Cellular pathological smear digital slice visible component separation and lightweight processing system and method based on gray threshold layering

Abstract

The invention provides a visible component separation and lightweight processing system and method for a digital section of a cell pathological smear based on gray threshold layering, and is applied to the technical field of medical data processing. The method comprises the following steps: processing a gray interval standard of preset cells and interstitial components and digital section image data of a cell pathological smear, and generating a cell component mask and an interstitial component mask by extracting a gray value of an image B channel; performing optimization processing on the generated cell component masks and interstitial component masks by utilizing a morphological algorithm, removing noisy points through corrosion and expansion operation of a 5 * 5 convolution kernel, extracting connected components by adopting a four-neighborhood expansion region growing algorithm, and generating optimized cell and interstitial active components; and carrying out lightweight reconstruction processing on the optimized effective components of the cells and the interstitial substances, removing redundant backgrounds in the original slices, compressing the data volume to be less than 1 / 10 of the original data volume, and generating independent RGB image views of the cell components and the interstitial components.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, and in particular to a system and method for the separation and lightweight processing of formed elements in digital sections of cytopathological smears based on grayscale threshold layering. Background Technology

[0002] Traditionally, pathologists manually examine smear samples under a microscope to determine cell morphology, structure, and distribution. However, this method has many limitations. On the one hand, manual interpretation relies heavily on the experience and expertise of the physician, and diagnostic results may vary significantly between different physicians, leading to low diagnostic consistency. On the other hand, manual interpretation is inefficient; when faced with a large number of samples, it not only consumes a significant amount of the physician's time and energy but is also prone to missed or misdiagnosed cases due to fatigue and other factors.

[0003] With the development of digital technology, digital slide technology has emerged. Through a full-slide imaging system, cytopathological smears can be converted into digital images, facilitating storage, transmission, and analysis. However, existing digital slide technology still faces many challenges in processing cytopathological smears. For example, cells and stromal components are intertwined in cytopathological smears, making accurate separation difficult with traditional image segmentation algorithms. While some threshold-based segmentation algorithms can distinguish foreground from background to some extent, they are ineffective for cells and stromal components with similar grayscale values ​​and complex features, easily leading to confusion between cells and stromal components and affecting the accuracy of subsequent diagnoses. Furthermore, digital slide data is enormous; a single full-field cytopathological smear digital slide can reach several gigabytes in size. This places extremely high demands on storage devices and transmission networks, making rapid storage and transmission difficult in resource-limited settings such as primary hospitals, severely restricting applications such as remote consultations.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this application is to provide a system and method for the separation and lightweight processing of formed elements in digital slides of cytopathological smears based on grayscale threshold layering. This system overcomes, to some extent, the problems existing in the prior art. It processes digital slides of cytopathological smears based on grayscale threshold layering, separating components through B-channel grayscale differences (cells 2-150, stroma 151-220). Noise is denoised using a 5×5 convolution kernel, and ≥10 pixel components are extracted using a four-neighbor algorithm. The 1000×1000 block storage is compressed to less than 1 / 10 of its original size, generating independent RGB views that support overlay. It is compatible with ordinary PCs, reducing storage by 90%, achieving second-level transmission, increasing slide reading efficiency by 85.1%, and raising the diagnostic consistency Kappa value to 0.89.

[0006] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0007] According to one aspect of this application, a method for separating formed elements and lightweight processing of digital slides of cytopathological smears based on grayscale threshold layering is provided, comprising: acquiring digital slide image data of cytopathological smears; generating preset grayscale range standards for cells and stromal components based on grayscale threshold layering information through image enhancement and standardization preprocessing, wherein the grayscale value range of cell components is 2-150, and the grayscale value range of stromal components is 151-220; the digital slide image data of cytopathological smears includes pleural effusion smears, peritoneal lavage fluid smears, and thyroid biopsy smears; processing the preset grayscale range standards for cells and stromal components and the digital slide image data of cytopathological smears, and generating cell component masks and stromal component masks by extracting the grayscale values ​​of the B channel of the image. A mask is used to initially separate cellular and stromal components. Morphological algorithms are then used to optimize the generated cellular and stromal component masks. Noise is removed by erosion and dilation operations using 5×5 convolutional kernels. A four-neighborhood expansion region growing algorithm is then used to extract connected components, filtering out invalid components smaller than 10 pixels to generate optimized effective cellular and stromal components. The optimized effective cellular and stromal components undergo lightweight reconstruction to remove redundant background from the original slices. A 1000×1000 pixel block strategy is used for storage, compressing the data volume to less than 1 / 10 of the original, generating independent RGB image views of cellular and stromal components. Independent visualization or overlay analysis of the two types of components is achieved through a component label association mechanism.

[0008] Another aspect of this application provides a device for separating and lightweighting cellular pathology smear digital slides based on grayscale threshold layering, comprising: an acquisition module for acquiring digital slide image data of cellular pathology smears, generating preset grayscale range standards for cells and stromal components based on grayscale threshold layering information through image enhancement and standardization preprocessing, wherein the grayscale value range for cell components is 2-150, and the grayscale value range for stromal components is 151-220, and the digital slide image data of cellular pathology smears includes pleural effusion smears, peritoneal lavage fluid smears, and thyroid biopsy smears; and a processing module for processing the preset grayscale range standards for cells and stromal components and the digital slide image data of cellular pathology smears, generating a cell component mask by extracting the grayscale value of the B channel of the image. Cellular and stromal component masks are used to initially separate cellular and stromal components. Morphological algorithms are then used to optimize the generated cellular and stromal component masks. Noise is removed by erosion and dilation operations using 5×5 convolutional kernels, and a four-neighborhood expansion region growing algorithm is used to extract connected components, filtering out invalid components smaller than 10 pixels to generate optimized effective cellular and stromal components. The optimized effective cellular and stromal components are then subjected to lightweight reconstruction to remove redundant background from the original slices. A 1000×1000 pixel block strategy is used for storage, compressing the data volume to less than 1 / 10 of the original, generating independent RGB image views of cellular and stromal components. Independent visualization or overlay analysis of the two types of components is achieved through a component label association mechanism.

[0009] According to another aspect of this application, an electronic device includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the executable instructions to perform the above-described method for separating and lightweighting formed elements in digital slides of cytopathological smears based on grayscale threshold layering.

