A medical microscopic cell image recognition method, device, equipment and medium
By using image recognition technology for grayscale conversion, threshold segmentation, and unsupervised clustering, cell feature points are automatically extracted and clustered, solving the problems of low efficiency and reliance on manual operation in traditional microscopic cell image processing, and achieving efficient and accurate cell counting and state judgment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JILIN UNIVERSITY
- Filing Date
- 2026-04-21
- Publication Date
- 2026-07-14
AI Technical Summary
The acquisition and processing of traditional medical microscopic cell fluorescence images rely on manual operation, which has problems such as high technical requirements, low efficiency, reliance on experience, and poor analysis accuracy due to visual fatigue. In particular, it is difficult to quantify in high-throughput screening and the judgment of subtle changes.
Using image recognition technology, cell feature points and clusters are automatically extracted through grayscale conversion, threshold segmentation, unsupervised clustering, and fuzzy C-means clustering, and the cell medical status is calculated to achieve standardization of cell counting and status judgment.
It improves detection efficiency, reduces human interference, ensures the standardization and repeatability of identification results, and can quickly process a large number of images and obtain accurate cell analysis results.
Smart Images
Figure CN122392053A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image recognition technology, and in particular to a method, apparatus, device, and medium for medical microscopic cell image recognition. Background Technology
[0002] In biomedical, life science research, and clinical diagnosis, medical fluorescence microscopy has become a core research tool. Essentially, it combines fluorescence labeling technology with microscopic imaging technology to address the needs of "specific identification, microscopic localization, and quantitative analysis" that traditional microscopic imaging (bright-field imaging) cannot meet. However, there are certain drawbacks in the acquisition and processing of medical fluorescence microscopy images. First, traditional fluorescence microscopy involves manual observation of cells under dark conditions. Each observation of cells within a field of view requires rapid selection, comparison, and photographing to prevent rapid quenching of fluorescence under light, demanding a high level of skill and expertise from the operator. Second, manually analyzing a single medium-resolution fluorescence image (low magnification) takes 10-20 minutes, which is insufficient for high-throughput screening of thousands of images simultaneously. Third, the judgment of cell fluorescence intensity, morphological boundaries, and aggregation states is highly dependent on the operator's experience, leading to labeling discrepancies. Human intervention can influence the definition of fluorescence intensity and boundaries, as well as the assessment and separation of cell overlap, resulting in biased analytical results. Fourth, weak fluorescence signals and subtle changes in cell morphology (e.g., shrunken nuclei, abnormal organelle distribution) are difficult to discern with the naked eye and even more challenging to quantify and analyze. Fifth, prolonged and repeated observation of fluorescence micrographs can lead to visual fatigue in the operator, further reducing analytical accuracy and causing missed or incorrect diagnoses. Summary of the Invention
[0003] In view of this, the purpose of this invention is to provide a method, apparatus, device, and medium for medical microscopic cell image recognition, capable of simultaneously and efficiently processing multiple microscopic cell images and rapidly obtaining accurate cell image analysis results. The specific solution is as follows: In a first aspect, this application discloses a method for medical microscopic cell image recognition, comprising: Obtain the target grayscale image after image coloring and image preprocessing; The target grayscale image is subjected to threshold segmentation, and then the image pixels representing cells in the target grayscale image are extracted to obtain cell feature points; The cell feature points are grouped using an unsupervised clustering method, and the number of clusters after grouping is taken as the total number of cells in the target grayscale image. The mean and uniformity information of the pixel coordinate feature values of all cell feature points in each subgroup are calculated to obtain the center pixel coordinates and cell medical status of each cell.
[0004] Optionally, obtaining the target grayscale image after image coloring and image preprocessing includes: Acquire fluorescence images of stained medical microscopy cells; The medical microscopic cell fluorescence image is converted to grayscale to obtain the target grayscale image.
