An online quality control review method and system for remote pathological diagnosis

Abstract

This invention provides an online quality control review method and system for remote pathological diagnosis, relating to the field of telemedicine technology. First, it receives a set of digital images of pathological slides uploaded by a remote pathological diagnosis terminal, containing microscopic imaging image units with depth-of-field markers under multiple fields of view. Next, it analyzes the pathological morphological features of these images to obtain the distribution characteristics of cell nuclei and the topological structure characteristics of cytoplasmic texture. Based on these features, it constructs a chain of cellular atypia evolution features and generates a set of quality control review markers. Based on the quality control review marker set, it determines the boundary contour chain of the lesion area and the chromatin feature offset vector path. Finally, it loads these markers into a preset template to generate a pathological quality control review report. This invention can efficiently and accurately perform quality control review of remote pathological diagnoses, improving diagnostic accuracy and consistency.

Description

Technical Field

[0001] This invention relates to the field of telemedicine technology, and more specifically, to an online quality control auditing method and system for remote pathology diagnosis. Background Technology

[0002] Traditional pathological diagnosis mainly relies on pathologists directly observing pathological slides under a microscope. While this method can provide relatively accurate diagnostic results, it has many limitations.

[0003] On the one hand, due to the uneven distribution of pathologist resources, patients in remote areas often struggle to access timely and high-quality pathological diagnostic services. On the other hand, the traditional pathological diagnostic process lacks effective quality control measures, and the accuracy and consistency of diagnostic results are easily affected by factors such as the pathologist's personal experience, skill level, and subjective judgment.

[0004] With the development of digital medical imaging technology, remote pathology diagnosis has gradually emerged. By digitizing pathology slides and uploading them to remote diagnostic terminals, remote pathology diagnosis has been achieved. However, existing online quality control and auditing methods for remote pathology diagnosis still have shortcomings. For example, some methods only perform simple quality assessments on the digital images of pathology slides, failing to deeply analyze the pathological morphological characteristics in the images and making it difficult to accurately determine key information such as lesion areas and cellular atypia, resulting in insufficient accuracy and effectiveness of quality control audits. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an online quality control audit method for remote pathological diagnosis, the method comprising: Receive a set of digital images of pathological slides uploaded by a remote pathological diagnosis terminal. The set of digital images of pathological slides includes microscopic imaging image units with depth-of-field markers that are continuously acquired under multiple fields of view. The pathological morphological features of the digital image set of the pathological slides are analyzed to obtain the morphological distribution features of cell nuclei and the topological features of cytoplasmic texture in the microscopic imaging image unit. The morphological distribution features of cell nuclei include nucleolar position offset parameters and chromatin edge continuity measures. The topological features of cytoplasmic texture include cytoplasmic fiber orientation consistency parameters and cytoplasmic vacuolar spatial distribution measures. Based on the morphological distribution characteristics of the cell nuclei and the topological structure characteristics of the cytoplasm texture, a cell atypia evolution feature chain based on the continuous measurement of chromatin edges is constructed, and a quality control review mark set for the set of digital images of the pathological slides is generated based on the cell atypia evolution feature chain. Based on the quality control audit mark set, determine the lesion region boundary contour chain and chromatin feature offset vector path of the pathological slide digital image set; The boundary contour chain of the lesion area and the chromatin feature offset vector path are loaded into a preset pathological diagnosis quality control report template to generate a pathological quality control review report instruction containing the mapping relationship between the cell atypia evolution feature chain and the boundary contour chain of the lesion area.

[0006] Furthermore, embodiments of the present invention also provide an online quality control and review system for remote pathological diagnosis, comprising: A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to perform the above-described online quality control auditing method for remote pathological diagnosis by executing the machine-executable instructions.

[0007] Based on the above, firstly, the system receives a set of digital images of pathological slides uploaded by a remote pathology diagnostic terminal. These images consist of continuously acquired microscopic imaging units with depth-of-field markers from multiple fields of view. The system then performs pathological morphological feature analysis on the set of digital images to obtain the distribution characteristics of cell nuclei and the topological structure characteristics of cytoplasmic texture. Based on these features, a cell atypia evolution feature chain based on continuous chromatin edge measurement is constructed, and a quality control audit mark set is generated. This allows for the analysis of cell atypia changes at the cellular level, accurately understanding the development trend of cellular lesions. Based on the quality control audit mark set, the system determines the boundary contour chain of the lesion area and the chromatin feature offset vector path, enabling precise location of lesion areas and chromatin changes. Finally, the boundary contour chain of the lesion area and the chromatin feature offset vector path are loaded into a preset pathology diagnostic quality control report template, generating a pathology quality control audit report instruction containing the mapping relationship between the cell atypia evolution feature chain and the lesion area boundary contour chain. This provides quality control audit information for pathologists, helping to improve the accuracy and consistency of remote pathology diagnosis and achieving efficient and accurate quality control auditing for remote pathology diagnosis. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the execution flow of the online quality control and auditing method for remote pathological diagnosis provided in an embodiment of the present invention.

[0009] Figure 2 This is a schematic diagram of exemplary hardware and software components of an online quality control and auditing system for remote pathology diagnosis provided in an embodiment of the present invention. Detailed Implementation

[0010] Figure 1 This is a flowchart illustrating an online quality control and auditing method for remote pathological diagnosis provided in one embodiment of the present invention, which will be described in detail below.

[0011] Step S110: Receive a set of digital images of pathological slides uploaded by a remote pathological diagnosis terminal. The set of digital images of pathological slides includes microscopic imaging image units with depth-of-field markers that are continuously acquired under multiple fields of view.