[0010] According to another aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a second processor, implements the above-described method for separating and lightweighting formed elements in digital sections of cytopathological smears based on grayscale threshold layering.

[0011] This application provides a system and method for the separation and lightweight processing of formed elements in digital slides of cytopathological smears based on grayscale threshold layering. It automates the processing of digital slides of cytopathological smears by using grayscale threshold layering, with the core technology separating cells and stromal components through B-channel grayscale differences. Digital slides of pleural effusion, peritoneal lavage fluid, and thyroid biopsy smears are acquired. After preprocessing, based on HE or Papanicolaou staining characteristics, dual threshold ranges (2-150) for cells and (151-220) for stromal components are set. The grayscale values ​​of the B-channel are extracted to generate a mask, which is then denoised using a 5×5 convolution kernel and a four-neighbor algorithm to extract effective components of ≥10 pixels. The data is stored in 1000×1000 blocks, and after redundancy removal, the data is compressed to less than 1 / 10 of its original size, generating independent RGB views for overlay analysis. The technology is compatible with ordinary PCs, reducing storage costs by 90%, reducing transmission time to a few seconds, improving slide reading efficiency by 85.1%, and increasing the diagnostic consistency Kappa value from 0.72 to 0.89.

[0012] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0013] Figure 1 The flowchart illustrates a method for separating formed elements and lightweighting cytopathological smear digital sections based on grayscale threshold layering, according to an embodiment of this application. Figure 2 This illustration shows an independent RGB view containing cellular and interstitial components provided in an embodiment of this application; Figure 3 This illustration shows a schematic diagram of a device for separating and lightweighting formed elements in digital slides of cytopathological smears based on grayscale threshold layering, according to an embodiment of this application. Detailed Implementation

[0014] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0015] The following is combined with Figure 1 This application describes a method for separating and lightweighting formed elements in digital slides of cytopathological smears based on grayscale threshold layering, according to an exemplary embodiment of this application. It should be noted that the following application scenarios are shown only to facilitate understanding of the spirit and principles of this application, and the embodiments of this application are not limited in any way. Rather, the embodiments of this application are applicable to any suitable scenario.

[0016] In one implementation, Figure 1The illustration schematically shows a process flow diagram of a method for separating formed elements and lightweighting cytopathological smear digital sections based on grayscale threshold layering according to an embodiment of this application.

[0017] S101: Acquire digital slide image data of cytopathological smears, and generate preset gray-level interval standards for cells and stromal components based on gray-level threshold layering information through image enhancement and standardization preprocessing.

[0018] In one implementation, the digital slide image data of cytopathological smears is analyzed using the principle of gray-level threshold layering to clarify the gray-level differences between cells and stromal components, generating preset gray-level value ranges for cellular components and stromal components. After acquiring a digital slide of a thyroid biopsy smear (GB-level high-resolution image with billions of pixels), the original data is first optimized through image enhancement and standardization preprocessing: transparent pixels in the slide are processed (the gray-level value of transparent areas is uniformly set to 255 to eliminate interference), and a downsampling factor (e.g., 4 in the example) is set to balance image resolution and processing efficiency.

[0019] After preprocessing, the grayscale distribution of image pixels was analyzed based on the principle of grayscale thresholding. The focus was on extracting the grayscale data of the B channel (blue channel). Statistical analysis revealed that in thyroid biopsy smears, areas labeled as typical cellular components such as cell nuclei and epithelial cells had B channel grayscale values ​​mainly distributed in the range of 2-150; while areas labeled as stromal components such as connective tissue and matrix had B channel grayscale values ​​concentrated in the range of 151-220. This difference in grayscale characteristics is directly related to the staining characteristics of cells and stroma under HE staining or Papanicolaou staining (cellular components have stronger adsorption of staining agents, resulting in lower B channel grayscale values; stromal components have weaker adsorption, resulting in higher B channel grayscale values).

[0020] Based on the above analysis results, a preset grayscale range standard for thyroid puncture smears was generated: the grayscale value range for cellular components was clearly defined as 2-150, and the grayscale value range for interstitial components was clearly defined as 151-220, providing a core judgment basis for the subsequent automated separation of cells and interstitial components.

[0021] In the process of matching the B-channel gray value data of HE-stained or Papanicolaou-stained pathological images with the dual threshold range standard (2-150 for cellular components and 151-220 for stromal components), the gray value data of the B-channel (blue channel) is first extracted based on the characteristics of HE-stained or Papanicolaou-stained pathological images. Under these two staining methods, cellular components (such as cell nuclei and epithelial cells) have a lower B-channel gray value because they have a stronger adsorption of the staining agent, while stromal components (such as connective tissue and matrix) have a higher B-channel gray value because they have a weaker adsorption, thus forming a natural gray value difference.

[0022] Based on this, a component classification benchmark is constructed: the gray value data of the B channel is matched one by one with the preset dual threshold range. If the gray value of the B channel of a pixel falls within the range of 2-150, the pixel is determined to be a cellular component; if it falls within the range of 151-220, it is determined to be a stromal component, thereby establishing the correspondence between "gray value - component type".

[0023] Through the above grayscale value range matching logic, the preliminary classification rules for cells and interstitial components are further clarified: for example, in a HE-stained pleural effusion smear, the grayscale value of a certain pixel in channel B is 90, which meets the cell component range standard and is marked as a cell component; in a Papanicolaou-stained thyroid puncture smear, the grayscale value of a certain pixel in channel B is 170, which meets the interstitial component range standard and is marked as an interstitial component.

[0024] Ultimately, by solidifying the matching logic and segmentation rules, a component recognition mechanism based on grayscale threshold stratification was established, enabling automated and standardized preliminary separation of cellular and stromal components in pathological images, providing basic data support for subsequent optimization processing.

[0025] Using pathological smear type as the coverage dimension, image data of cytopathological types including pleural effusion smears, peritoneal lavage fluid smears, and thyroid biopsy smears are integrated. Combined with grayscale threshold range standards, the system defines the objects to be processed in digital slide cytopathology. In the process of processing based on pathological smear type, it is first clarified that the system's processing objects cover all clinically common cytopathological types, such as pleural effusion smears, peritoneal lavage fluid smears, and thyroid biopsy smears. These smears are all digitized into high-resolution images (with a pixel scale of billions and original data volume in the gigabyte range) using a whole-slide imaging system.