[0005] Optionally, the step of performing threshold segmentation on the target grayscale image and then extracting image pixels representing cells from the target grayscale image to obtain cell feature points includes: Iterate through the image grayscale values of each pixel in the target grayscale image, and select the image pixels whose image grayscale values are greater than or equal to the candidate image grayscale threshold as foreground pixels; Image pixels whose grayscale values are less than the candidate image grayscale threshold are designated as background pixels. The foreground pixel ratio is determined based on the total number of foreground pixels and the total number of pixels in the target grayscale image; The background pixel ratio is determined based on the total number of background pixels and the total number of pixels in the target grayscale image; Based on the foreground pixel ratio, the background pixel ratio, the average gray level of the foreground, and the average gray level of the background, an inter-class variance is constructed. The target image gray level threshold that maximizes the inter-class variance is then selected. Target image pixels with gray levels greater than the target image gray level threshold are then used as cell feature points.
[0006] Optionally, after determining the target image pixels with gray values greater than the target image gray value threshold as cell feature points, the method further includes: Obtain the pixel coordinates and grayscale values of the target image pixels, and construct pixel coordinate feature values corresponding to cell feature points based on the pixel coordinates and grayscale values.
[0007] Optionally, the step of using an unsupervised clustering method to cluster the cell feature points, so as to use the number of clusters as the total number of cells in the target grayscale image, includes: The fuzzy C-means clustering method is used to divide the cell feature points in the initial population into two candidate clusters; wherein, the initial population is constructed based on all cell feature points to be clustered. Calculate the dispersion of each candidate cluster, and take the candidate clusters with dispersion greater than a preset dispersion threshold as new initial clusters, and jump to the step of dividing the cell feature points in the initial cluster into two candidate clusters using the fuzzy C-means clustering method, until the dispersion of all candidate clusters is less than or equal to the preset dispersion threshold. Obtain the current number of clusters after clustering, and use the number of clusters as the total number of cells in the target grayscale image.
[0008] Optionally, the uniformity information of the pixel coordinate feature values of all cell feature points in each cluster is calculated to obtain the cellular medical state of each cell, including: Calculate the Euclidean distance between all pairs of cell feature points in each cluster, and calculate the corresponding distance standard deviation based on the Euclidean distance and the mean distance to obtain the uniformity information of the cluster. If the uniformity information is less than or equal to a preset uniformity threshold, the cell's cellular medical state is determined to be a viable state. If the uniformity information is greater than a preset uniformity threshold, the cell is determined to be in a dead state in the cellular medicine context.
[0009] Optionally, the medical microscopic cell image recognition method further includes: The center pixel coordinates of cells identified as viable are marked on the medical microscopic cell fluorescence image using a first preset marker symbol. The center pixel coordinates of cells determined to be dead are marked on the medical microscopic cell fluorescence image by using a second preset marker symbol. The output includes cell count statistics and image files with cell status markers.
[0010] Secondly, this application discloses a medical microscopic cell image recognition device, comprising: The image acquisition module is used to acquire the target grayscale image after image coloring and image preprocessing; Feature point extraction processing is used to perform threshold segmentation on the target grayscale image, and then extract the image pixels representing cells in the target grayscale image to obtain cell feature points; Clustering is used to cluster the cell feature points using an unsupervised clustering method, so that the number of clusters after clustering is taken as the total number of cells in the target grayscale image. The state recognition process is used to calculate the mean and uniformity information of the pixel coordinate feature values of all cell feature points in each subgroup, so as to obtain the center point pixel coordinates and cell medical state of each cell respectively.
[0011] Thirdly, this application discloses an electronic device, including: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the aforementioned disclosed medical microscopic cell image recognition method.
[0012] Fourthly, this application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the steps of the aforementioned disclosed medical microscopic cell image recognition method.