[0012] In a remote pathology diagnostic system applied to a regional medical collaboration network, the fully automated digital pathology scanning equipment in a primary hospital generates a data stream conforming to a medical digital imaging transmission protocol after scanning a patient's tissue slide. Step S110 receives a set of digital images of the pathology slide uploaded by the remote pathology diagnostic terminal. This set includes microscopic imaging image units that are continuously scanned along the surface of the tissue slide under multiple microscopic fields of view. Each microscopic imaging image unit carries a depth-of-field marker to identify its relative position in the vertical optical axis direction. For a tissue slide with a thickness of 5 micrometers, the scanning equipment acquires images at three different focal planes in its vertical direction, with each image layer having a corresponding depth-of-field marker. Simultaneously, in the horizontal direction, the scanning equipment moves the stage along a serpentine path, acquiring hundreds of overlapping fields of view, each with a unique field-of-view position index. All of these microscopic imaging image units carrying depth-of-field markers and field-of-view position indices together constitute the set of digital images of the pathology slide.

[0013] Step S111: Receive the original scan data stream of pathological slides uploaded by the remote pathology diagnosis terminal through the medical digital imaging transmission protocol. The original scan data stream of pathological slides includes panoramic scan data of slides with a unified spatial reference system, which are continuously acquired under different objective magnifications.

[0014] In the raw scan data stream of pathological slides uploaded via medical digital imaging transmission protocols, the header and payload of each data packet conform to the standard protocol specifications for digital imaging and communication in medicine. The raw scan data stream contains two sets of panoramic scan data of the same tissue slide obtained using 10x and 40x objectives, respectively. All of these data share the same unified spatial reference system; that is, all image data are recorded with the same origin of the stage coordinate system as a reference for positional information.

[0015] Step S112: Perform data packet parsing processing on the original scan data stream of the pathological slide, extract the depth level marker field and field of view position index field from the packet header information, and perform data segmentation processing on the original scan data stream of the pathological slide according to the depth level marker field and field of view position index field to obtain a set of original image data segments with depth level attributes.

[0016] Each data packet in the raw scan data stream of pathological slides is parsed, and the depth-of-field marker field and field of view index field are extracted from the header information of each data packet. The depth-of-field marker field records the focal plane level to which the current image data belongs, and the field of view index field records the two-dimensional grid coordinates of the field of view to which the current image data belongs on the tissue slice. Based on the values ​​of the depth-of-field marker field and the field of view index field, data packets with the same depth-of-field marker field value and the same field of view index field value are grouped into the same raw image data segment. All raw image data segments constitute a set of raw image data segments with depth-of-field attributes.

[0017] Step S113: For each original image data segment in the set of original image data segments, determine the spatial hierarchy order of the original image data segment in the direction perpendicular to the optical axis according to the value of its corresponding depth level marker field, and perform layer stacking and arrangement of the original image data segments according to the spatial hierarchy order to generate a multi-layer image slice data stack with depth level as the vertical axis dimension.

[0018] The depth-of-field marker field value corresponding to the original image data segment is read. The magnitude of this value determines the spatial hierarchy of the original image data segment in the direction perpendicular to the optical axis. The smaller the value, the closer the segment is to the coverslip; the larger the value, the closer it is to the slide. Based on the ascending spatial hierarchy, all original image data segments are stacked hierarchically. The original image data segment with the smallest spatial hierarchy is placed at the top layer of the multi-layer image slice data stack, and the original image data segment with the largest spatial hierarchy is placed at the bottom layer. This generates a multi-layer image slice data stack with depth of field as the vertical axis.

[0019] Step S114: Obtain the image pixel matrix of each layer of the multi-layer image slice data stack, and perform color space conversion processing on the image pixel matrix to convert the pixel values ​​under the original sensor color space into color channel component values ​​under the standard pathological color space that is independent of the display device.

[0020] The image pixel matrix corresponding to each layer of the multi-layer image slice data stack is obtained. The original numerical format of this image pixel matrix is ​​the pixel value in the original sensor color space in Bayer array format. Color space conversion processing is performed on this image pixel matrix, specifically including demosaicing and white balance correction. Demosaicing uses an interpolation algorithm to supplement the values ​​of the two missing color channels for each pixel, and white balance correction eliminates illumination color shift by adjusting the gain coefficients of the red, green, and blue channels. After these processes, the pixel values ​​in the original sensor color space are converted into color channel component values ​​in a standard pathological color space independent of the display device, such as the L-star value, a-star value, and b-star value components in the International Commission on Illumination (ICI) standard color space.

[0021] Step S115: Based on the converted color channel component values, perform pixel-level registration on the image slices between adjacent depth layers in the multi-layer image slice data stack. By calculating the color gradient difference value at the same pixel position between adjacent layer slices, generate a displacement correction field to characterize the degree of pixel offset between depth layers.

[0022] For two adjacent slices in a multi-layer image slice data stack, let the upper slice be layer P and the lower slice be layer P+1. Traverse the coordinates of each pixel in layer P and obtain the pixel at the same coordinate position in layer P+1. Calculate the gradient difference values ​​of the two pixels in the three channels (L-value, a-value, b-value), where the gradient difference value in each channel is the difference of that channel's value along the perpendicular optical axis. Square the gradient difference values ​​of the three channels, sum them, and take the square root to obtain the comprehensive color gradient difference value at that coordinate position. Using the comprehensive color gradient difference value as the optimization objective, iteratively solve for the two-dimensional displacement vector that minimizes this comprehensive color gradient difference value. This two-dimensional displacement vector represents the pixel offset degree between depths of field for that pixel. Perform the above calculation for all pixels; the two-dimensional displacement vectors of all pixels constitute a displacement correction field.

[0023] Step S116: Use the displacement correction field to perform spatial position remapping processing on the image pixel matrix of each layer of the multi-layer image slice data stack, eliminate the image misalignment and offset between layers caused by the micro-vibration of the microscope stage, and obtain the multi-layer image slice data stack after spatial alignment.