[0026] When fusing image data of various pathological types, a pre-defined grayscale threshold range standard (2-150 for cellular components and 151-220 for stromal components) is uniformly used for screening based on the characteristics of different smears: In pleural effusion smears, the gray values ​​of the B channel of cellular components such as inflammatory cells and cancer cells were found to be concentrated in the range of 2-150, while the gray values ​​of stromal components such as fibrin were concentrated in the range of 151-220, which met the processing standards. In peritoneal lavage fluid smears, epithelial cells with a B channel gray value of 80-130 (belonging to the cellular range) and matrix components with a gray value of 160-210 (belonging to the interstitial range) are included in the treatment range; In thyroid biopsy smears, the gray values ​​of follicular epithelial cells are 30-140 (cellular range), and the gray values ​​of connective tissue are 155-200 (interstitial range), which meets the processing conditions.

[0027] Through the above process, all cytopathological image data such as pleural effusion smears, peritoneal lavage fluid smears, and thyroid puncture smears that meet the grayscale standard of "cellular components 2-150, interstitial components 151-220" are uniformly defined as the processing objects of the system, ensuring that subsequent component separation and lightweight processing can cover mainstream clinical cytopathological scenarios and achieve standardized and universal technical application.

[0028] High-resolution digital slide data is loaded using image enhancement and standardization preprocessing mechanisms. This is combined with grayscale thresholding to enhance the specificity of B-channel grayscale value extraction. This is then integrated with pre-defined component grayscale range standards to generate initial criteria for separating cells from stromal components. When loading high-resolution digital slide data using image enhancement and standardization preprocessing mechanisms, preprocessing operations are first performed specifically for the characteristics of cytopathological smear digital slides: transparent pixels in the slides are processed (the grayscale value of transparent areas is uniformly set to 255 to eliminate meaningless transparent background interference), and a downsampling factor is set. This balances image resolution and processing efficiency while preserving key diagnostic details, ensuring that high-resolution digital slides (billions of pixels) can be efficiently loaded and parsed.

[0029] After preprocessing, the core logic of grayscale threshold layering technology is combined to enhance the targeting of B channel grayscale value extraction—focusing on the B channel (blue channel) in HE staining or Papanicolaou staining pathological images that is closely related to cells and stromal components, and accurately extracting the grayscale value data of this channel through algorithms, ignoring information in other channels that is not significant for component separation, thereby improving the targeting of data extraction.

[0030] Subsequently, the extracted B-channel grayscale data is fused with the preset component grayscale range standard (cellular components 2-150, interstitial components 151-220): the B-channel grayscale value of each pixel in the loaded slice data is matched within the range. If the grayscale value falls within the range of 2-150, it is initially determined to be a cellular component; if it falls within the range of 151-220, it is initially determined to be an interstitial component.

[0031] Through the above process, the initial judgment criteria for separating cells and interstitial components are finally generated, namely, the preliminary component segmentation mask based on the gray-scale matching results. This provides basic data support for subsequent morphological optimization processing (such as noise removal and connected component extraction), ensuring that the accuracy of component separation is guaranteed from the initial stage.

[0032] S102, the preset grayscale range standards for cells and stromal components and digital slide image data of cytopathological smears are processed. By extracting the grayscale values ​​of the B channel of the image, cell component masks and stromal component masks are generated to achieve preliminary separation of cells and stromal components.

[0033] In one implementation, a grayscale threshold layering process is used to perform feature matching between preset grayscale range standards for cells and stromal components and digital slide image data of cytopathological smears. This extracts the B-channel grayscale values ​​of HE-stained or Papanicolaou-stained pathological images, generating cellular component masks and stromal component masks to achieve preliminary component segmentation. For a single HE-stained pleural effusion smear digital slide (digitally generated by a whole-slide imaging system, with an original data size of 5.4GB and a pixel scale of billions), when the grayscale threshold layering process is invoked, the data is first loaded through an image preprocessing module and standardized optimization is performed (processing transparent pixels, uniformly setting the grayscale value of transparent areas to 255 to eliminate background interference).

[0034] During the feature extraction stage, the system focuses on the B channel (blue channel) of the HE-stained image and uses an algorithm to accurately extract the gray value data of this channel. Because cellular components (such as cancer cells and inflammatory cells) in HE staining have a strong adsorption of hematoxylin (blue dye), the gray value of the B channel is relatively low; while stromal components (such as fibrin and matrix) have a weaker adsorption, the gray value of the B channel is relatively high. This provides a basis for component separation.

[0035] Subsequently, the system performs point-by-point feature matching between the extracted B-channel grayscale value of each pixel and the preset dual-threshold interval standard: if a pixel has a grayscale value of 70, which falls within the cellular component interval (2-150), it is determined to be a cellular component, and the pixel is marked as 255 (white) in the cellular component mask; if another pixel has a grayscale value of 190, which falls within the interstitial component interval (151-220), it is determined to be an interstitial component, and the pixel is marked as 255 (white) in the interstitial component mask.

[0036] In the binary mask generated by the above logic, pixels in non-target regions (such as grayscale values ​​of 1 or 221 and above) are marked as 0 (black), ultimately forming a clearly distinguishable cellular component mask and interstitial component mask, realizing the automated preliminary division of cellular and interstitial components, laying the foundation for subsequent morphological optimization.

[0037] By connecting the dual-threshold interval standard with the B-channel grayscale data of the image, a component separation benchmark model is constructed. Pixel affiliation weights are calculated through grayscale value interval matching logic, establishing an automated identification mechanism for cellular and stromal components. When connecting the generated B-channel grayscale data with the preset dual-threshold interval standard (cellular 2-150, stromal 151-220), the component separation benchmark model is first constructed based on the core algorithm of the grayscale threshold layering module. This model clarifies pixel affiliation through binary weight assignment logic: setting "pixel affiliation weight = 1" indicates that the pixel completely belongs to a certain component, and "weight = 0" indicates that it does not belong to that component at all.

[0038] For example, for HE-stained peritoneal lavage fluid smear B channel data: if a pixel has a gray value of 50, which falls within the cellular component range (2-150), the model automatically assigns it "cell weight = 1" and "mesenchymal weight = 0", clearly identifying it as a cellular component; if a pixel has a gray value of 160, which falls within the mesenchymal component range (151-220), it assigns "mesenchymal weight = 1" and "cell weight = 0", clearly identifying it as a mesenchymal component.