[0013] Therefore, this application discloses a method for acquiring a target grayscale image after image staining and image preprocessing; performing threshold segmentation on the target grayscale image; extracting image pixels representing cells from the target grayscale image to obtain cell feature points; using an unsupervised clustering method to cluster the cell feature points, with the number of clusters being used as the total number of cells in the target grayscale image; and calculating the mean and uniformity information of the pixel coordinate feature values of all cell feature points in each cluster to obtain the center point pixel coordinates and cell medical status of each cell. Thus, by using an unsupervised clustering method, the number of clusters is automatically determined, meaning that the operator does not need extensive experience to judge whether cells are adhered or have consistent morphology; the division is automatically completed, eliminating interference from human experience. Simultaneously, the extraction of cell feature points and unsupervised clustering, compared to traditional methods based on connected component counting, provides stronger stability and more accurate counting for cell overlap and adhesion. Finally, calculating uniformity information to obtain the cell medical status achieves standardized quantification of the cell medical status judgment process, making the recognition results standardized and repeatable. Attached Figure Description
[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0015] Figure 1 This is a flowchart of a medical microscopic cell image recognition method disclosed in this application; Figure 2 This is a flowchart of a thresholding method for images based on the maximum inter-class variance method disclosed in this application; Figure 3 This is a flowchart of an improved fuzzy C-means clustering method for clustering cell feature points disclosed in this application; Figure 4 This is a flowchart of a method for calculating the mean and uniformity of pixel coordinates of clusters of each cell feature point, as disclosed in this application. Figure 5 This is a diagram showing the identification results of the central location and cell viability status of each cell as disclosed in this application; Figure 6This is a flowchart of a specific medical microscopic cell image recognition method disclosed in this application; Figure 7 This is a schematic diagram of the structure of a medical microscopic cell image recognition device disclosed in this application; Figure 8 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation
[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0017] In biomedical, life science research, and clinical diagnosis, medical fluorescence microscopy has become a core research tool. Essentially, it combines fluorescence labeling technology with microscopic imaging technology to address the needs of "specific identification, microscopic localization, and quantitative analysis" that traditional microscopic imaging (such as bright-field imaging) cannot meet. Target biomolecules or structures are labeled with specific fluorescent probes, and the probes emit fluorescence at specific wavelengths when illuminated with excitation light. The fluorescence signal is then captured by a microscope, achieving both "visualization" and "quantification" of the target. Fluorescence imaging has moved from "macroscopic morphological observation" to "molecular mechanism analysis," upgrading from "qualitative description" to "precise quantitative analysis," and expanding from "static sample research" to "dynamic tracking of live samples," becoming a crucial bridge connecting the "molecular level" and the "cellular or tissue level," and "basic research" and "clinical application." Therefore, whether it's mechanism exploration in basic medical research, high-throughput screening in drug development, or precise typing in clinical diagnosis and data support in digital twins, fluorescence imaging, with its irreplaceable technological advantages, has become an indispensable core research tool in the biomedical field.
[0018] However, there are certain drawbacks in the acquisition and processing of medical microscopic cell fluorescence images. First, traditional fluorescence microscopy involves manual observation of cells under dark conditions. Each observation of cells within a field of view requires rapid selection, comparison, and photographing to prevent fluorescence quenching under light, demanding a high level of skill and expertise from the operator. Second, manually analyzing a single medium-resolution fluorescence image (low magnification) takes 10-20 minutes, insufficient for the high-throughput screening of thousands of images simultaneously. Third, the judgment of cell fluorescence intensity, morphological boundaries, and aggregation states is highly dependent on the operator's experience, leading to annotation discrepancies. Human intervention can influence the definition of fluorescence intensity, boundary determination, and the assessment and separation of overlapping cells, resulting in biased analysis results. Fourth, weak fluorescence signals and subtle changes in cell morphology (e.g., nuclear shrinkage, abnormal organelle distribution) are difficult to discern with the naked eye and even more challenging to quantify. Fifth, prolonged and repeated observation of fluorescence microscopic images can cause visual fatigue, further reducing analytical accuracy and leading to missed or incorrect diagnoses.
[0019] Therefore, this invention provides a medical microscopic cell image recognition scheme that overcomes the shortcomings of manual detection methods, achieves improved detection efficiency with lower hardware costs, and ensures better detection results.
[0020] like Figure 1 As shown, the present invention provides a method for medical microscopic cell image recognition, comprising: Step S11: Obtain the target grayscale image after image coloring and image preprocessing.