[0024] For each layer of the multi-layer image slice data stack, obtain the displacement correction field corresponding to that layer. For each pixel in the pixel matrix of that layer's image, read the corresponding two-dimensional displacement vector from the displacement correction field. Subtract this two-dimensional displacement vector from the original coordinates of the pixel to obtain the new coordinate position. Using bilinear interpolation, calculate the color channel component values ​​of the pixel at the new coordinate position based on the color channel component values ​​of the four integer pixels surrounding it. After traversing all pixels to complete the above remapping, obtain the spatially aligned image pixel matrix of that layer. Perform the same operation on all layer slices to obtain the spatially aligned multi-layer image slice data stack.

[0025] Step S117: Perform image fusion preprocessing on each layer of the multi-layer image slice data stack after spatial alignment. Based on the rate of change of color channel component values ​​of each pixel in adjacent depth layers, extract the pixels that meet the preset depth focus evaluation conditions as the focus clear pixels of the corresponding layer, and generate a focus clear pixel position mask.

[0026] For each pixel in the spatially aligned multi-layer image slice data stack, obtain the L-star value, a-star value, and b-star value of the corresponding pixel across all layer slices. Calculate the rate of change of the pixel's color channel component values ​​between adjacent layers, calculated as the absolute value of the channel component value of the next layer minus the absolute value of the channel component value of the previous layer. Sum the rates of change between all adjacent layers to obtain the total rate of change for that pixel. The preset depth-of-field focus evaluation condition is the layer slice with the largest total rate of change. For each pixel, compare its total rate of change across all layer slices, and mark the index of the layer slice with the largest total rate of change as the focus sharp layer for that pixel. After traversing all pixels, for each layer slice, collect the coordinates of all pixels marked with that layer index, and generate a binary mask with the same size as the image of that layer slice. The positions marked as focus sharp pixels are assigned a value of 1, and the remaining positions are assigned a value of 0. This binary mask is the focus sharp pixel position mask.

[0027] Step S118: Perform pixel-by-pixel weighted combination of the focus sharp pixel position mask of each layer and the color channel component values ​​of the slice of that layer. Based on the weight coefficient marked in the focus sharp pixel position mask, perform depth-of-field expansion synthesis processing on the multi-layer image slice data stack to obtain a microscopic imaging image unit with full field-of-field depth of field sharpness.

[0028] For each slice, the binary values ​​in the mask of the focused, sharpest pixel are used as weighting coefficients. For each pixel in the pixel matrix of that slice, the L-value, a-value, and b-value components are multiplied by their corresponding weighting coefficients. After performing this weighting operation on all slices, for the spatial coordinates of each pixel, the weighted L-value components of all slices are summed to obtain the fused L-value component. Similarly, the fused a-value and b-value components are obtained. The three channel component values ​​of all pixels after fusion constitute a new image pixel matrix, which is the microscopic imaging unit with full field-of-view depth of field sharpness.

[0029] Step S119: The field of view location index field is attached as a spatial location label to the corresponding microscopic imaging image unit, and the microscopic imaging image units with continuous field of view location index fields are combined into a pathological slide digital image set. Each microscopic imaging image unit in the pathological slide digital image set carries a corresponding depth level mark and field of view location index field.

[0030] The field of view location index field extracted in step S112 is used as a spatial location label and appended to the metadata of the microscopic imaging image unit generated in step S118. Following the continuity of row and column numbers in the field of view location index field, all microscopic imaging image units are combined. Units with the same row number and adjacent column numbers are stitched together horizontally, and units with adjacent row numbers and the same column numbers are stitched together vertically, forming a complete set of digital images of pathological slides. Each microscopic imaging image unit in this set of digital images of pathological slides carries a corresponding depth-of-field marker and field of view location index field.

[0031] Step S120: Perform pathological morphological feature analysis processing on the set of digital images of the pathological slides to obtain the morphological distribution features of cell nuclei and the topological features of cytoplasmic texture in the microscopic imaging image unit. The morphological distribution features of cell nuclei include nucleolar position offset parameters and chromatin edge continuity measures. The topological features of cytoplasmic texture include cytoplasmic fiber orientation consistency parameters and cytoplasmic vacuolar spatial distribution measures.

[0032] Step S121: Load the microscopic imaging image unit in the pathological slide digital image set into the pathological morphological feature analysis processor, perform color deconvolution separation processing on the microscopic imaging image unit, and according to the preset hematoxylin staining spectral absorption characteristic curve and eosin staining spectral absorption characteristic curve, demix the pixel color information of the microscopic imaging image unit into hematoxylin staining concentration distribution channel and eosin staining concentration distribution channel. In the hematoxylin staining concentration distribution channel, perform cell nucleus candidate region localization processing based on morphological top cap reconstruction, and perform grayscale opening operation on the hematoxylin staining concentration distribution channel by constructing a disk-shaped structural element with radius parameter adaptation to suppress background texture components with a size smaller than the radius parameter of the disk-shaped structural element, and obtain a cell nucleus foreground distribution map after background suppression.

[0033] For each pixel in a microscopic imaging unit, in the standard red-green-blue color space, color deconvolution separation processing treats this pixel color information as a linear combination of the absorption contributions of hematoxylin, eosin, and the background. By solving a system of linear equations based on Beer-Lambert's law, the red, green, and blue values ​​of each pixel are decomposed into the concentration contributions of hematoxylin and eosin. Traversing the entire microscopic imaging unit, the concentration contributions of all pixels constitute the hematoxylin and eosin concentration distribution channels. A disk-shaped structural element is constructed, with its radius parameter set according to the expected size range of typical cell nuclei in the microscopic imaging unit. This disk-shaped structural element is then used to perform grayscale opening operations on the hematoxylin concentration distribution channels, which involve erosion followed by dilation. The erosion operation eliminates bright details smaller than the disk-shaped structuring elements in the hematoxylin staining concentration distribution channel, while the dilation operation restores the size of the remaining main structures. Through grayscale opening, background texture components smaller than the radius parameter of the disk-shaped structuring elements are effectively suppressed, and the output is a background-suppressed nucleus-foreground distribution map.