[0039] By using grayscale value range matching logic (i.e., mathematical comparison of pixel values ​​with double thresholds), the model automatically calculates the classification of each pixel: for the critical grayscale value of 150 (upper limit of cell range), it is determined as "cell weight = 1"; for the grayscale value of 151 (lower limit of interstitial range), it is determined as "interstitial weight = 1", ensuring that there is no ambiguity in the range boundaries.

[0040] Ultimately, this logic was solidified into an automated recognition mechanism, which uses algorithm code (such as cell_mask[(b_channel>=B_LOW_CELL)&(b_channel<=B_HIGH_CELL)]=255) to automate the entire process, replacing the traditional manual pixel-by-pixel judgment method and avoiding human subjective error.

[0041] Using component type as the dividing dimension, a two-dimensional component separation matrix is ​​constructed by integrating cellular component mask data, interstitial component mask data, and B-channel grayscale value distribution data to achieve precise definition of cellular and interstitial components. After aligning the B-channel grayscale value data with dual threshold interval standards (cellular 2-150, interstitial 151-220), a two-dimensional component separation matrix is ​​constructed with "cellular components" and "interstitial components" as the core vertical axis and the horizontal axis integrating the three types of key data. The specific process is as follows: The vertical axis is defined as follows: it clearly divides the image into two categories, "cellular components" and "interstitial components," corresponding to the two types of components defined by the grayscale threshold layering module, ensuring that the matrix focuses on the accurate differentiation of the target components. Cellular component mask pixel distribution: Extract all pixel coordinates marked with 255 in the cellular component mask (e.g., (x1, y1), (x3, y3), etc.), and record their spatial distribution in the image, reflecting the actual area coverage of the cellular components. Interstitial component mask pixel distribution: Similarly, extract pixel coordinates marked with 255 in the interstitial component mask (e.g., (x2, y2), (x4, y4), etc.), and record the spatial distribution characteristics of the interstitial components. B-channel grayscale value frequency distribution: Statistically calculate the proportion of B-channel grayscale values ​​in the cellular (2-150) and interstitial (151-220) intervals of the entire image. For example, a thyroid biopsy smear might show that the 2-150 interval accounts for 30% of the pixels, and the 151-220 interval accounts for 25%, providing a quantitative basis for component proportions.

[0042] For critical pixels with gray values ​​of 150-151 (such as the mixed region in a pleural effusion smear), the matrix distinguishes them by cross-comparing three types of data: if the region has some pixel markers in the cell mask, the proportion of gray value 150 in the B channel is higher, and the spatial location is close to a known cell cluster, it is determined to be dominated by cellular components; otherwise, if the proportion of gray value 151 is higher and it is close to the interstitial region, it is classified as an interstitial component.

[0043] Finally, through spatial distribution and gray-level frequency correlation analysis of the matrix, a quantitative definition was made of the mixed region (the previously indistinguishable area of ​​interwoven cells and stroma) in a peritoneal lavage fluid smear: combining data such as the 70% pixel proportion of cell mask and the dense distribution of B-channel gray values ​​of 120-150 pixels, it was determined that 70% of the region consisted of cellular components (such as epithelial cells) and 30% consisted of stroma components (such as matrix), achieving precise quantification of component proportions. This two-dimensional component separation matrix enabled multi-dimensional fusion analysis from spatial distribution to gray-level features, providing a visualized and quantifiable basis for component separation.

[0044] Image data quality is optimized through morphological preprocessing mechanisms. Combined with grayscale threshold-based mask generation logic, component boundary features are enhanced and integrated with the component separation baseline model features to generate preliminary separation results of cells and stromal components. When performing morphological preprocessing on digital sections of pleural effusion smears, edge noise smoothing is performed according to the technical specifications of the region growth optimization module: preset morphological preprocessing logic (such as mild Gaussian blur) weakens interference signals such as residual red blood cells and carrier plate scratches in the sections, improving overall image quality. This step lays the foundation for subsequent component boundary enhancement.

[0045] After preprocessing, the core logic of mask generation, based on grayscale thresholding, is applied—only pixels with B-channel grayscale values ​​in the range of 2-150 (cells) and 151-220 (interstitial tissue) are retained, while invalid background pixels with grayscale values ​​<2 or >220 (such as blank carrier slice areas) are automatically filtered out. Through this filtering, the edge features of cell clusters and interstitial regions are significantly enhanced: for example, the originally blurred edges of epithelial cell clusters in pleural effusion smears become clearer due to the removal of invalid background pixels, and the boundary with the surrounding interstitial tissue becomes more defined.

[0046] Subsequently, the optimized edge feature data is fused with the component separation baseline model (weight judgment logic): the model performs secondary verification on each pixel through pixel assignment weights (cell weight 1 / interstitial weight 1). If a pixel simultaneously satisfies both "cell mask marking" and "grayscale value 2-150", it is determined to be a cellular component and marked in green; if it satisfies both "interstitial mask marking" and "grayscale value 151-220", it is marked as an interstitial component in red. For the critical overlapping region of grayscale value 150-151 (such as the transition zone where cells and interstitial tissue intertwine), due to the ambiguity of weight judgment, it is temporarily marked in yellow.

[0047] The final preliminary separation results (green / red / yellow three-color labeled images) completely preserved the diagnostically relevant cells and interstitial regions, and the component distribution was intuitively presented by color differentiation, providing clear basic data for subsequent steps (such as 5×5 convolution kernel erosion and dilation denoising, and four-neighbor region growth to extract connected components).

[0048] S103 uses morphological algorithms to optimize the generated cell component mask and interstitial component mask, removes noise through erosion and dilation operations of 5×5 convolution kernels, and then uses a four-neighbor expansion region growing algorithm to extract connected components, filters out invalid components smaller than 10 pixels, and generates optimized effective cell and interstitial components.

[0049] In one implementation, noise filtering is performed on the generated cellular component mask and interstitial component mask using morphological algorithms. Discrete noise is removed using erosion and dilation operations with a 5×5 convolution kernel. A four-neighborhood expansion region growing algorithm is introduced to extract connected components, achieving preliminary purification of the component masks. For the cellular component mask and interstitial component mask generated after preliminary separation of pleural effusion smears (at this stage, the mask contains discrete noise caused by uneven staining and background residue, such as isolated 1-2 pixel bright spots), morphological operations are performed using a 5×5 convolution kernel, as detailed below: Erosion operation: The mask is traversed by a 5×5 convolution kernel to eliminate small bumps at the edges (such as sporadic noise caused by residual red blood cells around cell clusters, or isolated pixels caused by carrier scratches), making the mask edges smoother. For example, noise with a diameter of less than 3 pixels in the cell component mask can be completely removed.

