Stem cell classification method and system based on image segmentation

Abstract

The invention relates to the technical field of image recognition, in particular to a stem cell classification method and system based on image segmentation. The method comprises the following steps: acquiring original stem cell image data, and performing Wasm processing on the original stem cell image data to obtain stem cell image quality parameters; performing adaptive preprocessing on the original stem cell image data based on the stem cell image quality parameters to obtain standardized stem cell image data; and performing multi-resolution pyramid construction on the standardized stem cell image data to obtain multi-level stem cell image representation data. According to the method, through multi-resolution pyramid construction, hierarchical processing of cell regions, accurate morphological feature extraction and hierarchical classification and distributed storage processing, the accuracy, reliability, adaptability and expandability of stem cell image classification are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a stem cell classification method and system based on image segmentation. Background Art

[0002] Due to various interference factors in the microscope imaging process, the accurate extraction and identification of cell areas will be affected. Although the existing image processing methods can improve the image quality to a certain extent, these methods often lack targeted adaptability and cannot automatically adjust parameters according to the specific conditions of different samples, resulting in unsatisfactory preprocessing results. In practical applications, stem cell images often have cell overlap and aggregation, which makes image processing more complicated. When facing cell overlapping areas, traditional image segmentation methods usually use simple methods based on threshold or edge detection. Although they can perform preliminary segmentation, due to the similarity of morphology, color and texture information between cells, they cannot effectively distinguish overlapping cells. The morphological characteristics of stem cells are crucial for their classification. However, traditional morphological feature extraction methods usually rely on rough edge detection and morphological filtering techniques, resulting in inaccurate recognition of cell structure and morphology. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a stem cell classification method and system based on image segmentation to solve at least one of the above technical problems.

[0004] To achieve the above object, a stem cell classification method based on image segmentation comprises the following steps:

[0005] Step S1: acquiring original stem cell image data, performing Wasm processing on the original stem cell image data to obtain stem cell image quality parameters; performing adaptive preprocessing on the original stem cell image data based on the stem cell image quality parameters to obtain standardized stem cell image data;

[0006] Step S2: constructing a multi-resolution pyramid for the standardized stem cell image data to obtain multi-level stem cell image representation data;

[0007] Step S3: performing multi-resolution hierarchical processing on the multi-level stem cell image representation data to obtain multi-level cell feature data, wherein the multi-level cell feature data includes potential cell region data, cell aggregation region data and cell overlap region data;

[0008] Step S4: performing stem cell adaptive algorithm segmentation based on the multi-level cell feature data to obtain segmented stem cell image data;

[0009] Step S5: extracting morphological features from the segmented stem cell image data, and performing hierarchical classification processing based on the morphological feature extraction results to obtain stem cell classification result data;

[0010] Step S6: Perform distributed storage processing on the stem cell classification result data to generate a persistent stem cell classification database.

[0011] The present invention can effectively improve the accuracy and reliability of stem cell image classification through the processing of each step. First, by obtaining stem cell image quality parameters through Wasm processing, the quality of the original image can be evaluated, and adaptive preprocessing can be performed based on these parameters, thereby overcoming the defect that the traditional method cannot automatically adjust according to different sample conditions in the image preprocessing stage, ensuring the standardization of image data, and improving the accuracy of subsequent processing. Through multi-resolution pyramid construction, stem cell image representation data at different levels can be obtained, so that the multi-scale characteristics of cell images can be better captured during the processing process, and the recognition ability of cell regions and their complex structures is improved. Further, through multi-resolution hierarchical processing, the cell region data is divided into potential cell regions, cell aggregation regions, and cell overlap regions, which can be more meticulously divided into regions, avoiding the problem of inaccurate processing of cell overlap and aggregation phenomena by traditional methods, thereby improving the effect of cell segmentation. Using stem cell adaptive algorithm segmentation, the segmentation strategy can be automatically adjusted according to the characteristics of different samples, overcoming the limitations of traditional image segmentation methods that insufficiently process cell overlap and morphological similarity, and ensuring the accuracy of segmentation results. The morphological feature extraction combined with hierarchical classification processing enables the cell classification process to better classify cells based on their morphological features, avoiding the influence of overly rough edge detection and morphological filtering in traditional methods, and improving the reliability of classification results. Finally, a persistent stem cell classification database is generated through distributed storage processing, ensuring efficient storage, management, and query of data, and supporting the processing requirements of large-scale data sets, further improving the adaptability and scalability of this method in practical applications.

[0012] Preferably, the present invention further provides a stem cell classification system based on image segmentation, which is used to execute the above-mentioned stem cell classification method based on image segmentation, and the stem cell classification system based on image segmentation comprises:

[0013] The image preprocessing and standardization module is used to obtain the original stem cell image data, perform Wasm processing on the original stem cell image data, and obtain the stem cell image quality parameters; based on the stem cell image quality parameters, the original stem cell image data is adaptively preprocessed to obtain standardized stem cell image data;

[0014] A multi-resolution pyramid construction module is used to construct a multi-resolution pyramid for the standardized stem cell image data to obtain multi-level stem cell image representation data;

[0015] A multi-resolution hierarchical feature extraction module is used to perform multi-resolution hierarchical processing on the multi-level stem cell image representation data to obtain multi-level cell feature data, wherein the multi-level cell feature data includes potential cell area data, cell aggregation area data and cell overlap area data;

[0016] A stem cell adaptive segmentation module is used to perform stem cell adaptive algorithm segmentation based on multi-level cell feature data to obtain segmented stem cell image data;

[0017] The morphological feature extraction and classification module is used to extract morphological features from the segmented stem cell image data, and perform hierarchical classification processing based on the morphological feature extraction results to obtain stem cell classification result data;

[0018] The distributed storage and persistence module is used to perform distributed storage processing on stem cell classification result data and generate a persistent stem cell classification database.

[0019] The present invention improves the quality of original stem cell image data through adaptive preprocessing and Wasm processing, and standardizes it to ensure the accuracy of subsequent analysis. Through multi-resolution pyramid construction, multi-level stem cell image data is generated, providing a multi-scale perspective for cell feature extraction and segmentation, and enhancing processing capabilities. Through multi-level cell feature extraction, potential cell areas, cell aggregation areas and cell overlap areas are accurately identified, providing rich feature data for subsequent stem cell segmentation and classification. Stem cell image segmentation is performed through an adaptive algorithm, and cell areas are accurately separated to provide clear stem cell image data for classification processing. Through morphological feature extraction and hierarchical classification, cell classification results are optimized to achieve efficient and accurate stem cell classification. Through distributed storage and persistent processing, efficient storage and management of stem cell classification result data are ensured, which is convenient for subsequent query and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0021] Figure 1 It is a schematic diagram of the steps of the stem cell classification method based on image segmentation of the present invention;

[0022] Figure 2 for Figure 1 Detailed step flow diagram of step S1;

[0023] Figure 3 for Figure 1Detailed step flow chart of step S2 in FIG. DETAILED DESCRIPTION

[0024] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0025] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0026] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0027] To achieve this, please refer to Figures 1 to 3 The present invention provides a stem cell classification method based on image segmentation, the method comprising the following steps:

[0028] Step S1: acquiring original stem cell image data, performing Wasm processing on the original stem cell image data to obtain stem cell image quality parameters; performing adaptive preprocessing on the original stem cell image data based on the stem cell image quality parameters to obtain standardized stem cell image data;

[0029] Step S2: constructing a multi-resolution pyramid for the standardized stem cell image data to obtain multi-level stem cell image representation data;

[0030] Step S3: performing multi-resolution hierarchical processing on the multi-level stem cell image representation data to obtain multi-level cell feature data, wherein the multi-level cell feature data includes potential cell region data, cell aggregation region data and cell overlap region data;

[0031] Step S4: performing stem cell adaptive algorithm segmentation based on the multi-level cell feature data to obtain segmented stem cell image data;

[0032] Step S5: extracting morphological features from the segmented stem cell image data, and performing hierarchical classification processing based on the morphological feature extraction results to obtain stem cell classification result data;

[0033] Step S6: Perform distributed storage processing on the stem cell classification result data to generate a persistent stem cell classification database.

[0034] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of a process flow of a stem cell classification method based on image segmentation according to the present invention. In this example, the stem cell classification method based on image segmentation includes the following steps:

[0035] Step S1: acquiring original stem cell image data, performing Wasm processing on the original stem cell image data to obtain stem cell image quality parameters; performing adaptive preprocessing on the original stem cell image data based on the stem cell image quality parameters to obtain standardized stem cell image data;

[0036] In the embodiment of the present invention, the original stem cell image data in the stem cell culture dish is obtained by a confocal microscope with a 10X objective lens, and the image resolution is 1024×1024 pixels; the obtained original stem cell image data is processed by an image analysis module compiled by WebAssembly, and the specific operation is to separate the image data according to the three channels of RGB, calculate the brightness mean, standard deviation, signal-to-noise ratio, contrast and clarity of each channel, and comprehensively form the stem cell image quality parameters; then, according to the stem cell image quality parameters, when the image brightness mean is lower than 30 or higher than 220, a histogram equalization algorithm is used; when the standard deviation is greater than 50, an image quality parameter is generated. Gaussian filtering was used for noise reduction, and the filter kernel size was set to 5×5. When the signal-to-noise ratio was lower than 15 dB, the median filter was used to remove salt and pepper noise, and the filter window size was 3×3. When the clarity index was lower than 0.65, the non-local mean denoising algorithm was applied, and the search window radius was set to 7. Through the above series of processing operations, image color space standardization, illumination non-uniformity correction and noise elimination were achieved, and finally standardized stem cell image data was output. The data size remained unchanged at 1024×1024 pixels, but the dynamic range was adjusted to between 0 and 255, the RGB channel mean was set to 128±5, and the standard deviation was controlled in the range of 40-60.