[0034] Step S122: In the cell nucleus foreground distribution map, watershed segmentation is performed based on the concentration gradient magnitude and gradient direction of the pixels. Boundary separation is performed on the densely populated regions of cell nuclei that are stuck together. By tracing the local maximum path of the concentration gradient magnitude, segmentation boundary ridges are generated between cell nucleus regions, resulting in a set of cell nucleus segmentation regions enclosed by independent cell nucleus contour lines. For each cell nucleus segmentation region in the set of cell nucleus segmentation regions, the geometric morphology description parameters of the cell nucleus segmentation region are extracted. The geometric morphology description parameters include the length of the major axis of the equivalent ellipse of the cell nucleus contour, the length of the minor axis of the equivalent ellipse, and the eccentricity parameter of the equivalent ellipse.

[0035] The concentration gradient magnitude and direction of each pixel in the nucleus foreground distribution map after background suppression are calculated. The concentration gradient magnitude represents the drastic change in hematoxylin concentration within the pixel's neighborhood, and the gradient direction represents the direction of the fastest change. The concentration gradient magnitude of each pixel is treated as terrain height to construct a topographic map. Watershed segmentation simulates the process of water flowing from high to low terrain, starting from a local minimum point. When water flows from different local minima meet, segmentation boundary ridges are constructed at the meeting point. These ridges are generated by tracing the local maxima path of the concentration gradient magnitude. All segmentation boundary ridges separate densely packed, overlapping nucleus regions, resulting in a series of segmented nucleus regions enclosed by independent nucleus contours. All these segmented nucleus regions form a set of segmented nucleus regions. For each segmented nucleus region, the coordinates of all pixels enclosing the nucleus contours of that region are obtained. Calculate the equivalent ellipse with the same second moment as the segmented region of the cell nucleus. The major and minor axis lengths of this equivalent ellipse are the same as the major and minor axis lengths of the equivalent ellipse of the cell nucleus outline. Calculate the eccentricity parameter of the equivalent ellipse based on its major and minor axis lengths. This eccentricity parameter is equal to the square root of the difference between the major and minor axis lengths of the equivalent ellipse divided by the major axis length of the equivalent ellipse.

[0036] Step S123: Extract the concentration distribution values ​​of hematoxylin staining within each segmented region of the cell nucleus, calculate the mean value and the local variation range of the concentration distribution values ​​within the segmented region of the cell nucleus, and mark the local variation range of the concentration distribution values ​​as the chromatin edge continuity measure of the segmented region of the cell nucleus.

[0037] Calculate the arithmetic mean of the hematoxylin concentration distribution values ​​of all pixels within the segmented region of the cell nucleus to obtain the mean value of the concentration distribution. Calculate the local variation range of the hematoxylin concentration distribution values ​​within the segmented region of the cell nucleus, specifically by averaging the absolute values ​​of the differences between the hematoxylin concentration distribution value of each pixel within the segmented region and the hematoxylin concentration distribution values ​​of its neighboring pixels. This local variation range is labeled as the chromatin edge continuity measure of the segmented region of the cell nucleus.

[0038] Step S124: Locate the local maximum point of the hematoxylin staining concentration distribution value in each cell nucleus segmentation region, mark the local maximum point as the nucleolar center position point, calculate the Euclidean distance offset of the nucleolar center position point relative to the geometric center point of the cell nucleus segmentation region, and mark the Euclidean distance offset as the nucleolar position offset parameter.

[0039] Within the segmented nucleus region, the concentration distribution of hematoxylin staining at all pixels is scanned. Pixels with concentrations higher than all their neighboring pixels are identified and marked as local maxima, which are then designated as the nucleolar center. The coordinates of the geometric center of the segmented nucleus region are calculated. The x-coordinate of the geometric center is the average of the x-coordinates of all pixels within the segmented region, and the y-coordinate is the average of the y-coordinates of all pixels within the segmented region. The Euclidean distance offset between the nucleolar center and the geometric center of the segmented region is calculated and designated as the nucleolar position offset parameter for the segmented nucleus region.

[0040] Step S125: In the eosin staining concentration distribution channel, perform cytoplasmic texture primitive extraction processing based on the local binary mode operator. Use a preset circular neighborhood radius parameter and neighborhood sampling point number parameter to traverse and scan the eosin staining concentration distribution channel to generate a cytoplasmic texture primitive encoding map that reflects the local texture arrangement pattern of the pixels. Calculate the uniformity and consistency measure of the local texture direction based on the cytoplasmic texture primitive encoding map. By counting the number of bitwise jumps in the texture primitive encoding within the circular neighborhood, mark the region with a jump count lower than a preset jump count threshold as a cytoplasmic fiber orientation consistent region, and mark the consistency measure value of the cytoplasmic fiber orientation consistent region as the cytoplasmic fiber orientation consistency parameter.