[0050] Dilation operation: After erosion, a reverse 5×5 convolution kernel operation is performed on the mask to repair the shrinkage of component edges caused by erosion (such as pixels missing from the edges of cell clusters due to erosion), restore the complete outline of cells and interstitial components, and ensure the morphological integrity of core components such as epithelial cell clusters and fibrous tissue.

[0051] After morphological denoising, a four-neighborhood expansion region growing algorithm is introduced: starting from the non-noise pixels (grayscale value 255) in the mask after erosion and dilation processing, adjacent pixels are scanned layer by layer in four directions (up, down, left, and right), merging continuous pixels of the same type (cells or stroma) into connected components. For example, scattered inflammatory cell pixels (originally isolated) in a pleural effusion smear are aggregated into complete cell cluster components, and fibrin filament pixels are connected into continuous stroma components.

[0052] Through the above process, more than 95% of discrete noise is removed, achieving preliminary purification of the composition mask and providing high-quality basic data for subsequent size filtering and precise optimization.

[0053] By integrating connected component feature data with minimum size filtering rules, a component validity evaluation model is constructed. Component retention weights are calculated using component area thresholds, establishing a dynamic removal mechanism for invalid components. The connected component feature data (including the number of pixels and spatial coordinate distribution of each component) extracted using a four-neighborhood region growing algorithm is integrated with minimum size filtering rules (≥10 pixels) to construct the component validity evaluation model.

[0054] The core logic of the model is based on "diagnostic relevance": In clinical pathological diagnosis, the morphological size of effective cells or stromal components (such as cancer cell clusters and connective tissue) is usually greater than 10 pixels, while components smaller than 10 pixels are mostly background noise (such as carrier plate impurities and residual staining particles). Therefore, the model sets retention weight rules through a pixel number threshold: if the number of pixels in a component is ≥10, it is determined to be an effective component with diagnostic significance, and the retention weight is assigned a value of 1; if the number of pixels is <10, it is determined to be an invalid component, and the retention weight is assigned a value of 0.

[0055] For example, in the cellular component mask of a thyroid biopsy smear, a connected component containing 8 pixels (less than 10 pixels) is identified by the model as background noise, has a retention weight of 0, and is marked as an object to be removed; another epithelial cell cluster component containing 50 pixels (≥10 pixels) has a retention weight of 1 and is marked as an object to be retained. Using this threshold judgment logic, the system automatically traverses all connected components and batches out components with fewer than 10 pixels, establishing a dynamic removal mechanism for invalid components.

[0056] Ultimately, after model screening, the invalid noise components in the component masks of various samples such as pleural effusion smears and peritoneal lavage fluid smears reached a removal rate of over 98%, ensuring that only effective components of cells and interstitium that meet the diagnostic criteria are retained, laying the foundation for subsequent accurate analysis.

[0057] Using component type as the optimization dimension, a two-dimensional component optimization matrix is ​​constructed by integrating cellular component mask optimization data, interstitial component mask optimization data, and connected component size data to accurately screen effective components. The integrity of connected components is enhanced through a region growth expansion mechanism. This is combined with the edge optimization effects of morphological erosion and dilation operations, and integrated with the features of the component effectiveness assessment model to generate optimized cellular and interstitial effective components. After screening effective connected components using a minimum size filtering rule (≥10 pixels), the components are further optimized through a region growth expansion mechanism. Taking a peritoneal lavage fluid smear as an example: for the screened cell cluster components, the system starts from the component edge pixels and expands the scan layer by layer in the four neighboring directions (up, down, left, and right), filling in the tiny voids (such as gaps between cells with a diameter <3 pixels) formed by staining gaps or slight noise within the component, ensuring the continuity and integrity of the cell cluster morphology and preventing the core components from being segmented by voids.

[0058] At the same time, combined with morphological optimization logic, a secondary erosion and dilation operation of 5×5 convolution kernel is performed on the component: first, the component edge is refined by erosion (such as eliminating the jagged edge caused by uneven staining of interstitial components), and then the edge excessive contraction is repaired by dilation, so that the boundary between cells and interstitium is smoother and clearer (such as the contour error of the connective tissue edge in thyroid puncture smear is reduced to within 2 pixels after processing).

[0059] After optimization, the component features and the component effectiveness evaluation model (with a retention weight ≥1) are fused and verified: all components after expansion and edge optimization must again meet the standards of "pixel count ≥10" and "continuous boundaries". The final generated cell and interstitial effective components achieve triple optimization - no redundant noise (removing components <10 pixels), complete morphology (filling holes), and clear boundaries (refining edges), providing high-precision effective component data support for subsequent lightweight reconstruction (data compression, block storage).

[0060] S104 performs lightweight reconstruction processing on the optimized effective components of cells and stromal tissue, removes redundant background from the original slices, and stores the data using a 1000×1000 pixel block strategy, compressing the data volume to less than 1 / 10 of the original, generating independent RGB image views of cellular and stromal components, and realizing independent visualization or overlay analysis of the two types of components through a component label association mechanism.

[0061] In one implementation, a lightweight reconstruction process is used to simplify the optimized cell and stromal components, removing redundant background from the original slides. A 1000×1000 pixel block strategy is employed for storage, compressing the data volume to less than 1 / 10 of the original, achieving efficient data simplification. When performing lightweight reconstruction on the optimized peritoneal lavage fluid smear cells and stromal components (which have undergone morphological denoising and size filtering to remove discrete noise and invalid components <10 pixels), a preset algorithm automatically identifies redundant background areas in the original slides, including blank carrier slice areas (pixel areas without any cells or stromal components) and residual red blood cell areas (interfering pixels with grayscale values ​​outside the effective range of 2-220). Data from these areas is directly removed, retaining only the optimized cell clusters (such as epithelial cells and inflammatory cells) and stromal tissues (such as matrix and fibrous tissue), which are diagnostically relevant and effective components.