[0037] Step S2: constructing a multi-resolution pyramid for the standardized stem cell image data to obtain multi-level stem cell image representation data;

[0038] In the embodiment of the present invention, the standardized stem cell image data is used as the initial level L0, with a resolution of 1024×1024 pixels, and a Gaussian kernel function is used to downsample L0, with a Gaussian kernel size of 5×5, a standard deviation σ=1.6, and a downsampling factor of 2, to generate a L1 level with a resolution of 512×512 pixels; then the Gaussian kernel downsampling with the same parameters is continued to be applied to the L1 level to generate a L2 level with a resolution of 256×256 pixels; and so on, a L3 level with a resolution of 128×128 pixels and a L4 level with a resolution of 64×64 pixels are gradually generated, thereby constructing a 5-layer structure multi-layer structure. Resolution pyramid; for each level Li (i = 0, 1, 2, 3, 4), calculate the Laplacian difference image Di = Li-upsample (Li+1), where the upsampling operation uses the bicubic interpolation algorithm to enlarge the size of Li+1 to the same as Li; for the bottom layer L4, directly use it as D4; combine the difference images Di of each layer with the original level Li into a multi-level stem cell image representation data {(L0, D0), (L1, D1), (L2, D2), (L3, D3), (L4, D4)}, each layer contains the original resolution image and the corresponding Laplacian difference image.

[0039] Step S3: performing multi-resolution hierarchical processing on the multi-level stem cell image representation data to obtain multi-level cell feature data, wherein the multi-level cell feature data includes potential cell region data, cell aggregation region data and cell overlap region data;

[0040] In the embodiment of the present invention, the multi-level stem cell image representation data {(L0, D0), (L1, D1), (L2, D2), (L3, D3), (L4, D4)} obtained in step S2 are subjected to feature extraction processing respectively. First, local contrast enhancement is performed on the L0 level (1024×1024 pixels). The contrast enhancement adopts an adaptive histogram equalization algorithm to divide the image into 16×16 tiles. The size of each tile is 64×64 pixels. The clip limit is set to 3.0. The potential cell area is identified by calculating the difference between the 90% and 10% percentiles of the pixel intensity in each tile. Then, the Otsu threshold segmentation method is applied to the L2 level (256×256 pixels) to determine the global threshold. The threshold is set to the minimum value between the two peaks in the bimodal distribution of the grayscale histogram to obtain the cell aggregation area data. Then, the zero crossing detection algorithm is used to identify the cell boundary on the D1 difference image. The zero crossing detection algorithm is used to identify the cell boundary. The detection used the Laplacian operator with a kernel size of 3×3, a smoothing factor α=0.2, and a boundary extraction threshold of 10 to obtain cell boundary information; then morphological operations were performed on the L3 level (128×128 pixels), including using a circular structure element with a radius of 3 pixels to perform an opening operation to remove noise, and then using a circular structure element with a radius of 5 pixels to perform a closing operation to fill holes and identify cell overlap areas; finally, the processing results of each level were mapped back to the original resolution L0 through the back projection technology. The back projection used the nearest neighbor interpolation algorithm to maintain the sharpness of the boundaries, and the boundary matching threshold was set to 0.85 to form a multi-level cell feature data set M={P,A,O} containing potential cell area data, cell aggregation area data and cell overlap area data, where P represents the potential cell area binary mask, A represents the cell aggregation area probability map, and O represents the cell overlap area labeling map. All feature data are stored at the same 1024×1024 resolution as L0.

[0041] Step S4: performing stem cell adaptive algorithm segmentation based on the multi-level cell feature data to obtain segmented stem cell image data;

[0042] The embodiment of the present invention takes a multi-level cell feature data set M={P,A,O} as input, first constructs an initial watershed marker map, performs distance transformation on the potential cell area data P, uses the Euclidean distance metric, and normalizes the distance range to [0,255]. Then, the H-maximum suppression algorithm is applied to the transformation result to extract the local maximum as the cell core area marker point, and the suppression radius is set to 15 pixels. Then, the probability value in the cell cluster area data A is used to construct a segmentation weight map W. The weight value calculation formula is W(x,y)=255-A(x,y)×(1-O(x,y)×0.5), ensuring that the cell cluster area has a lower weight and the boundary area has a higher weight. Then, the marker point matrix and the weight map W are input into the marker-controlled watershed algorithm. The neighborhood connection adopts 8 connectivity, the watershed gradient threshold is set to 25, and the iteration termination condition is set to 15 pixels. The condition is that the regional change rate is less than 0.01%; for the detected cell overlapping area data O, the concave point detection algorithm is used to identify the cell junction, the concave point detection angle threshold is set to 120 degrees, and the edge mutation threshold is set to 0.65. Then the conditional random field algorithm is applied to segment the overlapping cells. The energy function uses the Potts model, the smoothing factor λ is set to 0.8, and the number of iterations is 200 times; finally, the segmentation results are verified, and the noise areas with an area less than 200 pixels or a perimeter / area ratio greater than 0.75 are removed. The areas with an area greater than 2000 pixels and a roundness less than 0.5 are segmented again to obtain the final segmented stem cell image data, which contains the precise contour boundary and unique identifier of each independent cell, maintains the original 1024×1024 pixel resolution, and is stored in the form of a binary mask array, and each mask corresponds to an independent stem cell.

[0043] Step S5: extracting morphological features from the segmented stem cell image data, and performing hierarchical classification processing based on the morphological feature extraction results to obtain stem cell classification result data;

[0044] The embodiment of the present invention uses the segmented stem cell image data obtained in step S4 as input, and first extracts 28 morphological features for each segmented independent stem cell region, including area (calculated in terms of the number of pixels), perimeter (number of boundary pixels), equivalent diameter (circle diameter equal to the cell area), maximum inscribed circle radius, minimum circumscribed rectangle aspect ratio, minimum circumscribed ellipse eccentricity, circularity (4π×area / perimeter2), convex hull area, number of convex defects, total convex defect area, mean convex defect depth, elongation (length-to-width ratio), density (area / convex hull area), curvature mean, curvature standard deviation, 7 values ​​of Hu invariant moments, 5 coefficients of Zernike moments, grayscale mean, grayscale variance, grayscale entropy, texture energy, texture contrast, texture homogeneity and texture correlation; then the extracted features are normalized, using the Z-score normalization method, and the calculation formula is Z=(x-μ) / σ, where μ is the feature mean and σ is the standard deviation; Subsequently, a hierarchical clustering algorithm was used to classify the standardized features. The distance metric was Euclidean distance, the link criterion was Ward's minimum variance method, and the hierarchical tree truncation threshold was set to 0.35. Stem cells were divided into five main categories: pluripotent stem cells (circularity>0.85, area<1000 pixels), precursor cells (0.65<circularity<0.85, 1000<area<1500 pixels), differentiating cells (0.4<circularity<0.65, elongation>1.8), fully differentiated cells (circularity<0.4, elongation>2.5) and abnormal cells (convex defect area>25% of the total area or texture entropy>6.0); finally, the classification results were verified by calculating the intra-class distance and inter-class distance ratio to ensure classification robustness. The classification robustness threshold was set to 0.75 to generate stem cell classification result data. The data format is a structure for each cell, which contains cell ID, position coordinates, morphological feature vector and classification label.

[0045] Step S6: Perform distributed storage processing on the stem cell classification result data to generate a persistent stem cell classification database.

[0046] The embodiment of the present invention takes stem cell classification result data as input, firstly structures the data, and constructs a unique identifier for each stem cell classification result. The identifier consists of 16 hexadecimal characters, including the acquisition timestamp, sample number and cell number; then the stem cell classification result data is serialized into a binary object, using the ProtocolBuffers format, and the compression level is set to 7 to ensure data density; then a three-layer distributed storage architecture is constructed, the first layer is a memory database, using a Redis cluster, configuring 3 master nodes and 3 slave nodes, allocating 4GB of memory to each node, and setting the data expiration time to 72 hours; the second layer is a relational database, using a PostgreSQL database Cluster, create a partition table with dual partitions by time and cell type, the primary key index is based on the cell unique identifier, and the secondary index is based on morphological features and classification results; the third layer is object storage, using a distributed file system, the data block size is set to 128MB, the replication factor is 3, and the data sharding strategy is based on cell type and time range; finally, a data synchronization mechanism is used to ensure data consistency between the three layers of storage, using a two-phase commit protocol, the transaction timeout is set to 30 seconds, the number of retries is 3, and the synchronization interval is 5 minutes, completing the persistent storage of the stem cell classification database, which contains complete stem cell images, segmentation results, morphological features, and classification labels, and supports multi-dimensional query and retrieval by cell type, morphological feature range, and acquisition time.

[0047] The present invention can effectively improve the accuracy and reliability of stem cell image classification through the processing of each step. First, by obtaining stem cell image quality parameters through Wasm processing, the quality of the original image can be evaluated, and adaptive preprocessing can be performed based on these parameters, thereby overcoming the defect that the traditional method cannot automatically adjust according to different sample conditions in the image preprocessing stage, ensuring the standardization of image data, and improving the accuracy of subsequent processing. Through multi-resolution pyramid construction, stem cell image representation data at different levels can be obtained, so that the multi-scale characteristics of cell images can be better captured during the processing process, and the recognition ability of cell regions and their complex structures is improved. Further, through multi-resolution hierarchical processing, the cell region data is divided into potential cell regions, cell aggregation regions, and cell overlap regions, which can be more meticulously divided into regions, avoiding the problem of inaccurate processing of cell overlap and aggregation phenomena by traditional methods, thereby improving the effect of cell segmentation. Using stem cell adaptive algorithm segmentation, the segmentation strategy can be automatically adjusted according to the characteristics of different samples, overcoming the limitations of traditional image segmentation methods that insufficiently process cell overlap and morphological similarity, and ensuring the accuracy of segmentation results. The morphological feature extraction combined with hierarchical classification processing enables the cell classification process to better classify cells based on their morphological features, avoiding the influence of overly rough edge detection and morphological filtering in traditional methods, and improving the reliability of classification results. Finally, a persistent stem cell classification database is generated through distributed storage processing, ensuring efficient storage, management, and query of data, and supporting the processing requirements of large-scale data sets, further improving the adaptability and scalability of this method in practical applications.