[0041] Define a circular neighborhood radius parameter and a neighborhood sampling point number parameter. Using each pixel in the eosin staining concentration distribution channel as the center, determine a circular neighborhood according to the circular neighborhood radius parameter. Uniformly sample a number of neighborhood sampling points (the number of sampling points parameter) along the boundary of the circular neighborhood. The value of each sampling point is obtained through bilinear interpolation. Using the value of the center pixel as a threshold, binarize each sampling point within the circular neighborhood. A sampling point value greater than or equal to the center pixel value is recorded as 1, otherwise as 0. Arrange the resulting number of neighborhood sampling point parameters (the number of binary bits) into a binary number in a clockwise direction, then convert it to a decimal number. This decimal number is the local binary pattern encoding value of the center pixel. Traverse the entire eosin staining concentration distribution channel; the local binary pattern encoding values ​​of all pixels constitute a cytoplasmic texture primitive encoding map. For each pixel, count the number of bitwise transitions in the local binary pattern encoding values ​​of all sampling points within its circular neighborhood. The number of bitwise transitions refers to the number of changes from 0 to 1 or from 1 to 0 in the binary code. Regions with a number of bitwise jumps below a preset jump number threshold are marked as regions with consistent cytoplasmic fiber orientation, and the uniformity measurement value of such regions is marked as the cytoplasmic fiber orientation consistency parameter for those regions.

[0042] Step S126: Extract connected regions with gray values ​​lower than the preset cytoplasmic vacuoles gray value threshold from the eosin staining concentration distribution channel, calculate the area parameter and roundness parameter of the connected regions, and mark the connected regions that meet the area parameter and roundness parameter constraints as cytoplasmic vacuoles. Statistically count the number of cytoplasmic vacuoles per unit area and the degree of spatial dispersion as a measure of cytoplasmic vacuoles spatial distribution.

[0043] A preset grayscale threshold for cytoplasmic vacuoles is set, and pixels in the eosin staining concentration distribution channel with grayscale values ​​lower than this threshold are marked as candidate pixels. All connected regions composed of candidate pixels are extracted. For each connected region, its area parameter (the number of pixels within the region) is calculated. Simultaneously, the circularity parameter of the connected region is calculated, using the formula 4π multiplied by the area parameter divided by the square of the region's perimeter. The perimeter of the connected region is obtained by calculating the number of pixels at its boundary. Connected regions with an area parameter greater than a preset lower limit and less than a preset upper limit, and a circularity parameter greater than a preset circularity threshold, are marked as cytoplasmic vacuoles. After marking all cytoplasmic vacuoles, the number of cytoplasmic vacuoles per unit area and their spatial dispersion are statistically analyzed. This spatial dispersion is obtained by calculating the variance of the coordinates of the center points of all cytoplasmic vacuoles. These statistical results serve as a measure of the spatial distribution of cytoplasmic vacuoles.

[0044] Step S127: Combine the nucleolus position offset parameter of all cell nucleus segmentation regions within the same microscopic imaging unit with the chromatin edge continuity measure to form the cell nucleus morphology distribution feature, and combine the cytoplasmic fiber orientation consistency parameter with the cytoplasmic vacuoles spatial distribution measure within the same microscopic imaging unit to form the cytoplasmic texture topology feature.

[0045] Nucleolar position offset parameters and chromatin edge continuity measures for all nuclear segmentation regions within the microscopic imaging unit are collected. All nucleolar position offset parameters are arranged into a vector according to the index of their respective nuclear segmentation regions, and all chromatin edge continuity measures are also arranged into another vector according to the same index. These two vectors are combined to represent the nuclear morphological distribution characteristics of the microscopic imaging unit. Cytoplasmic fiber orientation consistency parameters and cytoplasmic vacuolar spatial distribution measures for the microscopic imaging unit are also collected. These two parameters are combined to represent the cytoplasmic texture topological structure characteristics of the microscopic imaging unit.

[0046] Step S130: Construct a cell atypia evolution feature chain based on chromatin edge continuity measurement according to the cell nuclear morphology distribution characteristics and the cytoplasmic texture topology characteristics, and generate a quality control audit mark set for the set of digital images of the pathological slides according to the cell atypia evolution feature chain.

[0047] Step S131: Extract the field of view position index field corresponding to the microscopic imaging image unit, sort all microscopic imaging image units in the pathological slide digital image set according to the spatial continuity relationship of the field of view position index field, obtain the spatial sequence of microscopic imaging image units along the preset tissue section extension direction, extract the chromatin edge continuity measure of the cell nucleus morphology distribution features of each microscopic imaging image unit in the spatial sequence of microscopic imaging image units, and use the chromatin edge continuity measure as the feature node value of the microscopic imaging image unit in the cell atypia evolution feature chain.

[0048] Extract the field of view location index field corresponding to each microscopic imaging image unit in the digital image set of pathological slides. The field of view location index field records the two-dimensional grid coordinates of the microscopic imaging image unit on the tissue section, including the row number and column number. Based on the row number and column number of the field of view location index field, determine the spatial adjacency relationship between microscopic imaging image units. Two microscopic imaging image units with the same row number and adjacent column numbers are considered to be left-right adjacent, and two microscopic imaging image units with adjacent row numbers and the same column number are considered to be top-bottom adjacent. Sort all microscopic imaging image units according to a preset tissue section extension direction, such as a serpentine path from left to right and from top to bottom, to obtain a one-dimensional spatial sequence of microscopic imaging image units. For each microscopic imaging image unit in the spatial sequence of microscopic imaging image units, extract the chromatin edge continuity measure from the cell nuclear morphology distribution characteristics of the microscopic imaging image unit obtained in step S127. Since the morphological distribution characteristics of cell nuclei include the continuity measure of chromatin edges in all segmented regions of cell nuclei within the microscopic imaging unit, the arithmetic mean of the continuity measures of chromatin edges in all segmented regions of cell nuclei within the microscopic imaging unit is taken as the representative value. This representative value is then used as the feature node value of the microscopic imaging unit in the cellular atypia evolution feature chain.