[0062] In the data storage stage, the continuous effective component region is divided into multiple 1000×1000 pixel sub-blocks according to the spatial coordinate order of the original slices (for example, the effective component distribution range of a certain peritoneal lavage fluid smear is relatively wide, and it is evenly divided into 20 independent sub-blocks). Each sub-block contains only the cell or interstitial effective component data of the corresponding region, avoiding redundant background occupying storage space.

[0063] Taking a peritoneal lavage fluid smear with an original data size of 5.4GB as an example, after the above process, invalid background data was completely removed, and only the effective component data was retained and integrated through block storage. The final file size was reduced to 391MB, with a compression rate of 93%. This not only met the core goal of compressing the data size from GB to MB, but also ensured that the target area could be quickly located when reading the data through the block strategy. It took into account both storage efficiency and access convenience, laying the foundation for subsequent remote transmission and rapid image reading.

[0064] By integrating the segmented storage of effective component data with independent component visualization rules, a component view generation model is constructed. The model defines the view output standard through RGB format conversion logic, establishing an independent display mechanism for cellular and interstitial components. The core function of this model is to achieve data visualization through RGB format conversion logic: converting a single-channel component mask (containing only grayscale values ​​of 0 and 255) into a three-channel color image, where the effective component area is displayed according to a preset hue, and the background area is set to black (grayscale value 0).

[0065] The cellular component view uses green as the primary color (RGB value set to (0, 255, 0)), while the stromal component view uses red as the primary color (RGB value set to (255, 0, 0)). This rule conforms to the technical requirements for component-independent RGB image output. For example, when converting cellular blocks in a thyroid biopsy smear, pixels marked as cellular components in the mask (original grayscale value 255) are assigned green, and effective components such as epithelial cell clusters and follicular cells are clearly presented in green, while background areas without components remain black. Similarly, pixels marked as stromal components in the stromal block (original grayscale value 255) are assigned red, and components such as connective tissue and matrix are displayed in red to avoid confusion with cellular components.

[0066] The independent view generated by this model establishes a dedicated display mechanism for cellular and interstitial components: doctors can view the cellular view (green component) or the interstitial view (red component) separately, replacing the traditional view mode where cells and interstitial components are mixed and superimposed in slides, thus solving the problem of identification difficulties caused by component interweaving during traditional slide reading.

[0067] Using component type as the output dimension, a two-dimensional visualization matrix is ​​constructed by integrating cellular effective component data, interstitial effective component data, and segmented storage information to achieve independent view generation and associated management of the two types of components. In the process of constructing the two-dimensional visualization matrix with "cellular component view" and "interstitial component view" as the output dimensions, the vertical axis dimension is first clearly defined as the two core view types to ensure that the matrix focuses on the associated management of cellular and interstitial components.

[0068] The horizontal axis integrates three types of key data, as follows: Cellular effective component block storage path: Records the file storage location of each cell block, such as "cell_block01.png" and "cell_block02.png", which correspond to 1000×1000 pixel cell component sub-blocks after lightweight processing; Interstitial effective component block storage path: Similarly, records the storage location of interstitial blocks, such as "interstitial_block01.png" and "interstitial_block02.png", each corresponding to the same original slice area as the cell blocks; Block spatial coordinate information: Marks the specific location range of each block in the original slice, for example, "block01" corresponds to the upper left corner of the original slice's X coordinate (1-1000) and Y coordinate (1-1000), and "block02" corresponds to the middle left area of ​​the X (1-1000) and Y (1001-2000).

[0069] Through this two-dimensional matrix, the system can establish a mapping relationship between segmented paths and spatial locations: when a doctor queries "segment 01", the matrix can quickly locate the corresponding "cell_segment 01.png" and "mesenchyma_segment 01.png", and clearly identify that both originate from the upper left corner of the original slide. This mechanism enables the precise generation and associated management of independent views of cells and mesenchyma. Doctors can quickly access two types of views at the same spatial location through the matrix for comparative analysis (such as observing the relationship between cell distribution and mesenchymal support in a certain area), improving the efficiency of spatial positioning and the convenience of analysis when reading slides.

[0070] The correspondence between cells and interstitial components is strengthened through a component tagging association mechanism. The compression characteristics of lightweight data ensure the integrity of diagnostic information, and the data is integrated with the features of the component view generation model to achieve independent visualization or overlay analysis of the two types of components. When adding component tags to the cellular and interstitial component views of pleural effusion smears, specific types are labeled for the identified effective components in the cellular view: for example, densely distributed green areas are labeled as "epithelial cell clusters," and scattered green dots are labeled as "inflammatory cells." Red areas in the interstitial view are labeled with types such as "fibrous tissue" and "matrix," ensuring that the tags match the morphological characteristics of the components.

[0071] Through a tag association mechanism, cell tags and mesenchymal tags within the same segment are bound together (e.g., the "epithelial cell cluster" tag in "segment 03" is mapped to the "fibrous tissue" tag in the same area), achieving a dual association between spatial location and component type. Combined with the core characteristic of lightweight data—retaining 100% diagnostic information after compression (such as the preservation of key features like cell morphology and mesenchymal distribution)—doctors can flexibly switch view modes using the system's functions. When viewing the cell view alone, the green labels "epithelial cell clusters" and "inflammatory cells" are clearly displayed, making it easy to focus on cell morphology analysis; when viewing the mesenchymal view alone, the red label "fibrous tissue" is highlighted, making it easy to assess the mesenchymal structure; when selecting the overlay view, the system overlays the green cell area with the red mesenchymal area, and the overlapping transition area is displayed in yellow, intuitively showing the spatial distribution relationship between cells and mesenchymal tissue (such as whether there is fibrous tissue proliferation around the epithelial cell clusters).

[0072] By using label association and flexible view switching, the time cost for doctors to identify components in mixed views is reduced, and the diagnostic consistency (Kappa value) is improved from 0.72 in the traditional method to 0.89, significantly improving the efficiency of image reading and diagnostic accuracy.

[0073] Furthermore, this application utilizes lightweight, MB-level files (e.g., peritoneal lavage fluid smears compressed from 5.4GB to 391MB, and thyroid biopsy smears compressed from 8.1GB to 562MB), whose data volume is less than 1 / 10 of the original slides, making them perfectly compatible with the storage and transmission capabilities of ordinary PCs. Without relying on high-performance servers or GPU support, a single lightweight file can be directly stored on a regular hard drive, and transmission over a 10Mbps network takes only a few seconds (e.g., a 10MB file takes 0.8 seconds to transfer). This solves the pain point of traditional GB-level slides experiencing transmission delays (over 30 minutes) in remote consultations at primary hospitals, providing efficient data support for cross-institutional real-time consultations.