[0048] Preferably, step S1 comprises the following steps:

[0049] Step S11: acquiring original stem cell image data, and compiling the original stem cell image data using a Wasm module to obtain a high-performance image processing module;

[0050] Step S12: performing quality assessment on the original stem cell image data based on a high-performance image processing module to obtain stem cell image quality parameters;

[0051] Step S13: performing image quality analysis and preprocessing strategy determination on stem cell image quality parameters to obtain an optimal preprocessing strategy;

[0052] Step S14: performing image noise reduction and contrast enhancement on the original stem cell image data based on the optimal preprocessing strategy to obtain enhanced stem cell image data;

[0053] Step S15: performing illumination correction and edge optimization on the enhanced stem cell image data to obtain standardized stem cell image data.

[0054] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in the embodiment of the present invention, step S1 includes the following steps:

[0055] Step S11: acquiring original stem cell image data, and compiling the original stem cell image data using a Wasm module to obtain a high-performance image processing module;

[0056] In the embodiment of the present invention, a microscope system is equipped to collect raw stem cell image data. The magnification of the microscope objective lens is set to 20X, the numerical aperture is 0.75, the image acquisition device is a 12-megapixel CCD camera, the pixel size is 3.75 μm×3.75 μm, the exposure time is set to 25 milliseconds, the acquisition resolution is 2048×2048 pixels, and the color depth is 24-bit RGB color; the raw stem cell image data is compiled by WebAssembly module. First, the image processing algorithm source code written in C++, including four core functions of Gaussian blur, median filtering, Laplace edge detection, and threshold segmentation, is converted into LLVM intermediate code through the Emscripten compiler, and the compilation instruction is "emcc-O3-s WASM=1", the optimization level is set to 3; then the LLVM intermediate code is compiled into a WebAssembly binary module to generate a .wasm file with a size of 256KB; then the JavaScript glue code is built to implement memory management and function calls, and the glue code size is 32KB; then the module is instantiated, the initial memory size is set to 16MB, and the maximum size can be expanded to 64MB, and the exported function interface includes four main functions: processImage, detectEdges, enhanceContrast, and correctIllumination; finally, the module performance is optimized, the SIMD instruction set is applied to accelerate, the data processing is parallelized into 8 channels, and the memory access mode is optimized to linear continuous access, completing the construction of a high-performance image processing module.

[0057] Step S12: performing quality assessment on the original stem cell image data based on a high-performance image processing module to obtain stem cell image quality parameters;

[0058] The embodiment of the present invention applies a high-performance image processing module to the original stem cell image data. By calling the processImage function in the module, the original image of 2048×2048 pixels is loaded into the WebAssembly linear memory, the memory start address is set to 0x1000, and 16MB of continuous space is allocated; then a multi-parameter quantitative evaluation is performed, firstly calculating the global brightness statistical parameters, including the average brightness values ​​μR, μG, and μB of the three RGB channels, and the calculation formula is μc=∑I(x,y,c) / (W×H), where c∈{R,G,B}, W=H=2048; then calculating The contrast parameter is calculated by Michelson contrast, the formula is C = (Imax-Imin) / (Imax+Imin), and the contrast values ​​of the three channels CR, CG, and CB are obtained respectively; then the image noise level is measured, and the mean square error calculation method is used to calculate the local variance σ2local for each 7×7 pixel block, and the mean of all local variances is taken as the noise estimation value σnoise; then the image clarity is evaluated, and the Tenengrad algorithm is applied. The horizontal and vertical gradients Gx and Gy are calculated using the 3×3 Sobel operator, and then the gradient amplitude G(x, y) = √(Gx 2 +Gy2), the clarity index S = ∑G(x,y)2, and the value range is normalized to [0,1]; then the color balance is measured, and the proportional deviation of the RGB three channels is calculated δRG = |μR / μG-1|, δRB = |μR / μB-1|, δGB = |μG / μB-1|; finally, the frequency characteristics are analyzed by Fourier transform, the image is converted to the frequency domain, and the ratio of high-frequency to low-frequency components ρHF / LF is calculated. All the calculation results are combined to form the stem cell image quality parameter set Q = {μR, μG, μB, CR, CG, CB, σnoise, S, δRG, δRB, δGB, ρHF / LF}, and the parameter set is stored in the form of a floating-point array.

[0059] Step S13: performing image quality analysis and preprocessing strategy determination on stem cell image quality parameters to obtain an optimal preprocessing strategy;

[0060] The embodiment of the present invention first extracts four indicators, namely signal-to-noise ratio, contrast, brightness variance, and edge sharpness, from the stem cell image quality parameters, and determines the preprocessing strategy through the threshold judgment rule. When the signal-to-noise ratio is lower than 30 dB, the non-local mean denoising algorithm is triggered. The OpenCV The fastNlMeansDenoisingColored function of 4.5 is used, and the parameters are set to h=7, templateWindowSize=7, and searchWindowSize=21. When the contrast is lower than 50, limited contrast adaptive histogram equalization is performed, using the CLAHE algorithm and setting clipLimit=2.0 and tileGridSize=(8,8). When the brightness variance exceeds 100, multi-scale Retinex illumination correction is applied, and a three-layer decomposition structure with Gaussian kernel scales σ=15, 30, and 60 is adopted. When the edge sharpness is lower than 0.5, the Sobel operator edge optimization is started, the convolution kernel size is fixed to 3×3, and the gradient calculation uses the bidirectional difference of dx=1, dy=0 and dx=0, dy=1. Finally, the priority of each strategy is sorted by the parameter weight matrix [0.4, 0.3, 0.2, 0.1] to generate the optimal preprocessing strategy configuration file containing algorithm parameters and execution order.

[0061] Step S14: performing image noise reduction and contrast enhancement on the original stem cell image data based on the optimal preprocessing strategy to obtain enhanced stem cell image data;

[0062] The embodiment of the present invention applies the optimal preprocessing strategy to the original stem cell image data. First, adaptive noise reduction processing is performed according to the noise type. When σnoise<10, Gaussian filtering is applied for smoothing processing, the Gaussian kernel size is set to 5×5, and the standard deviation σ=1.2; when 10≤σnoise<25, bilateral filtering is applied to retain edge details, the spatial domain standard deviation is set to 6.0, and the range standard deviation is set to 25.0; when σnoise≥25, a non-local mean filtering algorithm is applied to remove high-intensity noise, the search window size is set to 21×21, the comparison window size is set to 7×7, and the filter strength parameter is set to 1.5 times the noise standard deviation; then contrast enhancement processing is performed. When the image contrast C<0.3, global histogram equalization is applied, and the 256-level grayscale mapping function adopts the cumulative distribution function; when 0.3≤C<0.5, limited contrast adaptive histogram equalization (CLAHE) is applied to segment the image into 8×8 grid blocks, each block size is 256×256 pixels, the contrast limit threshold is set to 3.5, and the bilinear interpolation coefficient α is 0.75; when C ≥ 0.5 and there is a local area with insufficient contrast, the multi-scale Retinex algorithm is applied, using three Gaussian filter kernels σ1 = 15, σ2 = 80, σ3 = 250, with weight coefficients w1 = 0.3, w2 = 0.5, w3 = 0.2, and color restoration factor β = 1.2; after implementing the above-mentioned adaptive denoising and contrast enhancement operations, the enhanceContrast function in the high-performance image processing module is called to complete the processing. The function input is the original stem cell image data and the preprocessing parameter set, and the output is the enhanced stem cell image data. After processing, the image resolution remains at 2048×2048 pixels, and the bit depth is still 24 bits, but the noise level is reduced to less than 30% of the original value, and the contrast is increased to 1.5-2.5 times that of the original image.

[0063] Step S15: performing illumination correction and edge optimization on the enhanced stem cell image data to obtain standardized stem cell image data.

[0064] The embodiment of the present invention takes the enhanced stem cell image data as input, first performs non-uniform illumination correction, adopts a polynomial fitting background illumination model, divides the image into 256 sub-regions according to a 16×16 grid, and each sub-region is 128×128 pixels in size, calculates the average brightness value of each sub-region, constructs a brightness distribution matrix M, applies cubic B-spline interpolation to generate a continuous illumination model function B(x, y), calculates the correction coefficient K(x, y)=μtarget / B(x, y), where μtarget is set to 128, and multiplies each pixel value I(x, y) of the original image by the correction coefficient K(x, y) of the corresponding position to obtain an illumination balanced image; then performs edge optimization processing, first applies the Canny edge detection algorithm, sets the low threshold to 10% of the image intensity, sets the high threshold to 25% of the image intensity, and the Gaussian smoothing factor σ=1.0, and detects cells. The initial position of the boundary; then the edge is enhanced by combining morphological operations, and a linear structure element with a length of 5 and a width of 1 is used for directional dilation. The directional angle interval is set to 45 degrees. The dilation in 8 directions is performed and the union is taken; then the sub-pixel edge positioning technology is applied, the second-order moment of the local grayscale gradient of the edge point is calculated, and the quadratic curve is fitted to obtain the precise boundary position. The edge positioning accuracy reaches 0.1 pixel; finally, the color space is standardized, the image is converted from RGB space to Lab color space, the L channel mean is adjusted to 50, the standard deviation is 15, the a channel mean is 0, the standard deviation is 5, the b channel mean is 0, the standard deviation is 5, and then converted back to RGB space to output standardized stem cell image data. The data size is still 2048×2048 pixels, the color depth is 24 bits, but it has uniform lighting conditions, clear cell boundaries and standardized color representation.