[0021] In this embodiment, a stained medical microscopic cell fluorescence image is acquired; the medical microscopic cell fluorescence image is then converted to grayscale to obtain a target grayscale image. It is understood that medical microscopic cell fluorescence images are typically color images, with color information primarily derived from fluorescent probes of specific wavelengths, used to indicate the structure or activity state of target cells. However, during thresholding and feature extraction, the grayscale levels (brightness information) of the image can characterize the contrast difference between cells and the background, while color information not only increases computational complexity but also introduces redundant interference. Therefore, by converting to grayscale, the RGB three-channel image is compressed into a single-channel grayscale image, effectively reducing data dimensionality and improving processing efficiency while preserving the cell fluorescence intensity characteristics. This target grayscale image serves as the input for all subsequent recognition and clustering operations.
[0022] Step S12: Perform threshold segmentation on the target grayscale image, and then extract the image pixels representing cells in the target grayscale image to obtain cell feature points.
[0023] In this embodiment, the grayscale values of each pixel in the target grayscale image are traversed, and pixels with grayscale values greater than or equal to a candidate image grayscale threshold are designated as foreground pixels; pixels with grayscale values less than the candidate image grayscale threshold are designated as background pixels; the foreground pixel ratio is determined based on the total number of foreground pixels and the total number of pixels in the target grayscale image; the background pixel ratio is determined based on the total number of background pixels and the total number of pixels in the target grayscale image; an inter-class variance is constructed based on the foreground pixel ratio, the background pixel ratio, the average foreground grayscale, and the average background grayscale, and a target image grayscale threshold that maximizes the inter-class variance is selected. Then, target image pixels with grayscale values greater than the target image grayscale threshold are designated as cell feature points. It can be understood that the maximum inter-class variance method is used to perform threshold segmentation on the target grayscale image to obtain cell feature points in the image, such as... Figure 2 As shown, the grayscale threshold of the target image that maximizes the inter-class variance is determined by traversing the grayscale values of the image, thereby extracting cell feature points that are greater than the grayscale threshold of the target image. The specific process is as follows: For a candidate image grayscale threshold of the target grayscale image T Calculate which image pixel in the target grayscale image belongs to the foreground (grayscale value ≥ 0.5). T The proportion of pixels in the entire image. ω 0 and belong to the background (grayscale value less than 0) T The proportion of pixels in the entire image. ω 1. The calculation method is as follows: ω 0= N 0 / ( M × N ); ω 1= N 1 / ( M × N ); in, M × N This represents the image resolution (total number of pixels). N 0 For image grayscale values greater than or equal to a threshold T The number of pixels, N 1 For image grayscale values less than a threshold T The number of pixels.
[0024] Furthermore, regarding this candidate grayscale threshold... T Calculate the average gray value of the foreground of the image respectively. μ 0 and average gray value of image background μ 1. The calculation method is as follows: ; ; in, P i For pixel grayscale values greater than or equal to T The image grayscale value of the pixel, i.e. i =1,2,…, N 0 and P i ≥ T ; Q j For pixel grayscale less than T The image grayscale value of the pixel, i.e. j =1,2,…, N 1 and Q j < T .
[0025] For this grayscale value T Based on the foreground pixel ratio, background pixel ratio, foreground average gray level, and background average gray level obtained above, the maximum inter-class variance of the image is constructed. G The calculation method is as follows: G = ω 0× ω 1×( μ 0- μ 1) 2 ; Iterate through the entire image's grayscale values, when G Maximum grayscale value T That is, the grayscale threshold of the target image. T optima , grayscale values greater than T optimal The pixels were identified as cell feature points.
[0026] In this embodiment, the pixel coordinates and grayscale values of the target image pixels are obtained, and pixel coordinate feature values corresponding to the cell feature points are constructed based on the pixel coordinates and grayscale values. This can be understood as follows: for a given cell feature point, the pixel coordinates of each cell feature point are obtained. Taking the k-th cell feature point as an example, its pixel coordinates are (…). P uk , P vk At the same time, its grayscale value is obtained. T k Then, based on pixel coordinates and grayscale values, a three-dimensional feature vector corresponding to the cell feature points is constructed, which is the pixel coordinate feature value. .
[0027] Step S13: Use an unsupervised clustering method to cluster the cell feature points, and use the number of clusters after clustering as the total number of cells in the target grayscale image.