[0049] Step S132: Calculate the gradient parameter of the change in chromatin edge continuity measure between adjacent microscopic imaging image units in the spatial sequence of microscopic imaging image units. The gradient parameter is the ratio of the absolute value of the difference between the chromatin edge continuity measure of the subsequent microscopic imaging image unit and the chromatin edge continuity measure of the preceding microscopic imaging image unit to the spatial distance parameter between adjacent microscopic imaging image units.

[0050] For each pair of adjacent microscopic imaging units in the spatial sequence, the feature node values ​​of the preceding and subsequent microscopic imaging units are obtained. Simultaneously, the spatial distance parameter between these two units is acquired. This spatial distance parameter is calculated based on the field-of-view index fields of both units. The spatial distance between two adjacent units (left and right) is equal to the physical step size of the field of view in the horizontal direction, and the spatial distance between two adjacent units (upper and lower) is equal to the physical step size of the field of view in the vertical direction. The absolute value of the difference between the chromatin edge continuity metric of the subsequent and preceding microscopic imaging units is calculated. This absolute value is divided by the spatial distance parameter, and the resulting ratio is the gradient parameter between the pair of adjacent microscopic imaging units.

[0051] Step S133: Determine the evolution direction identifier between adjacent feature nodes in the cell atypia evolution feature chain based on the positive and negative polarity of the change gradient parameter. When the chromatin edge continuity measure of the subsequent microscopic imaging image unit is greater than that of the preceding microscopic imaging image unit, set the evolution direction identifier to an increasing identifier; otherwise, set the evolution direction identifier to a decreasing identifier.

[0052] The relationship between the chromatin edge continuity measure of a subsequent microscopic imaging unit and that of a preceding microscopic imaging unit is determined. When the subsequent value is greater than the preceding value, it indicates an increase in the roughness of the nuclear chromatin edges, suggesting an intensified trend of cellular atypia. In this case, the evolution direction identifier between the pair of adjacent feature nodes is set to an increasing indicator. When the subsequent value is less than or equal to the preceding value, it indicates a decrease or no change in the roughness of the nuclear chromatin edges, suggesting a weakening or stable trend of cellular atypia. In this case, the evolution direction identifier between the pair of adjacent feature nodes is set to a decreasing indicator.

[0053] Step S134: Obtain the cytoplasmic vacuoles spatial distribution measure in the cytoplasmic texture topology features of each microscopic imaging image unit in the spatial sequence of the microscopic imaging image units, calculate the distribution density change rate of the cytoplasmic vacuoles spatial distribution measure between adjacent microscopic imaging image units, and use the distribution density change rate as an auxiliary evolution constraint parameter for feature nodes in the cell heterogeneity evolution feature chain.

[0054] The spatial distribution metric of cytoplasmic vacuoles is extracted from the cytoplasmic texture topology features of each microscopic imaging unit obtained in step S127. The spatial distribution metric of cytoplasmic vacuoles includes the number of cytoplasmic vacuole regions per unit area and the degree of spatial dispersion. For each pair of adjacent microscopic imaging units in the spatial sequence, the spatial distribution metrics of cytoplasmic vacuoles for the preceding unit and the following unit are obtained. The absolute value of the difference between the number of cytoplasmic vacuole regions in the following unit and the number of cytoplasmic vacuole regions in the preceding unit is calculated and divided by the spatial distance parameter to obtain the rate of change in the number of cytoplasmic vacuoles. Similarly, the rate of change in the spatial dispersion of cytoplasmic vacuoles is calculated. The weighted sum of the two rates of change is taken as the distribution density change rate, which is the auxiliary evolution constraint parameter between the pair of adjacent feature nodes.

[0055] Step S135: Construct the node connection relationship of the cell heterogeneity evolution feature chain based on the evolution direction identifier and the auxiliary evolution constraint parameter. When the evolution direction identifier is an increasing identifier and the auxiliary evolution constraint parameter is greater than the preset evolution constraint threshold, establish a strengthening connection edge between adjacent feature nodes. When the evolution direction identifier is a decreasing identifier or the auxiliary evolution constraint parameter is less than or equal to the preset evolution constraint threshold, establish a weakening connection edge between adjacent feature nodes.

[0056] For each pair of adjacent feature nodes, read the evolution direction identifier obtained in step S133 and the auxiliary evolution constraint parameter obtained in step S134. A preset evolution constraint threshold is established. Judgment condition: When the evolution direction identifier is increasing and the auxiliary evolution constraint parameter is greater than the preset evolution constraint threshold, it indicates that cellular atypia is intensified while cytoplasmic vacuolation is also significantly altered; both synergistically indicate disease progression. In this case, a strengthening connection edge is established between the pair of adjacent feature nodes. When the evolution direction identifier is decreasing, or the auxiliary evolution constraint parameter is less than or equal to the preset evolution constraint threshold, it indicates that cellular atypia is not intensified or cytoplasmic vacuolation changes are not significant. In this case, a weakened connection edge is established between the pair of adjacent feature nodes.

[0057] Step S136: Based on the distribution range of chromatin edge continuity measurement values ​​of all feature nodes in the cell atypia evolution feature chain, determine the atypia evolution amplitude parameter of the cell atypia evolution feature chain. The atypia evolution amplitude parameter is the difference between the maximum and minimum chromatin edge continuity measurement in the cell atypia evolution feature chain.

[0058] Traverse all feature nodes in the cellular atypia evolution feature chain and obtain the continuous chromatin edge measurement value corresponding to each feature node. Find the maximum and minimum values ​​among these values. Calculate the atypia evolution amplitude parameter, which is the maximum value minus the minimum value. This atypia evolution amplitude parameter reflects the overall change in the roughness of the nuclear chromatin edge from one end of the tissue section to the other; the larger the value, the more drastic the change in cellular atypia throughout the entire section.