[0065] The present invention is compiled by Wasm module to achieve high-performance image processing, improve image processing efficiency and computing performance. Based on the high-performance processing module, the quality of the original stem cell image data is evaluated, and the image quality parameters are accurately obtained to provide a reliable basis for subsequent processing. Through image quality analysis and preprocessing strategy determination, the preprocessing process is optimized to ensure that image processing is more adapted to the needs of different samples. Through image noise reduction and contrast enhancement, the image quality is improved, the visualization effect of cell structure and region is enhanced, and clearer image data is provided for subsequent segmentation. Through illumination correction and edge optimization, illumination interference is eliminated and edge recognition accuracy is improved, image data standardization is ensured, which is conducive to subsequent cell segmentation and classification.

[0066] Preferably, step S2 comprises the following steps:

[0067] Step S21: performing Gaussian filtering and step-by-step downsampling processing on the standardized stem cell image data to obtain a multi-scale stem cell image sequence;

[0068] Step S22: organizing and arranging different resolution levels based on the multi-scale stem cell image sequence to obtain a hierarchical stem cell image pyramid structure;

[0069] Step S23: performing adjacent-level differential operations on the hierarchical stem cell image pyramid structure, and normalizing the adjacent-level differential operation results to obtain standardized feature representation data;

[0070] Step S24: performing information fusion processing of each level based on the standardized feature representation data to obtain multi-level stem cell image representation data.

[0071] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed step flow diagram of step S2 in the embodiment of the present invention, step S2 includes the following steps:

[0072] Step S21: performing Gaussian filtering and step-by-step downsampling processing on the standardized stem cell image data to obtain a multi-scale stem cell image sequence;

[0073] In the embodiment of the present invention, the standardized stem cell image data is first input into a Gaussian filter with a convolution kernel size of 5×5 and a standard deviation of 1.6 for smoothing to reduce image noise and retain key features; then, a step-by-step downsampling operation is performed on the filtered stem cell image, and the downsampling adopts a bilinear interpolation algorithm to reduce the image size to 1 / 2, 1 / 4, 1 / 8, and 1 / 16 of the original size, respectively. The specific implementation process is as follows: Gaussian filtering is performed on the original image I0(x, y) (resolution is W×H) to obtain I0'(x, y), and then I0'(x, y) is downsampled to obtain the first layer image I1(x, y) (resolution is W / 2×H / 2 ), continue to apply Gaussian filtering with the same parameters to I1(x,y) to obtain I1'(x,y), and then downsample to obtain the second layer image I2(x,y) (resolution is W / 4×H / 4), and so on to obtain a total of 5 layers of stem cell images with different resolutions; in the downsampling process, a Gaussian filter with the same parameters is applied for preprocessing before each image downsampling; each layer of image maintains the same pixel depth; for RGB color stem cell images, the above-mentioned Gaussian filtering and downsampling operations are performed on the three channels of R, G, and B respectively, and finally a multi-scale stem cell image sequence is obtained, which retains the morphological characteristics and cell boundary information of stem cells at different resolutions.

[0074] Step S22: organizing and arranging different resolution levels based on the multi-scale stem cell image sequence to obtain a hierarchical stem cell image pyramid structure;

[0075] The embodiment of the present invention strictly arranges and organizes the multi-scale stem cell image sequence in order from high resolution to low resolution to construct a hierarchical stem cell image pyramid structure; first, the original resolution image I0(x, y) (W×H) is used as the bottom layer of the pyramid, and the downsampled images I1(x, y) (W / 2×H / 2), I2(x, y) (W / 4×H / 4), I3(x, y) (W / 8×H / 8) and I4(x, y) (W / 16×H / 16) are used as the second, third, fourth and fifth layers of the pyramid respectively; then, an index relationship between the levels is established, and for the pixel point (x, y) of any level in the pyramid, a mapping relationship with the pixel point (x / 2, y / 2) of the next level is established through a formula to ensure the spatial correspondence between the levels; then, memory space is allocated for each level image and a pointer to the image is created. The pointer structure of adjacent levels contains the address of the image data of the current level, the image size information, and the address references of the adjacent upper and lower levels. For color stem cell images, independent pyramid structures are established for the three channels of R, G, and B to maintain the hierarchical consistency between channels. To ensure the integrity of the pyramid structure, a quadtree data structure is used in the storage process. Each parent node is connected to four child nodes. The parent node stores low-resolution image information, and the child node stores high-resolution image information of the corresponding area. After the pyramid structure is created, the continuity between the levels is verified by calculating the mean and standard deviation of each layer of the image to ensure that the images of different levels in the pyramid structure can correctly represent the characteristics of stem cells at different scales. Finally, a hierarchical stem cell image pyramid structure containing 5 levels and with a strict spatial correspondence is output.

[0076] Step S23: performing adjacent-level differential operations on the hierarchical stem cell image pyramid structure, and normalizing the adjacent-level differential operation results to obtain standardized feature representation data;

[0077] In the embodiment of the present invention, the lower resolution image Il(x,y) is first upsampled to the same size as the adjacent higher resolution image Ih(x,y) by a bilinear interpolation algorithm. The interpolation process uses the formula f(x,y)=(1-a)(1-b)f(i,j)+a(1-b)f(i+1,j)+(1-a)bf(i,j+1)+abf(i+1,j+1) to calculate the pixel value of the interpolation point, where a=xi, b=yj, and (i,j) is the integer coordinate point closest to the upper left corner of the interpolation point; then, a pixel-level difference operation is performed on the upsampled image Il'(x,y) and the corresponding high resolution image Ih(x,y), and the calculation formula is D(x,y)=Ih(x,y)-Il'(x,y) to obtain a difference image D(x,y); the above difference calculation is repeated for all adjacent levels in the pyramid to obtain four sets of differences Images D1(x,y), D2(x,y), D3(x,y) and D4(x,y) respectively represent the detail information between adjacent levels; then each group of differential images is normalized using the min-max normalization method, and the calculation formula is Dn(x,y)=(D(x,y)-Dmin) / (Dmax-Dmin)×255, where Dmin and Dmax are the minimum and maximum values ​​of the differential image D(x,y), respectively; for color stem cell images, the same normalization process is performed on the differential images of the three channels of R, G, and B; in order to eliminate noise interference in the differential images, hard threshold processing with a threshold of 10 is applied to the normalized differential images, and the differential values ​​with absolute values ​​less than the threshold are set to 0; finally, the four groups of normalized differential images are combined with the original five-layer pyramid image to form standardized feature representation data.

[0078] Step S24: performing information fusion processing of each level based on the standardized feature representation data to obtain multi-level stem cell image representation data.

[0079] In the embodiment of the present invention, the original images and differential images of all levels are first uniformly scaled to the same resolution W×H as the original stem cell image. The scaling method adopts a bicubic interpolation algorithm, and the interpolation kernel function is h(x)=-(α+2)|x|3+(α+3)|x|2 when |x|<1, and h(x)=α|x|3-5α|x|2+8α|x|-4α when 1≤|x|<2, wherein the parameter α is set to -0.5; then, different weights are assigned to images of each level based on the importance of image edge and texture features, and the weight of the original image I0(x, y) is set to 0.35, and the weights of the differential images D1(x, y), D2(x, y), D3(x, y) and D4(x, y) are set to 0.25 and 0.20, respectively. , 0.12 and 0.08, the sum of the weights is 1; then for each pixel position (x, y), perform the weighted sum operation M(x, y) = 0.35 × I0(x, y) + 0.25 × D1(x, y) + 0.20 × D2(x, y) + 0.12 × D3(x, y) + 0.08 × D4(x, y) to generate the fused image M(x, y); for the edge information in the fused image, apply the Sobel operator for enhancement processing, the Sobel operator in the horizontal direction is [-1, -2, -1; 0, 0, 0; 1, 2, 1], and the Sobel operator in the vertical direction is [-1, 0, 1; -2, 0, 2; -1, 0, 1], and calculate the gradient size of each pixel point G(x, y) = √(Gx 2 +Gy2), where Gx and Gy are the gradient values ​​in the horizontal and vertical directions respectively; a non-maximum suppression operation is applied to the enhanced image to retain the local gradient maximum point, suppress the non-maximum point, and highlight the boundary characteristics of the stem cells; finally, a histogram equalization operation is performed on the processing result to improve the image contrast. The calculation formula is T(rk)=(L-1) / MN×∑i=0 to k ni, where L is the number of gray levels, M×N is the image size, and ni is the number of pixels with a gray value of i; the multi-level stem cell image representation data obtained by the above operation.

[0080] The present invention reduces image noise and realizes multi-scale image representation through Gaussian filtering and step-by-step downsampling, thereby enhancing the ability to extract information at different scales. By organizing and arranging multi-scale images and constructing a hierarchical pyramid structure, different resolution levels of the image are effectively integrated, providing richer multi-level information for subsequent processing. Through adjacent level differential operations and normalization processing, the detailed features of the image are effectively extracted, and the recognition and expression of image information are improved. Through the fusion of information at each level, multi-level stem cell image representation data is formed, providing comprehensive cell image information for subsequent analysis and processing.

[0081] Preferably, the step S23 of performing adjacent level difference calculation on the hierarchical stem cell image pyramid structure includes:

[0082] Performing interpolation and alignment processing on adjacent levels of the hierarchical stem cell image pyramid structure to obtain scale-matched stem cell image data;

[0083] The embodiment of the present invention performs interpolation alignment processing on adjacent levels of the hierarchical stem cell image pyramid structure to achieve the acquisition of scale-matched stem cell image data. First, an image pyramid method is used to generate multi-level image data, in which each layer of images represents a different scale and has a different resolution. Then, in view of the size difference between adjacent levels, the image of each layer is scale-aligned by an interpolation algorithm (such as bilinear interpolation or cubic interpolation) so that the adjacent level images match in size. This interpolation alignment operation needs to be completed under the premise of ensuring that the characteristic information of the stem cell image is as undistorted as possible. During the operation, it is necessary to accurately control the transition between pixels during the interpolation calculation process to avoid over-smoothing or over-enhancing the image information, and ensure that the alignment process between levels can retain key information. After the interpolation is completed, the scale-matched image data is obtained.