[0028] In this embodiment, the fuzzy C-means clustering method is used to divide the cell feature points in the initial population into two candidate clusters. The initial population is constructed based on all cell feature points to be clustered. The dispersion of each candidate cluster is calculated, and candidate clusters with dispersion greater than a preset dispersion threshold are used as new initial populations. The process then proceeds to the step of dividing the cell feature points in the initial population into two candidate clusters using the fuzzy C-means clustering method, until the dispersion of all candidate clusters is less than or equal to the preset dispersion threshold. The number of clusters after the current clustering is obtained, and this number is used as the total number of cells in the target grayscale image. It can be understood that the fuzzy C-means clustering method divides all cell feature points into two clusters, and the calculation method for the clustering process is as follows: Figure 3 As shown: First, for the center point (v) of each group i ) and the weights of each sample point corresponding to each group center ( ), set the objective function J : Min J = ; in, N The total number of samples, The weights are subject to the following constraints: st , for k =1,2,3,…, N ; The calculation method is as follows: ; in, , p This is a fuzzy weighting index. Based on this, update the center of each group: ; The calculation process for dividing all cell feature points into two groups using the fuzzy C-means clustering method is as follows: Randomly assign initial weights Conforms to the restricted formula .
[0029] Calculate the initial centroid position .
[0030] Will and Substituting into the objective function J = .
[0031] renew and Then, repeat the process. and Substituting into the objective function J = and updates and The steps are repeated until J reaches its minimum or converges to a stable value, thus obtaining the two group centers. And the final clustering results.
[0032] After obtaining the clustering results (divided into two clusters), the dispersion of the two clusters is calculated separately, as follows: First, the feature vectors of cell feature points within the same group are written in the following form: , ; Secondly, calculate the mean of each feature value of the sample. , c This indicates the number of samples in the group.
[0033] Next, the zero mean is applied to each sample, that is, .
[0034] Then, the covariance matrix K of the sample is calculated. ( (square formation) among them, A represents the matrix after zero-mean processing.
[0035] make Let c be the eigenvalues of the covariance matrix K. Then the formula for calculating the cluster dispersion is: ; in, for Mean: , for The standard deviation is: ; Furthermore, for the cluster with the largest dispersion value, the above-described clustering steps are used to divide it into two more clusters. After obtaining the two new clusters, the dispersion values of all existing clusters are calculated again. For all the resulting clusters, if any cluster still has a dispersion value greater than the threshold... T f Then repeat the process of regrouping and recalculating the dispersion values of the new groups until all cluster dispersion values are less than or equal to the threshold. T f If the grouping is complete, the number of groups obtained is the number of cells.
[0036] Step S14: Calculate the mean and uniformity information of the pixel coordinate feature values of all cell feature points in each subgroup to obtain the center pixel coordinates and cell medical status of each cell.
[0037] In this embodiment, as Figure 4 As shown, the mean pixel coordinates of each cell feature point cluster are calculated to determine the center position of each cell feature point cluster. j The center pixel coordinates of each group ( , The calculation formula is: ; ; in, m j For the first j The number of feature points of each group of cells and They represent the first j The first in the group i The row and column coordinates of each feature point.
[0038] Simultaneously, the Euclidean distance between all pairs of cell feature points in each cluster is calculated, and the corresponding standard deviation of distance is calculated based on the Euclidean distance and the average distance to obtain the uniformity information of the cluster. If the uniformity information is less than or equal to a preset uniformity threshold, the cell's medical state is determined to be viable; if the uniformity information is greater than the preset uniformity threshold, the cell's medical state is determined to be dead. It can be understood that, by using the distance between cell feature points within the same cluster to calculate the uniformity of that cluster, firstly, the distance between cell feature points within the same cluster (assuming there are...) is calculated... m j The distance between every two samples (in samples), i.e.: , , and .
[0039] Secondly, it can be used m j The evenness of the cluster is calculated using the standard deviation of the distance, i.e.: ; ; The cell viability is determined by comparing uniformity with a threshold. ; in, This indicates the preset uniformity threshold.
[0040] In this embodiment, a first preset marker symbol is used to mark the center pixel coordinates of cells determined to be alive on the medical microscopic cell fluorescence image; a second preset marker symbol is used to mark the center pixel coordinates of cells determined to be dead on the medical microscopic cell fluorescence image; and an image file containing cell count statistics and cell status markers is output. Figure 5 As shown, using red " Cells marked at the center of the image (center pixel coordinates) are living cells, and cells marked with a yellow "+" are dead cells. Then, the image file containing the above cell status markers and the cell count statistics obtained above is output for subsequent use.