[0074] like Figure 2 As shown, the lightweight file contains independent RGB views of cellular and stromal components (cellular components are marked in green, and stromal components are marked in red). Doctors can directly focus on clearly marked diagnostic areas in the view (such as "epithelial cell clusters" and "inflammatory cells" in pleural effusion smears or "follicular epithelial cells" in thyroid puncture smears) without having to search through each field of view in the redundant background of the original slides (blank carrier slides, residual red blood cells). Combined with the component label association mechanism (such as the correspondence between cells and stromal in the same area), doctors can quickly locate key lesion features, reducing the average slide reading time from 28.5 minutes in the traditional method to 4.3 minutes, an improvement rate of 85.1%. Moreover, the standardized view output reduces subjective judgment differences, and the diagnostic consistency (Kappa value) increases from 0.72 to 0.89, significantly reducing the rate of missed diagnoses and the risk of misdiagnosis.

[0075] This application aims to achieve automated separation and lightweight processing of cellular and stromal components through B-channel grayscale differences. First, digital slices of pleural effusion, peritoneal lavage fluid, and thyroid biopsy smears are acquired. After preprocessing, based on the B-channel grayscale characteristics of HE or Papanicolaou staining, dual threshold ranges are set for cells (2-150) and stromal (151-220). A component mask is generated by extracting B-channel grayscale values, and denoising is achieved using morphological operations with 5×5 convolution kernels. Then, a four-neighborhood region growing algorithm is used to extract connected components of ≥10 pixels to optimize effective components. Finally, the data is stored in 1000×1000 pixel blocks, redundant background is removed, and the data is compressed to less than 1 / 10 of its original size, generating independent RGB views that support overlay analysis.

[0076] This technology requires no GPU support and can run on a regular PC, reducing storage costs by more than 90%, reducing remote transmission time from 30 minutes to a few seconds, improving image viewing efficiency by 85.1%, and increasing diagnostic consistency (Kappa value) from 0.72 to 0.89, effectively solving the problems of digital slice storage, transmission, and image viewing efficiency.

[0077] In one implementation, such as Figure 3 As shown, this application also provides a device for separating formed elements and lightweighting cytopathological smears by digital sectioning based on grayscale threshold layering, comprising: The acquisition module 301 is used to acquire digital slide image data of cytopathological smears. Through image enhancement and standardization preprocessing, it generates preset gray range standards for cells and stromal components based on gray threshold layering information. The gray value range of cell components is 2-150, and the gray value range of stromal components is 151-220. The digital slide image data of cytopathological smears includes pleural effusion smears, peritoneal lavage fluid smears, and thyroid puncture smears. The processing module 302 is used to process the preset grayscale range standards of cell and stromal components and digital slide image data of cytopathological smears. By extracting the grayscale values ​​of the B channel of the image, cell component masks and stromal component masks are generated to achieve preliminary separation of cell and stromal components. The generated cell and stromal component masks are optimized using morphological algorithms. Noise is removed by erosion and dilation operations of 5×5 convolution kernels. Then, a four-neighbor expansion region growing algorithm is used to extract connected components and filter out invalid components smaller than 10 pixels to generate optimized effective cell and stromal components. The optimized effective cell and stromal components are then subjected to lightweight reconstruction processing to remove redundant background in the original slides. A 1000×1000 pixel block strategy is used for storage to compress the data volume to less than 1 / 10 of the original, generating independent RGB image views of cell and stromal components. Independent visualization or overlay analysis of the two types of components is achieved through a component label association mechanism.

[0078] The computer-readable storage medium provided in the above embodiments of this application and the method for separation and lightweight processing of formed elements in digital slides of cytopathological smears based on grayscale threshold layering provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0079] All embodiments in this application are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for evaluating a method for separating and lightweighting formed elements in digital slides of cytopathological smears based on grayscale threshold layering, an electronic device, an electronic device, and a readable storage medium are basically similar to the embodiments of the aforementioned method for separating and lightweighting formed elements in digital slides of cytopathological smears based on grayscale threshold layering. Therefore, the description is relatively simple, and relevant parts can be referred to in the description of the embodiments of the aforementioned method for separating and lightweighting formed elements in digital slides of cytopathological smears based on grayscale threshold layering.

Claims

1. A method for cytological smear digital slide object formation component separation and light-weight processing based on gray scale thresholding layering, characterized by, The method comprises the following steps: Obtain cell pathology smear digital slice image data, preprocess by image enhancement and standardization, and generate preset cell and interstitial component gray interval standards based on gray threshold layering information, wherein the cell component gray value interval is 2-150, and the interstitial component gray value interval is 151-220, and the cell pathology smear digital slice image data includes pleural effusion smear, peritoneal lavage fluid smear, and thyroid puncture smear; Process the preset cell and interstitial component gray interval standards and the cell pathology smear digital slice image data, extract the image B channel gray value, generate cell component masks and interstitial component masks, and realize preliminary separation of cell and interstitial components; Optimize the generated cell component masks and interstitial component masks by using a morphological algorithm, remove noise points by erosion and expansion operations of a 5*5 convolution kernel, and then extract connected components by using a four-neighborhood expansion region growing algorithm, filter out invalid components smaller than 10 pixels, and generate optimized cell and interstitial effective components; Perform lightweight reconstruction processing on the optimized cell and interstitial effective components, remove redundant backgrounds in the original slice, store by using a 1000*1000 pixel block strategy, compress the data volume to less than 1 / 10 of the original, generate independent RGB image views of cell components and interstitial components, and realize independent visualization or superposition analysis of the two types of components by using a component label association mechanism.