[0084] Based on the scale-matched stem cell image data, pixel-level subtraction operations are performed on adjacent layers to obtain inter-layer difference information data;

[0085] The embodiment of the present invention performs pixel-level subtraction operations of adjacent layers based on scale-matched stem cell image data to extract inter-layer difference information data. First, the scale-matched image data is used as input, and each layer of the image is processed by pixel-level point-by-point operations. In each layer of the image, the corresponding pixel values ​​of the two adjacent layers are taken for difference calculation, that is, the pixel values ​​of the higher resolution layer are subtracted from the pixel values ​​of the lower resolution layer to obtain a difference image. The difference image reflects the changes in image content at different scales and has certain inter-layer information. In order to ensure the accuracy of the calculation process, the difference of each pixel point needs to be processed layer by layer to ensure the pixel alignment accuracy of the two layers of images. During the operation, the calculation accuracy of the subtraction operation can be controlled to ensure that the difference information of the adjacent layers can be fully reflected in the difference image without losing detail information. After completing the subtraction operation, the inter-layer difference information data is obtained.

[0086] Perform nonlinear enhancement processing on the inter-layer difference information data to obtain significant feature area data;

[0087] The embodiment of the present invention first takes the inter-layer difference information data as input. Nonlinear enhancement processing transforms the pixel values ​​in the image by applying a series of nonlinear functions. These nonlinear functions can enhance the relatively weak detail information in the image while maintaining the overall structure of the image. Especially in stem cell images, small structural differences have an important impact on the classification results, so these details need to be made more significant through enhancement processing. The parameter settings in the enhancement process should be adjusted according to the image characteristics. Common adjustment methods include controlling the enhancement intensity and processing range to ensure that the details of the local area are enhanced without affecting the integrity of the global information. In the enhancement process, a layer-by-layer operation method is adopted to perform nonlinear transformation on each layer of the image separately, and finally the processed results of each layer are merged into significant feature area data.

[0088] Based on the data of the significant feature area, the low signal-to-noise ratio area is adaptively filtered by threshold to obtain pure differential feature data;

[0089] The embodiment of the present invention first inputs the significant feature area data, which contains the stem cell image information after nonlinear enhancement processing. Next, the adaptive threshold method is used to process the low signal-to-noise ratio area. The specific process is: by calculating the local statistical characteristics of each pixel in the image, the noise level of the area where each pixel is located is determined. Then, an adaptive threshold is set, and the pixel values ​​below the threshold are judged as noise and filtered out, while the pixels above the threshold are retained to ensure that the important feature areas in the image are not weakened. The threshold is determined based on the local signal-to-noise ratio of each area in the image, and is dynamically adjusted based on the image content, so as to carry out targeted processing of the noise in different areas. The entire filtering process is performed pixel by pixel, and is screened in combination with the local area characteristics, so as to retain the useful difference information in the stem cell image, filter out the noise with low signal-to-noise ratio, and finally obtain pure differential feature data.

[0090] Perform multi-scale hierarchical integration on pure differential feature data to obtain hierarchical feature set data;

[0091] The embodiment of the present invention first inputs pure differential feature data, which contains cell image features after filtering through low signal-to-noise ratio areas. The integration process is completed by fusing the feature data layer by layer at multiple scales. First, according to the image pyramid structure, the pure differential feature data is divided into multiple scale levels, representing different resolutions and image details. At each layer, the feature information is extracted according to its resolution and integrated with the data of its adjacent levels. The integration operation includes pixel-level weighted averaging or interpolation processing to ensure that feature data of different scales can be effectively aligned during integration and maintain the multi-level information of the image. In specific implementation, the influence of image features at each scale is adjusted by a weighting coefficient, and the determination of the weight can be set based on the complexity of image details or the level of resolution. Finally, the features of all scales are summarized and integrated into a unified hierarchical feature set data.

[0092] Based on the hierarchical feature set data, the features of each level are weighted and combined to obtain the differential operation results of adjacent levels.

[0093] The embodiment of the present invention first inputs hierarchical feature set data, which already contains stem cell image features extracted at different scales and resolutions. In order to achieve feature weighted combination, the feature data of each layer is first weighted, and the weight setting is based on the resolution of each layer of the image, the richness of details, and the contribution of the layer to the final classification result. The specific operation includes: calculating the contribution of each layer of features, and the importance of the hierarchical features can be determined by analyzing the pixel information entropy or texture complexity index of each layer. Then, the weighting coefficient is applied to enlarge or reduce the features of each layer to ensure that important scale features can dominate when combined. Next, by adding or pixel-by-pixel splicing the weighted features of each layer, the feature data of different scales are integrated into a unified feature vector or feature map to complete the weighted combination of features at each level. Finally, the difference operation results of adjacent levels are obtained.

[0094] The present invention improves the alignment accuracy of each layer of the image through scale matching to ensure the consistency of image information. It effectively extracts inter-layer differences and highlights the changes in cell structure at different resolutions. By enhancing significant features, the recognition of cell areas is improved, making key information more prominent. By filtering out noise areas, the purity of image data is improved to ensure the accuracy of subsequent analysis. By integrating information at different scales, the multi-dimensional expression of features is enhanced, and the effect of cell recognition is improved. By assigning different weights to features at each level and optimizing feature combinations, the differential operation results are made more accurate.

[0095] Preferably, the multi-resolution hierarchical processing of the multi-level stem cell image representation data in step S3 includes:

[0096] Perform preliminary cell region recognition on the low-resolution layer in the multi-level stem cell image representation to obtain potential cell region data;

[0097] The embodiment of the present invention first extracts the stem cell image data of the low-resolution layer, and uses image processing methods, such as threshold segmentation or edge detection algorithms, to perform preliminary cell region identification. For the low-resolution layer, the cell region in the image is usually fuzzy, so the potential cell region is identified by calculating the brightness or color characteristics of the local area. For example, the Otsu method is used to perform global threshold segmentation on the grayscale image to distinguish the cell region from the background region in the image. For further screening of potential cell regions, morphological processing (such as expansion and corrosion) can be used to remove noise and fill small holes to obtain a more accurate preliminary cell region outline. Next, by calculating the connectivity of adjacent regions, the potential cell regions in the image are separated and marked to ensure that the region of each potential cell can be accurately identified. Finally, the potential cell region data is obtained.

[0098] Performing cell aggregation area analysis on the medium-resolution layer in the multi-level stem cell image representation to obtain cell aggregation area data;

[0099] The embodiment of the present invention first extracts the stem cell image data of the medium resolution layer, and performs image preprocessing, such as noise removal and contrast enhancement, to improve the recognizability of the cell area. Then, a clustering algorithm (such as K-means clustering or density peak clustering) is applied to analyze the cell area in the image to identify the cell aggregation area. The aggregation area refers to the cell area with a higher density in the image. The execution of the clustering algorithm clusters by calculating the pixel value features of each area in the image, and divides the cells in the image into different groups or aggregation areas. In the process of identifying the cell aggregation area, morphological operations (such as expansion and corrosion) are further applied to smooth the edges of the aggregation area, remove isolated noise points, and fill the holes caused by unclear boundaries of the aggregation area. In addition, connectivity analysis is used to ensure that the cell areas in the aggregation area are effectively connected to avoid misjudgment of the aggregation area due to loss of details in the image resolution or processing process. Finally, the cell aggregation area data is obtained.

[0100] Finely process the cell overlap region of the high-resolution layer in the multi-level stem cell image representation to obtain cell overlap region data;

[0101] The embodiment of the present invention first extracts the stem cell image data of the high-resolution layer, which can usually show the tiny details between cells more clearly. The cell overlap area is usually the area where cells contact or overlap each other, and these areas have an important influence on cell classification. In order to refine the cell overlap area, the image is first accurately detected by edge detection, and the Canny edge detection algorithm is used to extract the boundary information of the cell. Then, the boundary is smoothed in combination with morphological operations (such as opening and closing operations) to remove some unclear areas caused by noise or resolution limitations. On this basis, the overlapping area is divided by connectivity analysis to ensure that the cells in the overlapping area are separated. By calculating the contact area and overlap characteristics of the cell boundary, the overlapping area is further refined. In the identification of the overlapping area, the contacting cells are segmented using the region growth algorithm to avoid the overlapping area being misjudged as a single cell. Finally, the cell overlap area data is obtained.

[0102] The potential cell area data, cell aggregation area data and cell overlapping area data are recorded as multi-level cell feature data.

[0103] The embodiment of the present invention first inputs potential cell area data, cell aggregation area data and cell overlapping area data. The potential cell area data comes from the preliminary identification of the cell area of ​​the low-resolution layer, the cell aggregation area data comes from the aggregation area analysis of the medium-resolution layer, and the cell overlapping area data comes from the fine processing of the overlapping area of ​​the high-resolution layer. When these three types of data are represented uniformly, the data of each level are first normalized by an image processing method. Then, the three types of data are fused in hierarchical order using image overlay technology to form a multi-level cell feature set. The fusion process combines the cell area features of each layer with the data of other layers through pixel-level weighted overlay, where the weighting coefficient is adjusted according to the resolution and importance of the data of each layer. In this way, multi-level cell feature data is obtained.