[0041] Therefore, this application discloses a method for acquiring a target grayscale image after image staining and image preprocessing; performing threshold segmentation on the target grayscale image; extracting image pixels representing cells from the target grayscale image to obtain cell feature points; using an unsupervised clustering method to cluster the cell feature points, with the number of clusters being used as the total number of cells in the target grayscale image; and calculating the mean and uniformity information of the pixel coordinate feature values of all cell feature points in each cluster to obtain the center point pixel coordinates and cell medical status of each cell. Thus, by using an unsupervised clustering method, the number of clusters is automatically determined, meaning that the operator does not need extensive experience to judge whether cells are adhered or have consistent morphology; the division is automatically completed, eliminating interference from human experience. Simultaneously, the extraction of cell feature points and unsupervised clustering, compared to traditional methods based on connected component counting, provides stronger stability and more accurate counting for cell overlap and adhesion. Finally, calculating uniformity information to obtain the cell medical status achieves standardized quantification of the cell medical status judgment process, making the recognition results standardized and repeatable.
[0042] like Figure 6 As shown, this embodiment provides a specific medical microscopic cell image recognition method, including image input and preprocessing, cell feature point extraction, cell cluster counting, cell center localization, and medical status judgment. The specific execution flow of each step is as follows: Step 1: Image Acquisition and Grayscale Conversion. The input is a raw medical microscopic cell fluorescence image. Since the raw image is typically an RGB image containing color information, its color primarily originates from the labeling of specific fluorescent probes. Although color information is meaningful for distinguishing different fluorescence channels, in cell morphology analysis and quantity statistics, the grayscale levels (brightness information) of the image are sufficient to characterize the contrast relationship between cells and the background. Therefore, this embodiment converts the input microscopic cell fluorescence image into a grayscale image. This preprocessing operation not only preserves the fluorescence intensity characteristics of the cells but also compresses the three-channel data into a single channel, effectively reducing the computational complexity of subsequent algorithms and improving processing efficiency.
[0043] Step Two: Threshold Segmentation and Cell Feature Point Extraction. After obtaining the grayscale image, this embodiment uses threshold segmentation technology to extract pixels representing cells. Specifically, the optimal segmentation threshold is automatically determined using the maximum inter-class variance method. By traversing the image's grayscale levels, the inter-class variance between the foreground (cellular region) and background (non-cellular region) is calculated under different thresholds. The grayscale value corresponding to the maximum inter-class variance is the optimal segmentation threshold. This threshold is used to perform binarization segmentation on the grayscale image, and pixels with grayscale values higher than the threshold are identified as cell feature points. These feature points not only contain their pixel coordinate information but also retain the original grayscale value information, collectively forming a three-dimensional feature vector for cluster processing.
[0044] Step 3: Cell Counting Based on Unsupervised Clustering. For the extracted discrete cell feature points, this embodiment introduces an improved unsupervised fuzzy clustering method for cluster analysis. The core of this step is that it does not require pre-setting the specific morphology or approximate number of cells in the image, but automatically identifies the number of clusters through recursive splitting. First, the fuzzy C-means algorithm is used to divide all current feature points into two candidate clusters. Then, a dispersion index is introduced to measure the density of feature points within each cluster. For clusters with excessive dispersion (exceeding a preset threshold), it indicates that there may be multiple cells inside, so these are used as new initial groups, and the fuzzy C-means algorithm is applied again for bisection. Through this recursive process of judgment, splitting, and re-judgment, the dispersion of all clusters meets the preset requirements. The number of clusters obtained at this point is the total number of cells in the microscopic image. This method has good splitting ability for partially adhered or blurred-boundary cells, improving the accuracy of counting.