[0059] Step S137: Based on the number of consecutive identical evolution direction identifiers between adjacent feature nodes in the cell heterogeneity evolution feature chain, determine the heterogeneity evolution trend consistency parameter of the cell heterogeneity evolution feature chain. The heterogeneity evolution trend consistency parameter is the maximum sequence length value of the evolution direction identifier continuously maintaining an increasing or decreasing identifier.

[0060] Traverse the evolution direction identifier sequence between all adjacent feature nodes in the cell atypia evolution feature chain. Calculate the maximum length of consecutive identical identifiers in this evolution direction identifier sequence. Initialize the current consecutive length value to 1 and the maximum consecutive length value to 1. Starting from the second evolution direction identifier, sequentially check if the current identifier is the same as the previous identifier. If they are the same, increment the current consecutive length value by 1 and update the maximum consecutive length value to the larger of the current and maximum consecutive length values. If they are different, reset the current consecutive length value to 1. After traversal, the maximum consecutive length value is the atypia evolution trend consistency parameter. This atypia evolution trend consistency parameter reflects the stability of the cell atypia change trend; a larger value indicates that cell atypia continues to intensify or weaken over a longer spatial range, and the trend is more consistent.

[0061] Step S138: Map the anomalous evolution amplitude parameter and the anomalous evolution trend consistency parameter to a preset quality control audit mark mapping table to generate a quality control audit mark set corresponding to the set of digital images of the pathological slides. The quality control audit mark set includes mark symbols used to indicate the slide quality level and mark symbols used to indicate the risk tendency of lesions.

[0062] A pre-defined quality control audit mark mapping table is established, which defines the correspondence between different value ranges of the atypia evolution amplitude parameter and the atypia evolution trend consistency parameter, and the slide quality grade mark symbol and the lesion risk tendency mark symbol. The atypia evolution amplitude parameter value calculated in step S136 and the atypia evolution trend consistency parameter value calculated in step S137 are input into this mapping table. Based on the combination of their respective value ranges, the corresponding slide quality grade mark symbol and the corresponding lesion risk tendency mark symbol are retrieved from the table. The marks obtained from these table lookups constitute the quality control audit mark set.

[0063] Step S140: Determine the lesion region boundary contour chain and chromatin feature offset vector path of the pathological slide digital image set based on the quality control audit mark set.

[0064] Step S141: Extract the marker symbols used to indicate the risk tendency of lesions from the quality control audit mark set, and determine the range of lesion risk areas that need to be processed by boundary tracking in the digital image set of pathological slides based on the preset risk tendency level value corresponding to the marker symbols.

[0065] Extract markers indicating lesion risk propensity from the quality control audit marker set generated in step S138. A risk propensity level mapping relationship is preset, with each lesion risk propensity marker corresponding to a risk propensity level value. The corresponding risk propensity level value is determined based on the marker, and this value determines the range of the lesion risk area requiring boundary tracking. A higher risk propensity level value indicates a larger area requiring boundary tracking, meaning a larger tracking range extending from the high-risk microscopic imaging unit to its neighborhood.

[0066] Step S142: In the spatial sequence of microscopic imaging image units in the digital image set of the pathological slide, locate the microscopic imaging image unit corresponding to the feature node in the cell atypia evolution feature chain where the continuous measurement of chromatin edge exceeds the preset atypia determination threshold, and mark the located microscopic imaging image unit as the core region unit of the lesion.

[0067] A predefined atypia threshold is established. All feature nodes in the cellular atypia evolution feature chain are traversed, identifying those with chromatin edge continuity measurements exceeding the predefined atypia threshold. The positions of the corresponding microscopic imaging units within the microscopic imaging unit spatial sequence are obtained. These located microscopic imaging units are marked as lesion core region units. The lesion core region units represent the areas with the most significant cellular atypia in the tissue section and serve as the starting seed points for lesion tracking.

[0068] Step S143: Obtain the hematoxylin staining concentration distribution channel of the core lesion region unit, perform boundary evolution tracking processing based on active contour model in the hematoxylin staining concentration distribution channel, take the cell nucleus segmentation region with the highest chromatin edge continuity metric in the core lesion region unit as the initial contour line, drive the initial contour line to iteratively deform and evolve towards the true boundary of the lesion region according to the hematoxylin staining concentration gradient field, and obtain the boundary contour of the core lesion region.

[0069] The hematoxylin staining concentration distribution channels of microscopic imaging units labeled as the core lesion region are obtained. Within this core lesion region, the nuclear segmentation region with the highest chromatin edge continuity is identified, and its contour line is used as the initial contour line. An active contour model is constructed, which undergoes iterative deformation driven by both internal and external energy. The internal energy constrains the smoothness and continuity of the contour line, while the external energy attracts the contour line to regions with high concentration gradients based on the hematoxylin staining concentration gradient field. Regions with high concentration gradients correspond to boundary positions where hematoxylin staining concentration changes drastically. During iteration, control points on the contour line move in the direction that minimizes the total energy. Evolution stops when the number of iterations reaches a preset upper limit or the contour line energy change is less than a preset threshold; the contour line at this point represents the boundary contour of the core lesion region.

[0070] Step S144: Spatial propagation processing is performed on the boundary contour of the lesion core region along the sequence direction of the spatial sequence of the microscopic imaging image units to adjacent microscopic imaging image units. The boundary contour of the lesion core region determined in the preceding microscopic imaging image unit is used as the initial constraint contour line for boundary tracking of the adjacent subsequent microscopic imaging image units. The boundary evolution direction in the adjacent subsequent microscopic imaging image units is guided by the constraint contour line to obtain the boundary contour of the lesion extension region of the adjacent subsequent microscopic imaging image units.