[0104] The present invention can quickly locate potential cell areas and provide a preliminary framework for subsequent detailed analysis. It can identify cell aggregation areas and improve the recognition accuracy of cell aggregation phenomena. It can accurately process cell overlapping areas and solve the recognition problems caused by cell overlap. It can integrate information at different resolution levels into comprehensive cell features to ensure the comprehensiveness and accuracy of subsequent analysis.

[0105] Preferably, step S4 comprises the following steps:

[0106] Step S41: Processing the potential cell region data using a region growing algorithm to obtain initial segmentation region data;

[0107] The embodiment of the present invention first inputs the potential cell region data and extracts the cell region preliminarily identified in the image. The region growing algorithm is processed by selecting seed points, which are usually certain pixels in the potential cell region that can reflect the edge or center of the cell. The algorithm starts from these seed points and gradually expands according to the similarity of the neighborhood pixels (such as grayscale value, color or texture features) until the set threshold conditions are met, such as the similarity measure reaches a predetermined standard or the growth area reaches the maximum preset area. The selection of neighborhood pixels in the expansion process is usually based on the connectivity principle, and the coherence of the region growth is ensured by judging the spatial relationship between pixels (such as 8 neighborhoods or 4 neighborhoods). The region growing algorithm continuously expands the cell region until the entire potential cell region is covered and generates preliminary segmentation region data.

[0108] Step S42: optimizing the cell boundary active contour model based on the initial segmented region data to obtain accurate cell boundary data;

[0109] The embodiment of the present invention first uses the initial segmentation area data as input, and then optimizes the active contour model of the cell boundary. By combining the initial segmentation area data with the active contour model, the boundary accuracy is optimized. The active contour model is optimized based on the image gradient information. By calculating the gradient field of the image, the area with more obvious brightness changes in the image is identified as a preliminary estimate of the cell boundary. Through model iteration, the boundary gradually approaches the actual position of the cell boundary in the image. During the iteration process, the boundary is adjusted according to the change of the image gradient value, and a mechanical model is used to simulate the contraction and expansion of the boundary to ensure that it conforms to the shape characteristics of the cell. The optimization process is achieved by minimizing the energy function. The energy function is calculated based on the difference between the cell boundary and the image feature. The boundary is adjusted in the direction of reducing energy, and finally accurate cell boundary data is obtained.

[0110] Step S43: performing watershed transformation on the cell clustering area data and the cell overlapping area data to obtain separated cell unit data;

[0111] The embodiment of the present invention first uses precise cell boundary data and preliminary segmentation area data to perform watershed transformation on the cell aggregation area data and the cell overlap area data. Watershed transformation is a segmentation method based on image gradient, which is used to process cell areas with different densities. First, the gradient of the cell aggregation area and the cell overlap area images is calculated to obtain the gradient map of the image. Then, the gradient map is converted into a topographic map model, and the segmentation is performed by simulating the process of water flow from a low point to a high point. Each pixel in the image is regarded as the "starting point of the water flow", and gradually expands from the low gray value area to the surrounding area according to the gradient information. When the water flow encounters a high gray value area, the flow is "blocked", forming a watershed line, and then the image is divided into different areas. Finally, the separated cell unit data is obtained.

[0112] Step S44: verifying the integrity of the cell based on the precise cell boundary data and the separated cell unit data to obtain valid cell object data;

[0113] The embodiment of the present invention first verifies the integrity of the cell based on the precise cell boundary data and the separated cell unit data. The process ensures the structural integrity of each cell by analyzing the shape, size and continuity of the boundary of the cell. First, the shape feature analysis is performed on each separated cell unit to calculate the smoothness, convex hull and boundary closure parameters of the cell boundary. If there are cracks or obviously irregular shapes on the boundaries of some cells, the cell is considered to be incomplete and needs to be further processed. At this time, by comparing the boundaries of the precise cell boundary data with the separated cell unit data, it is detected whether the boundaries are consistent. If the boundaries fail to coincide accurately, there is a situation of mis-segmentation or overlap. By setting a threshold, the area and shape of the cell unit are restricted, and those cell units that do not meet the requirements in boundary matching are filtered out. Finally, those cells that are morphologically complete and consistent with the precise boundary data are retained to obtain valid cell object data.

[0114] Step S45: performing shape constraint filtering on the valid cell object data to obtain pure cell set data;

[0115] After obtaining valid cell object data, the embodiment of the present invention performs shape constraint filtering to further improve the accuracy of cell data. By analyzing and filtering the shapes of valid cell objects, cells that do not meet the characteristics of stem cells are excluded. First, shape constraints are set based on the shape characteristics of the cells. By calculating the geometric features of each cell object, it is determined whether it meets the preset cell morphology requirements. For example, for the shape of stem cells, the cell boundary is required to be relatively smooth and close to a circle, and the aspect ratio of the cell is within a reasonable range. For cell objects that do not meet these conditions, they are removed or adjusted based on the results of geometric calculations. For cell objects that meet the shape constraints, they are retained to form pure cell collection data.

[0116] Step S46: reconstructing the original image region labels based on the pure cell collection data to obtain segmented stem cell image data.

[0117] The embodiment of the present invention firstly uses the boundary information of each cell in the pure cell set data to re-mark these cells into the original image. By matching the position of the cell object in the original image with its corresponding marked area, the position and shape of each cell in the image are reconstructed. By extracting the cell contour data in the pure cell set and using these data as a reference, the cell contour is aligned with the original image at the pixel level. For each valid cell object, its corresponding pixel area is extracted from the original image, and the cell area in the original image is accurately reconstructed based on the cell contour, and finally the marked segmented stem cell image data is formed. In this process, each cell area is accurately mapped back to the original image, and the pixel values ​​of these areas are updated to the corresponding cell marking values. The reconstructed image data is the segmented stem cell image data for subsequent analysis or classification.

[0118] The present invention can effectively extract potential cell regions through the region growing algorithm and provide preliminary cell segmentation regions. Through active contour model optimization, cell boundaries can be accurately determined to improve the accuracy of segmentation results. Watershed transformation effectively handles cell aggregation and overlapping areas to achieve accurate separation of cell units. Cell integrity verification ensures the validity of segmentation results and removes incomplete or erroneous cell segmentation regions. Shape constraint filtering excludes areas that do not conform to stem cell morphology to obtain pure cell collection data. Through marker reconstruction, high-quality stem cell image segmentation results are generated to provide reliable data for subsequent analysis.

[0119] Preferably, step S43 includes the following steps:

[0120] Step S431: performing distance transformation calculation on the cell aggregation area data, and extracting the local maximum value of the image area on the distance transformation calculation result to obtain a set of cell seed points;

[0121] The embodiment of the present invention first obtains the cell aggregation area data and performs a distance transformation calculation. The distance transformation generates a distance map by calculating the distance from each pixel point in the image to the nearest background pixel, in which each pixel value represents its distance to the nearest background pixel. For the cell aggregation area data, the purpose of the distance transformation is to highlight the structural information of the cell area. Then, the local maximum value of the image area is extracted for the distance transformation result. The local maximum value extraction searches for the maximum value point in each local area by performing a neighborhood comparison in the distance transformation map. These local maximum value points represent the seed points of the cells, which are usually located at the center of the cell nucleus and can more accurately identify the position of the cells. In order to ensure the quality of the seed points, the local maximum values ​​can be thresholded to remove noise and screen out seed points with higher credibility. Finally, a set of cell seed points is obtained, which contains the core position of each cell in the cell aggregation area.

[0122] Step S432: assigning a unique identifier to the cell seed point set to obtain a cell marking image;

[0123] After obtaining the cell seed point set, the embodiment of the present invention assigns a unique identifier to each seed point. First, each seed point in the cell seed point set will be assigned a unique identifier according to its position in the image. To this end, a pixel coordinate system in the image is used to assign a unique integer identifier to each seed point using a marking algorithm. Each seed point in the image is traversed and a unique identifier is assigned to each seed point one by one according to its spatial position. The assignment of identifiers is determined based on the position in the image. After assigning the identifiers, a cell marking image is generated, in which each cell seed point in the image corresponds to a unique identifier, while the non-seed point area remains as a background or unmarked state.

[0124] Step S433: performing watershed algorithm processing on the cell overlap area data based on the cell marker image, and performing boundary smoothing optimization on the watershed algorithm processing result to obtain fine cell boundary line data;

[0125] The embodiment of the present invention first uses a cell marker image to perform a watershed algorithm on the cell overlap area data. By calculating the gradient map of the cell marker image, the intensity change information of different areas in the image is obtained. Then, the gradient map is converted into a topographic map model, in which the low gray value area represents the depressed area and the high gray value area represents the ridge line. The watershed algorithm searches for boundaries and performs segmentation in the gradient map by simulating the process of water flow expanding from a low gray area to a high gray area. Each cell seed point in the cell marker image is used as the starting point of the watershed algorithm and gradually expands in the topographic map until it encounters the boundary lines of other cells. After watershed segmentation, the preliminary separation area of ​​the cell is obtained. Next, the watershed algorithm processing result is subjected to boundary smoothing optimization. Boundary smoothing is achieved by filtering or curve fitting technology to reduce the jagged boundaries caused by image noise or gradient calculation errors. Finally, fine cell boundary line data is obtained.

[0126] Step S434: marking the connected domain of the separation region based on the fine cell boundary line data to obtain separation cell unit data.

[0127] After obtaining the fine cell boundary line data, the embodiment of the present invention performs separation region connected domain labeling based on these boundary line data. First, using the binarization result of the image (i.e., boundary line data), each independent cell region is marked as a connected region. The connected domain labeling algorithm traverses each pixel in the image to check whether it belongs to the same connected region as the adjacent pixels. If the adjacent pixels belong to the same region, the pixel is merged with the marked region until all cell regions are accurately marked. The connected domain labeling algorithm can be implemented using a depth-first search (DFS) or a breadth-first search (BFS) method, which marks each connected region by recursion or queue. Finally, separation cell unit data is obtained.