[0045] Step Four: Cell Center Location and Medical Status Assessment. After clustering the feature points, this embodiment performs individualized analysis on each cell. First, for all feature points within each cluster, the arithmetic mean of their pixel coordinates is calculated. By restoring the zero-mean value of the coordinate system, the precise center pixel coordinates of the cell in the image can be obtained, achieving physical location of each cell. Second, this embodiment determines the cell's viability by analyzing the uniformity of feature point distribution within the same cluster. Specifically, the pairwise Euclidean distances between all feature points within the cluster are calculated, and the standard deviation of these distances is used as a uniformity index. For live cells, fluorescent markers are usually uniformly distributed throughout the cytoplasm, resulting in a uniform feature point distribution and a smaller calculated uniformity value. However, for apoptotic or necrotic cells, the fluorescence signal is often concentrated in the nucleus or diffused, resulting in an uneven feature point distribution and a larger uniformity value. By comparing the calculated uniformity with a preset medical status threshold, each cell can be automatically determined to be either viable or dead.
[0046] Therefore, through the above four steps, a raw microscopic cell fluorescence image is ultimately transformed into structured data output containing the total number of cells, the coordinates of the center of each cell, and the life or death status of each cell.
[0047] like Figure 7 As shown, the present invention also discloses a medical microscopic cell image recognition device, comprising: Image acquisition module 11 is used to acquire the target grayscale image after image coloring and image preprocessing; Feature point extraction processing 12 is used to perform threshold segmentation processing on the target grayscale image, and then extract the image pixels representing cells in the target grayscale image to obtain cell feature points; Clustering processing 13 is used to perform clustering processing on the cell feature points using an unsupervised clustering method, so that the number of clusters after clustering is used as the total number of cells in the target grayscale image. The state recognition processing 14 is used to calculate the mean and uniformity information of the pixel coordinate feature values of all cell feature points in each subgroup, so as to obtain the center point pixel coordinates and cell medical state of each cell respectively.
[0048] Therefore, by using a threshold segmentation method, cell feature points are obtained from medical microscopic cell fluorescence images, and corresponding feature vectors are formed based on the pixel coordinates and gray values of these cell feature points. Then, an improved unsupervised fuzzy clustering method is applied to cluster the cell feature points in the image. The clustering stopping point is determined by measuring the cluster dispersion, and the final number of clusters represents the number of cells in the image. Finally, the center of each cluster is calculated to obtain the corresponding cell pixel position, and the medical state of the cells is determined based on the uniformity of the clusters. By employing image recognition technology and principal component analysis technology, automatic digital recognition of medical microscopic cell fluorescence images can be achieved, enabling automatic detection of cell number and state in medical microscopic cell fluorescence images.
[0049] Furthermore, embodiments of this application also disclose an electronic device, Figure 8 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.
[0050] Figure 8 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the medical microscopic cell image recognition method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.
[0051] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.
[0052] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.
[0053] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.
[0054] The operating system 221 manages and controls the various hardware devices and computer programs 222 on the electronic device 20 to enable the processor 21 to perform calculations and processing on the massive amounts of data 223 in the memory 22. It can be Windows Server, Netware, Unix, Linux, etc. The computer program 222, in addition to including a computer program capable of performing the medical microscopic cell image recognition method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, may further include computer programs capable of performing other specific tasks. The data 223 may include data received by the electronic device from external devices, as well as data collected by its own input / output interface 25.
[0055] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned disclosed medical microscopic cell image recognition method. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0056] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.
[0057] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly in hardware, software modules executed by a processor, or a combination of both. The software module may be located in random access memory (RAM), memory, read-only memory (ROM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, removable disks, CD-ROMs (Compact Disc-Read Only Memory), or any other form of storage medium known in the art.
[0058] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0059] The solution provided by the present invention has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. A method for medical microscopic cell image recognition, characterized in that, include: Obtain the target grayscale image after image coloring and image preprocessing; The target grayscale image is subjected to threshold segmentation, and then the image pixels representing cells in the target grayscale image are extracted to obtain cell feature points; The cell feature points are grouped using an unsupervised clustering method, and the number of clusters after grouping is taken as the total number of cells in the target grayscale image. The mean and uniformity information of the pixel coordinate feature values of all cell feature points in each subgroup are calculated to obtain the center pixel coordinates and cell medical status of each cell.