2. The method of claim 1, wherein, Obtain cell pathology smear digital slice image data, preprocess by image enhancement and standardization, and generate preset cell and interstitial component gray interval standards based on gray threshold layering information, which comprises the following steps: Analyze the characteristics of the cell pathology smear digital slice image data by using the gray threshold layering principle, determine the gray difference characteristics of the cell and interstitial components, and generate preset cell component gray value intervals and interstitial component gray value intervals; Construct a component classification benchmark by using the B channel gray value data of the HE staining or Papanicolaou staining pathological image and the double-threshold interval standard, determine the preliminary division rule of the cell and interstitial components by using the gray value interval matching logic, and establish a gray threshold layering component recognition mechanism; Cover the pathological smear types as dimensions, fuse the image data of the cell pathology types including pleural effusion smear, peritoneal lavage fluid smear, and thyroid puncture smear, combine the gray threshold interval standard, and realize the definition of the cell pathology digital slice processing object; Load high-resolution digital slice data by using the image enhancement and standardization preprocessing mechanism, combine the gray threshold layering to strengthen the pertinence of B channel gray value extraction, and generate the initial judgment basis for the separation of cell and interstitial components by fusing the preset component gray interval standard.

3. The method of claim 2, wherein, Process the preset cell and interstitial component gray interval standards and the cell pathology smear digital slice image data, extract the image B channel gray value, generate cell component masks and interstitial component masks, and realize preliminary separation of cell and interstitial components, which comprises the following steps: The preset cell and interstitial component gray scale interval standard and cell pathology smear digital slice image data are matched by a gray scale threshold layering processing procedure, the B channel gray scale value of the HE staining or Pap staining pathological image is extracted, the cell component mask and the interstitial component mask are generated, and the preliminary division of the components is realized; The double threshold interval standard and the image B channel gray scale value data are connected, a component separation reference model is constructed, the pixel attribution weight is calculated through a gray scale value interval matching logic, and an automatic identification mechanism of the cell and interstitial components is established; A two-dimensional component separation matrix is constructed by taking the component type as the division dimension, fusing the cell component mask data, the interstitial component mask data and the B channel gray scale value distribution data, and the accurate definition of the cell and interstitial components is realized; The image data quality is optimized through a morphological preprocessing mechanism, the component boundary features are strengthened by combining the gray scale threshold layering mask generation logic, and the preliminary separation result of the cell and interstitial components is generated by fusing the features of the component separation reference model.

4. The method of claim 1, wherein, The generated cell component mask and interstitial component mask are optimized by a morphological algorithm, the noise points are removed through the erosion and expansion operation of a 5*5 convolution kernel, the connected components are extracted by using a four-neighborhood expansion region growing algorithm, the invalid components smaller than 10 pixels are filtered out, and the optimized cell and interstitial effective components are generated, including: The generated cell component mask and interstitial component mask are filtered by a morphological algorithm, the discrete noise points are removed through the erosion and expansion operation of a 5*5 convolution kernel, the connected components are extracted by introducing a four-neighborhood expansion region growing algorithm, and the preliminary purification of the component mask is realized; The connected component feature data and the minimum size filtering rule are connected, a component effectiveness evaluation model is constructed, the component retention weight is calculated through a component area threshold judgment logic, and a dynamic invalid component elimination mechanism is established; A two-dimensional component optimization matrix is constructed by taking the component type as the optimization dimension, fusing the cell component mask optimization data, the interstitial component mask optimization data and the connected component size data, and the accurate screening of the effective components is realized; The connected component integrity is strengthened through a region growing expansion mechanism, the edge optimization effect of the morphological erosion and expansion operation is combined, the features of the component effectiveness evaluation model are fused, and the optimized cell and interstitial effective components are generated.

5. The method of claim 4, wherein, The optimized cell and interstitial effective components are subjected to lightweight reconstruction processing, the redundant background in the original slice is removed, a 1000*1000 pixel block strategy is adopted for storage, the data volume is compressed to less than 1 / 10 of the original, the independent RGB image views of the cell component and the interstitial component are generated, and the independent visualization or superposition analysis of the two types of components is realized through a component label association mechanism, including: The optimized cell and interstitial effective components are subjected to data simplification through a lightweight reconstruction processing procedure, the redundant background in the original slice is removed, a 1000*1000 pixel block strategy is adopted for storage, the data volume is compressed to less than 1 / 10 of the original, and the efficient simplification of the data is realized; The effective component data of the docking block storage is matched with the component independent visualization rule, a component view generation model is constructed, a view output standard is defined through RGB format conversion logic, and an independent display mechanism of cell and interstitial components is established; Taking the component type as the output dimension, the cell effective component data, the interstitial effective component data and the block storage information are fused to construct a two-dimensional visualization matrix, and independent view generation and association management of the two types of components are realized; The corresponding relationship between the cell and interstitial components is strengthened through the component label association mechanism, the compression characteristics of lightweight data are combined to guarantee the integrity of the diagnostic information, and the independent visualization or superimposed analysis of the two types of components is realized by combining with the characteristics of the component view generation model.

6. A device for cytological smear digital slide component separation and lightweight processing based on gray scale thresholding, characterized by, The device comprises: An acquisition module is configured to acquire cell pathology smear digital slice image data, perform image enhancement and standardization preprocessing, and generate a preset cell and interstitial component gray interval standard based on gray threshold layering information, wherein the cell component gray value interval is 2-150, the interstitial component gray value interval is 151-220, and the cell pathology smear digital slice image data includes pleural effusion smear, peritoneal lavage fluid smear, and thyroid puncture smear; A processing module is configured to process the preset cell and interstitial component gray interval standard and the cell pathology smear digital slice image data, generate cell component masks and interstitial component masks by extracting image B channel gray values, realize preliminary separation of cell and interstitial components, perform optimization processing on the generated cell component masks and interstitial component masks by using a morphological algorithm, remove noise points through 5*5 convolution kernel erosion and inflation operations, and then extract connected components by using a four-neighborhood expansion region growing algorithm, filter out invalid components smaller than 10 pixels, generate optimized cell and interstitial effective components, and perform lightweight reconstruction processing on the optimized cell and interstitial effective components to remove redundant backgrounds in the original slice, store the data by using a 1000*1000 pixel block strategy, compress the data to less than 1 / 10 of the original data, generate independent RGB image views of cell components and interstitial components, and realize independent visualization or superimposed analysis of the two types of components through a component label association mechanism.

7. An electronic device, comprising: It comprises: A first processor; And a memory for storing executable instructions of the first processor; The first processor is configured to execute the executable instructions to perform the method for separating and lightweight processing of cell pathology smear digital slice components based on gray threshold layering according to any one of claims 1-5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the second processor to realize the method for separating and lightweight processing of cell pathology smear digital slice components based on gray threshold layering according to any one of claims 1-5.

Images