[0128] The present invention accurately locates the seed points of cells through distance transformation and local maximum extraction, providing an initial basis for cell segmentation. A unique identifier is assigned to each cell seed point to ensure that each cell can be independently identified in subsequent processing. The watershed algorithm and boundary smoothing optimization effectively separate the overlapping areas of cells and provide accurate cell boundary data. The connected domain marking effectively separates the cell units, laying the foundation for the independent analysis of stem cells.

[0129] Preferably, step S5 comprises the following steps:

[0130] Step S51: Calculate geometric morphological parameters of the segmented stem cell image data to obtain basic cell morphological feature data;

[0131] After obtaining the segmented stem cell image data, the embodiment of the present invention first calculates the geometric morphological parameters of the image. Geometric morphological analysis mainly involves the morphological characteristics of cells in the image. First, by extracting the boundary contour of each cell area, the area and perimeter of each cell are calculated. On this basis, the shape factor (such as the circularity of the cell) is calculated, which reflects the degree of proximity of the cell morphology to the circle. The aspect ratio of the cell is further calculated to evaluate the proportional characteristics of its shape. Using morphological transformation tools such as corrosion, expansion, opening and closing operations, the noise in the image is further removed to extract more accurate cell shape information. Through the calculation of these geometric morphological parameters, the basic morphological feature data of each cell can be extracted.

[0132] Step S52: performing cell region gray-level co-occurrence matrix analysis based on the cell morphology basic feature data to obtain cell texture feature data;

[0133] The embodiment of the present invention performs grayscale co-occurrence matrix analysis of cell regions based on basic cell morphology feature data. First, the segmented stem cell image data is converted into a grayscale image. This conversion process involves converting the RGB channel data of the color image into grayscale levels. Then, the grayscale co-occurrence matrix (GLCM) of the cell region in the image is calculated, which is used to describe the grayscale relationship between pixel pairs in the image. The calculation of the grayscale co-occurrence matrix involves counting the frequency of each pair of pixels (i, j) in the image at a specified direction and distance. After obtaining the grayscale co-occurrence matrix, multiple texture feature indicators are further calculated, including contrast, correlation, energy, and uniformity. Through these calculations, data describing cell texture features are obtained.

[0134] Step S53: Calculate the spatial density of cell regional distribution based on the cell texture feature data to obtain cell density feature data;

[0135] The embodiment of the present invention first extracts the boundary information of the cell region based on the segmented stem cell image data based on the cell texture feature data, and subdivides the cell region to calibrate the position and region size of each cell. The spatial distribution density is calculated using these cell region information. The calculation process includes dividing the image region into a plurality of small grid cells and counting the number of cells in each small cell. The distribution density of the cells in space is obtained by calculating the number of cells in each grid cell. At the same time, in order to ensure the accuracy of the density calculation, a weighted calculation method can be applied in a specific cell aggregation area or sparse area according to the boundary and morphological characteristics of the cells to improve the accuracy of the density estimation. Finally, the density characteristic data of the cell region is obtained according to the density distribution of cells in different regions.

[0136] Step S54: performing multi-dimensional feature fusion on the cell morphology basic feature data, the cell texture feature data and the cell density feature data to obtain a comprehensive cell feature vector;

[0137] The embodiment of the present invention performs multi-dimensional feature fusion on the basic feature data of cell morphology, cell texture feature data and cell density feature data, and standardizes the morphological features, texture features and spatial density feature data of each cell to ensure that each type of feature is consistent in the numerical range. The standardized data is spliced ​​to form a multi-dimensional vector, which contains the geometric morphology, texture pattern and spatial distribution density characteristics of the cell. In the fusion process, in order to avoid the scale differences of different features affecting the final fusion effect, a weighted summation of different feature data can be used to ensure that each feature occupies an appropriate proportion in the comprehensive process. Through this fusion process, the multi-dimensional features of the cell are converted into a unified comprehensive cell feature vector.

[0138] Step S55: Perform support vector machine classification processing on the comprehensive cell feature vector to obtain primary cell classification data;

[0139] The embodiment of the present invention performs support vector machine (SVM) classification processing on the comprehensive cell feature vector to obtain primary cell classification data. First, a support vector machine model is constructed using the fused multidimensional feature vector as input. The support vector machine divides the cell feature space into different categories by constructing an optimal hyperplane, and then assigns each cell to a corresponding category. During the training process, a linear kernel function is used for feature mapping to map low-dimensional data to a high-dimensional feature space to improve the accuracy of classification. After the training is completed, the comprehensive feature vector of the new cell sample is input into the support vector machine model, and the position of the hyperplane and the category boundary are predicted to obtain primary cell classification data.

[0140] Step S56: Perform deep learning segmentation of cell subclasses based on the primary cell classification data to obtain stem cell classification result data.

[0141] In the embodiment of the present invention, first, a deep neural network model is established using primary cell classification data as input. The input layer of the network model receives feature information in the primary cell classification data, and is processed by a multi-layer convolutional neural network (CNN) to extract deeper cell features. The convolutional layer and pooling layer in the network are optimized through multiple iterations to analyze the details of the cell image to distinguish different cell subclasses. In order to enhance the expressiveness of the model, a residual connection (ResNet) is used to solve the gradient disappearance problem in deep network training. The output layer of the network is classified according to a predetermined number of categories, and finally the classification result data of stem cells is output. To ensure the accuracy of the deep learning model, a large-scale data set is used for training, and hyperparameters are adjusted through cross-validation to avoid overfitting. During the training process, an optimization algorithm (such as the Adam optimizer) is used to update the weights.

[0142] The present invention extracts the basic morphological features of cells through geometric morphological parameter calculation, providing basic morphological information for subsequent analysis. Through grayscale co-occurrence matrix analysis, the texture features of the cell area are revealed, and the texture information of the cell image is provided. The distribution space density of the cell area is calculated to reflect the spatial distribution characteristics of the cell. Through multidimensional feature fusion, the cell morphology, texture and density information are combined to generate a more comprehensive comprehensive cell feature vector. A support vector machine is used for classification processing to achieve preliminary classification of stem cells and distinguish different cell types. Cell subclasses are further subdivided through deep learning to obtain more refined stem cell classification results.

[0143] Preferably, step S6 comprises the following steps:

[0144] Step S61: performing data structure optimization conversion on stem cell classification result data to obtain standardized storage format data;

[0145] The embodiment of the present invention performs data structure optimization conversion on the stem cell classification result data. First, a standardized data structure is defined according to the characteristic data of the stem cell classification result. Through structured design, the storage format of all classification result data is unified, for example, using a tabular data structure or JSON format. Then, based on predefined fields, such as cell categories, characteristic values, and identifiers, standardized conversion of data fields is performed to avoid format inconsistencies between different data sources. At the same time, the numerical data in the classification results is standardized. Then, for each cell category and its characteristic data, a suitable data compression and encoding method (such as Huffman encoding or Base64 encoding) is applied to further optimize data storage, reduce data redundancy, and improve storage efficiency. After these processes, standardized storage format data is finally obtained.

[0146] Step S62: constructing a multidimensional index of cell classification information based on the standardized storage format data to obtain a cell classification retrieval index;

[0147] The embodiment of the present invention first constructs a multidimensional index structure of cell classification information based on standardized storage format data. Suitable data fields, such as cell categories, cell morphological features, texture features and density features, are selected as index dimensions. By indexing each dimension, it is ensured that each cell classification data can be quickly retrieved from multiple angles. During the construction process, a B-tree or KD-tree data structure is used to spatially divide the cell data and sort them according to cell categories and related features. Then, based on the different categories of data in the cell classification results, an index of cell categories is constructed, and these index information are further associated with the original data. In addition, for cell data with high-dimensional features, a clustering algorithm is used to perform dimensionality reduction processing while retaining the key features of the cell data. Next, for each dimensional index, an incremental update process is performed to ensure that the index structure can be updated in a timely manner as new data is added. Through these operations, an efficient, dynamically updateable cell classification retrieval index is finally formed.

[0148] Step S63: performing hash partitioning processing on the cell classification retrieval index, and performing shard storage on the standardized storage format data based on the hash partitioning processing result to obtain a multi-node stem cell data cluster;

[0149] The embodiment of the present invention first performs hash partitioning based on the cell classification retrieval index. First, a relatively stable and highly distinguishable feature field in the cell classification information, such as cell category, morphological parameter or texture feature, is selected as the basis for hash partitioning. For these features, a consistent hash algorithm (such as MD5, SHA-256 or a custom hash function) is applied to perform hash calculation on each cell classification information to obtain a corresponding hash value. According to the hash value, the data is divided into multiple data partitions. Each partition contains a portion of data. According to the distribution of the hash value, the data is evenly distributed between each partition, thereby avoiding excessive load on some partitions. After the hash partition is completed, the cell data of each partition is stored in slices. Each partition data is stored in a separate storage unit, such as a database table, a file system or a distributed storage node, and each storage unit is regarded as an independent slice. When storing, the storage path is mapped according to the hash value of the partition. Finally, a multi-node stem cell data cluster is formed.

[0150] Step S64: performing consistency verification processing on the multi-node stem cell data cluster and metadata integration to obtain a persistent stem cell classification database.

[0151] The embodiment of the present invention first performs a consistency check process on a multi-node stem cell data cluster. During the check process, a hash algorithm is used to re-hash the content of each data shard. Specifically, the same hash algorithm as that used for hash partitioning is used to compare the hash value of each data shard with the original value one by one to see if it matches. If the data is found to be inconsistent, the data recovery mechanism is triggered to restore the lost or damaged part from the original data source or backup data. After the consistency check is completed, metadata integration processing is performed. Metadata integration classifies and organizes the index information, storage path, and data type metadata of each data shard to establish a unified metadata management process. The metadata of each shard is recorded in a special metadata table. At this point, the storage information of all cell classification data has been fully integrated according to a specific storage structure and index system to form a complete persistent stem cell classification database.