2. The medical microscopic cell image recognition method according to claim 1, characterized in that, The process of acquiring the target grayscale image after image coloring and image preprocessing includes: Acquire fluorescence images of stained medical microscopy cells; The medical microscopic cell fluorescence image is converted to grayscale to obtain the target grayscale image.
3. The medical microscopic cell image recognition method according to claim 1, characterized in that, The step of performing threshold segmentation on the target grayscale image and then extracting image pixels representing cells from the target grayscale image to obtain cell feature points includes: Iterate through the image grayscale values of each pixel in the target grayscale image, and select the image pixels whose image grayscale values are greater than or equal to the candidate image grayscale threshold as foreground pixels; Image pixels whose grayscale values are less than the candidate image grayscale threshold are designated as background pixels. The foreground pixel ratio is determined based on the total number of foreground pixels and the total number of pixels in the target grayscale image; The background pixel ratio is determined based on the total number of background pixels and the total number of pixels in the target grayscale image; Based on the foreground pixel ratio, the background pixel ratio, the average gray level of the foreground, and the average gray level of the background, an inter-class variance is constructed. The target image gray level threshold that maximizes the inter-class variance is then selected. Target image pixels with gray levels greater than the target image gray level threshold are then used as cell feature points.
4. The medical microscopic cell image recognition method according to claim 3, characterized in that, After defining the target image pixels with gray values greater than the target image gray value threshold as cell feature points, the method further includes: Obtain the pixel coordinates and grayscale values of the target image pixels, and construct pixel coordinate feature values corresponding to cell feature points based on the pixel coordinates and grayscale values.
5. The medical microscopic cell image recognition method according to claim 4, characterized in that, The step of using an unsupervised clustering method to cluster the cell feature points, and using the number of clusters as the total number of cells in the target grayscale image, includes: The fuzzy C-means clustering method is used to divide the cell feature points in the initial population into two candidate clusters; wherein, the initial population is constructed based on all cell feature points to be clustered. Calculate the dispersion of each candidate cluster, and take the candidate clusters with dispersion greater than a preset dispersion threshold as new initial clusters, and jump to the step of dividing the cell feature points in the initial cluster into two candidate clusters using the fuzzy C-means clustering method, until the dispersion of all candidate clusters is less than or equal to the preset dispersion threshold. Obtain the current number of clusters after clustering, and use the number of clusters as the total number of cells in the target grayscale image.
6. The medical microscopic cell image recognition method according to claim 1, characterized in that, Calculate the uniformity information of pixel coordinate feature values of all cell feature points in each cluster to obtain the cellular medical state of each cell, including: Calculate the Euclidean distance between all pairs of cell feature points in each cluster, and calculate the corresponding distance standard deviation based on the Euclidean distance and the mean distance to obtain the uniformity information of the cluster. If the uniformity information is less than or equal to a preset uniformity threshold, the cell's cellular medical state is determined to be a viable state. If the uniformity information is greater than a preset uniformity threshold, the cell is determined to be in a dead state in the cellular medicine context.
7. The medical microscopic cell image recognition method according to any one of claims 1 to 6, characterized in that, Also includes: The center pixel coordinates of cells identified as viable are marked on the medical microscopic cell fluorescence image using a first preset marker symbol. The center pixel coordinates of cells determined to be dead are marked on the medical microscopic cell fluorescence image by using a second preset marker symbol. The output includes cell count statistics and image files with cell status markers.
8. A medical microscopic cell image recognition device, characterized in that, include: The image acquisition module is used to acquire the target grayscale image after image coloring and image preprocessing; Feature point extraction processing is used to perform threshold segmentation on the target grayscale image, and then extract the image pixels representing cells in the target grayscale image to obtain cell feature points; Clustering is used to cluster the cell feature points using an unsupervised clustering method, so that the number of clusters after clustering is taken as the total number of cells in the target grayscale image. The state recognition process is used to calculate the mean and uniformity information of the pixel coordinate feature values of all cell feature points in each subgroup, so as to obtain the center point pixel coordinates and cell medical state of each cell respectively.
9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the medical microscopic cell image recognition method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, Used to store a computer program; wherein, when the computer program is executed by a processor, it implements the steps of the medical microscopic cell image recognition method as described in any one of claims 1 to 7.