition point set ; Finally, calculate the number of phase transition points in the region ; Calculate the area of ​​the region: According to the following formula Calculate the area of ​​a region, where the points are arranged clockwise and the last point loops back to the starting point: Where, For the region area; The phase transition point density per unit area is calculated using the following formula: Using the same method, the phase transition point density per unit area in the corresponding region of image a is calculated: If the corresponding areas in image a and image b satisfy , it is believed that this region has undergone a significant phase change response after light activation and belongs to the high phase change region, which characterizes the sensitivity or activation effect of the biological calibrant to external stimuli.

7. An intelligent recognition device for phase separation in cell fluorescence images, characterized in that: include: An acquisition module, used for acquiring an image a of the biological cell before light stimulation and an image b after light stimulation; A grayscale processing module, configured to perform grayscale processing on the image a and the image b respectively to obtain corresponding single-channel grayscale images; A binarization processing module is used to perform binarization processing on the grayscale images corresponding to the image a and the image b using the Otsu adaptive threshold segmentation method, determine the optimal grayscale segmentation threshold, and obtain a suspected phase change point set P; A region partitioning module is used to geometrically partition the suspected phase change point set P using a Delaunay triangulation algorithm, construct a triangular mesh structure and generate a triangle index matrix, and calculate the circumscribed circle parameters of each triangle; The dense area determination module is used to merge the phase transition points based on the circumscribed circle area and the adjacency relationship of the triangular structure to obtain the dense area of ​​the biological calibration object in image a and image b; An overlap region determination module is used to determine the effective overlap region of the biological calibration object dense regions of image a and image b based on the Weiler-Atherton polygon clipping algorithm, and obtain the target biological calibration object dense region; The index calculation module is used to calculate the phase change point density per unit area in the target biological calibrant dense area as the phase change degree index.

8. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor calls and executes the computer program, the method realizes the intelligent recognition method for phase separation in cell fluorescence images according to any one of claims 1 to 6.

Images