[0152] The present invention ensures the standardization of the storage format of stem cell classification results through data structure optimization and conversion, and improves the compatibility and operability of data processing. Through multidimensional index construction, the retrieval efficiency of cell classification information is improved, and fast and accurate classification data query is supported. Through hash partition processing and shard storage, data is distributed to multiple nodes, data storage management is optimized, and system scalability and load balancing are improved. Through consistency verification and metadata integration, the integrity and accuracy of the data are ensured, and it is persisted as a reliable stem cell classification database.

[0153] Preferably, the present invention further provides a stem cell classification system based on image segmentation, which is used to execute the above-mentioned stem cell classification method based on image segmentation, and the stem cell classification system based on image segmentation comprises:

[0154] The image preprocessing and standardization module is used to obtain the original stem cell image data, perform Wasm processing on the original stem cell image data, and obtain the stem cell image quality parameters; based on the stem cell image quality parameters, the original stem cell image data is adaptively preprocessed to obtain standardized stem cell image data;

[0155] A multi-resolution pyramid construction module is used to construct a multi-resolution pyramid for the standardized stem cell image data to obtain multi-level stem cell image representation data;

[0156] A multi-resolution hierarchical feature extraction module is used to perform multi-resolution hierarchical processing on the multi-level stem cell image representation data to obtain multi-level cell feature data, wherein the multi-level cell feature data includes potential cell area data, cell aggregation area data and cell overlap area data;

[0157] A stem cell adaptive segmentation module is used to perform stem cell adaptive algorithm segmentation based on multi-level cell feature data to obtain segmented stem cell image data;

[0158] The morphological feature extraction and classification module is used to extract morphological features from the segmented stem cell image data, and perform hierarchical classification processing based on the morphological feature extraction results to obtain stem cell classification result data;

[0159] The distributed storage and persistence module is used to perform distributed storage processing on stem cell classification result data and generate a persistent stem cell classification database.

[0160] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is not limited by the above description, so it is intended to include all changes within the meaning and scope of the equivalent elements of the application documents in the present invention.

[0161] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A stem cell classification method based on image segmentation, characterized in that: The following steps are involved: Step S1: acquiring original stem cell image data, performing Wasm processing on the original stem cell image data to obtain stem cell image quality parameters; performing adaptive preprocessing on the original stem cell image data based on the stem cell image quality parameters to obtain standardized stem cell image data; Step S2: constructing a multi-resolution pyramid for the standardized stem cell image data to obtain multi-level stem cell image representation data; Step S3: performing multi-resolution hierarchical processing on the multi-level stem cell image representation data to obtain multi-level cell feature data, wherein the multi-level cell feature data includes potential cell region data, cell aggregation region data and cell overlap region data; Step S4: performing stem cell adaptive algorithm segmentation based on the multi-level cell feature data to obtain segmented stem cell image data; Step S5: extracting morphological features from the segmented stem cell image data, and performing hierarchical classification processing based on the morphological feature extraction results to obtain stem cell classification result data; Step S6: Perform distributed storage processing on the stem cell classification result data to generate a persistent stem cell classification database.

2. The stem cell classification method based on image segmentation according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: acquiring original stem cell image data, and compiling the original stem cell image data using a Wasm module to obtain a high-performance image processing module; Step S12: performing quality assessment on the original stem cell image data based on a high-performance image processing module to obtain stem cell image quality parameters; Step S13: performing image quality analysis and preprocessing strategy determination on stem cell image quality parameters to obtain an optimal preprocessing strategy; Step S14: performing image noise reduction and contrast enhancement on the original stem cell image data based on the optimal preprocessing strategy to obtain enhanced stem cell image data; Step S15: performing illumination correction and edge optimization on the enhanced stem cell image data to obtain standardized stem cell image data.

3. The stem cell classification method based on image segmentation according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing Gaussian filtering and step-by-step downsampling processing on the standardized stem cell image data to obtain a multi-scale stem cell image sequence; Step S22: organizing and arranging different resolution levels based on the multi-scale stem cell image sequence to obtain a hierarchical stem cell image pyramid structure; Step S23: performing adjacent-level differential operations on the hierarchical stem cell image pyramid structure, and normalizing the adjacent-level differential operation results to obtain standardized feature representation data; Step S24: performing information fusion processing of each level based on the standardized feature representation data to obtain multi-level stem cell image representation data.

4. The stem cell classification method based on image segmentation according to claim 3, characterized in that: The step S23 of performing adjacent level difference operation on the hierarchical stem cell image pyramid structure includes: Performing interpolation and alignment processing on adjacent levels of the hierarchical stem cell image pyramid structure to obtain scale-matched stem cell image data; Based on the scale-matched stem cell image data, pixel-level subtraction operations are performed on adjacent layers to obtain inter-layer difference information data; Perform nonlinear enhancement processing on the inter-layer difference information data to obtain significant feature area data; Based on the data of the significant feature area, the low signal-to-noise ratio area is adaptively filtered by threshold to obtain pure differential feature data; Perform multi-scale hierarchical integration on pure differential feature data to obtain hierarchical feature set data; Based on the hierarchical feature set data, the features of each level are weighted and combined to obtain the differential operation results of adjacent levels.

5. The stem cell classification method based on image segmentation according to claim 1, characterized in that: The multi-resolution hierarchical processing of the multi-level stem cell image representation data in step S3 includes: Perform preliminary cell region recognition on the low-resolution layer in the multi-level stem cell image representation to obtain potential cell region data; Performing cell aggregation area analysis on the medium-resolution layer in the multi-level stem cell image representation to obtain cell aggregation area data; Finely process the cell overlap region of the high-resolution layer in the multi-level stem cell image representation to obtain cell overlap region data; The potential cell area data, cell aggregation area data and cell overlapping area data are recorded as multi-level cell feature data.

6. The stem cell classification method based on image segmentation according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Processing the potential cell region data using a region growing algorithm to obtain initial segmentation region data; Step S42: optimizing the cell boundary active contour model based on the initial segmented region data to obtain accurate cell boundary data; Step S43: performing watershed transformation on the cell clustering area data and the cell overlapping area data to obtain separated cell unit data; Step S44: verifying the integrity of the cell based on the precise cell boundary data and the separated cell unit data to obtain valid cell object data; Step S45: performing shape constraint filtering on the valid cell object data to obtain pure cell set data; Step S46: reconstructing the original image region labels based on the pure cell collection data to obtain segmented stem cell image data.

7. The stem cell classification method based on image segmentation according to claim 6, characterized in that: Step S43 includes the following steps: Step S431: performing distance transformation calculation on the cell aggregation area data, and extracting the local maximum value of the image area on the distance transformation calculation result to obtain a set of cell seed points; Step S432: assigning a unique identifier to the cell seed point set to obtain a cell marking image; Step S433: performing watershed algorithm processing on the cell overlap area data based on the cell marker image, and performing boundary smoothing optimization on the watershed algorithm processing result to obtain fine cell boundary line data; Step S434: marking the connected domain of the separation region based on the fine cell boundary line data to obtain separation cell unit data.

8. The stem cell classification method based on image segmentation according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: Calculate geometric morphological parameters of the segmented stem cell image data to obtain basic cell morphological feature data; Step S52: performing cell region gray-level co-occurrence matrix analysis based on the cell morphology basic feature data to obtain cell texture feature data; Step S53: Calculate the spatial density of cell regional distribution based on the cell texture feature data to obtain cell density feature data; Step S54: performing multi-dimensional feature fusion on the cell morphology basic feature data, the cell texture feature data and the cell density feature data to obtain a comprehensive cell feature vector; Step S55: Perform support vector machine classification processing on the comprehensive cell feature vector to obtain primary cell classification data; Step S56: Perform deep learning segmentation of cell subclasses based on the primary cell classification data to obtain stem cell classification result data.

9. The stem cell classification method based on image segmentation according to claim 1, characterized in that: Step S6 includes the following steps: Step S61: performing data structure optimization conversion on stem cell classification result data to obtain standardized storage format data; Step S62: constructing a multidimensional index of cell classification information based on the standardized storage format data to obtain a cell classification retrieval index; Step S63: performing hash partitioning processing on the cell classification retrieval index, and performing shard storage on the standardized storage format data based on the hash partitioning processing result to obtain a multi-node stem cell data cluster; Step S64: performing consistency verification processing on the multi-node stem cell data cluster and integrating metadata to obtain a persistent stem cell classification database.

10. A stem cell classification system based on image segmentation, characterized in that: For executing the stem cell classification method based on image segmentation as claimed in claim 1, the stem cell classification system based on image segmentation comprises: The image preprocessing and standardization module is used to obtain the original stem cell image data, perform Wasm processing on the original stem cell image data, and obtain the stem cell image quality parameters; based on the stem cell image quality parameters, the original stem cell image data is adaptively preprocessed to obtain standardized stem cell image data; A multi-resolution pyramid construction module is used to construct a multi-resolution pyramid for the standardized stem cell image data to obtain multi-level stem cell image representation data; A multi-resolution hierarchical feature extraction module is used to perform multi-resolution hierarchical processing on the multi-level stem cell image representation data to obtain multi-level cell feature data, wherein the multi-level cell feature data includes potential cell area data, cell aggregation area data and cell overlap area data; A stem cell adaptive segmentation module is used to perform stem cell adaptive algorithm segmentation based on multi-level cell feature data to obtain segmented stem cell image data; The morphological feature extraction and classification module is used to extract morphological features from the segmented stem cell image data, and perform hierarchical classification processing based on the morphological feature extraction results to obtain stem cell classification result data; The distributed storage and persistence module is used to perform distributed storage processing on stem cell classification result data and generate a persistent stem cell classification database.

Images