[0071] The boundary contour of the lesion core region obtained in step S143 is used as the initial contour. It is propagated along the spatial sequence of the microscopic imaging image units to adjacent subsequent microscopic imaging image units. The boundary contours already determined in the preceding microscopic imaging image units are projected onto the adjacent subsequent microscopic imaging image units, serving as the initial constraint contour lines for boundary tracking of those subsequent units. In the hematoxylin staining concentration distribution channel of that subsequent unit, starting from these initial constraint contour lines, the boundary evolution tracking process of the active contour model is performed again. Since the initial constraint contour lines are already close to the true boundary, the evolution process only needs fine-tuning to converge, obtaining the boundary contour of the lesion extension region of that subsequent microscopic imaging image unit. This process is iterated sequentially, propagating gradually along the sequence direction, to obtain the boundary contour of the lesion extension region of each adjacent subsequent unit.

[0072] Step S145: Connect all microscopic imaging image units in the spatial sequence of microscopic imaging image units that contain the boundary contour of the core region of the lesion or the boundary contour of the extended region of the lesion according to the spatial adjacency relationship, extract the set of boundary contour coordinate points corresponding to each microscopic imaging image unit, and perform end-to-end splicing of the boundary contour coordinate point sets between adjacent microscopic imaging image units to generate a lesion region boundary contour chain that extends continuously along the extension direction of the tissue slice.

[0073] Collect all microscopic imaging units in the spatial sequence of microscopic imaging units that contain the boundary contours of the core or extended regions of the lesion. For each unit, extract the coordinates of all pixels on its boundary contour to form a set of coordinate points. Following the order of the spatial sequence of microscopic imaging units, concatenate the coordinate point sets of adjacent units end-to-end, connecting the end point of the boundary contour of the previous unit to the beginning point of the boundary contour of the next unit. After concatenation, a long chain-like contour spanning multiple consecutive microscopic imaging units is obtained; this long chain-like contour is the lesion region boundary contour chain extending continuously along the direction of tissue section extension.

[0074] Step S146: Extract the nucleolar position offset parameter of each feature node in the cell heterogeneity evolution feature chain, and convert the nucleolar position offset parameter into a two-dimensional plane offset vector with the geometric center point of the cell nucleus as the starting point. The direction of the two-dimensional plane offset vector is determined by the azimuth angle of the nucleolar center position point relative to the geometric center point of the cell nucleus, and the magnitude of the two-dimensional plane offset vector is determined by the Euclidean distance offset.

[0075] For each feature node in the cell atypia evolution feature chain, the nucleolar position offset parameter of all nuclear segmentation regions within the corresponding microscopic imaging unit is obtained. Each nucleolar position offset parameter includes an Euclidean distance offset and the azimuth information of the nucleolar center position relative to the nuclear geometric center. Starting from the nuclear geometric center, a two-dimensional planar offset vector is constructed along the azimuth direction, with its magnitude equal to the Euclidean distance offset. Thus, the nucleolar position offset parameter of each nuclear segmentation region is converted into a two-dimensional planar offset vector.

[0076] Step S147: Based on the spatial positional relationship between adjacent microscopic imaging image units in the spatial sequence of the microscopic imaging image units, perform spatial interpolation fitting on the two-dimensional planar offset vectors of all cell nucleus regions within each microscopic imaging image unit to generate an offset vector field distribution map covering the entire microscopic imaging image unit region. Extract the regions in the offset vector field distribution map whose offset vector magnitude is greater than a preset offset significance threshold as significant chromatin feature offset regions, and extract the principal angular distribution direction of the offset vector direction within the significant chromatin feature offset regions. Use the principal angular distribution direction as the dominant chromatin feature offset direction within the microscopic imaging image unit.

[0077] For each microscopic imaging unit, two-dimensional planar offset vectors of all nucleus regions within that unit are obtained. These vectors are distributed along discrete coordinates of the nucleus center points. Using radial basis function interpolation, with the vector values ​​at these discrete points as known data points, an offset vector field covering the continuous space of the entire microscopic imaging unit region is fitted. Each pixel in this offset vector field has a corresponding offset vector. All pixels in the offset vector field are traversed, and pixels with offset vector magnitudes greater than a preset offset significance threshold are marked. The connected regions formed by these pixels constitute the chromatin feature offset significant regions. Within these chromatin feature offset significant regions, the orientation angles of all pixel offset vectors are statistically analyzed, generating an orientation angle histogram. The angle value corresponding to the angle interval with the highest frequency in the histogram is taken as the dominant orientation of the angular distribution in that region. This dominant orientation of the angular distribution is then used as the dominant orientation of the chromatin feature offset within that microscopic imaging unit.

[0078] Step S148: Connect the dominant directions of chromatin feature offset of each microscopic imaging image unit in the spatial sequence of microscopic imaging image units according to the spatial adjacency order of the microscopic imaging image units to generate a chromatin feature offset vector path with spatial position as the horizontal axis and offset direction angle as the vertical axis.

[0079] For each microscopic imaging unit in the spatial sequence of microscopic imaging units, the dominant direction angle value of the chromatin feature migration obtained in step S147 is acquired. Data points are plotted sequentially on a two-dimensional plane using the spatial sequence index of the microscopic imaging unit as the horizontal axis and the dominant direction angle value as the vertical axis. Data points corresponding to adjacent indices are connected by line segments to form a broken line. This broken line is the chromatin feature migration vector path, reflecting how the dominant direction of the nucleus-lamin migration changes as the tissue slice space extends.

[0080] Step S150: Load the boundary contour chain of the lesion area and the chromatin feature offset vector path into the preset pathological diagnosis quality control report template to generate a pathological quality control review report instruction containing the mapping relationship between the cell atypia evolution feature chain and the boundary contour chain of the lesion area.

[0081] Step S151: Obtain a preset pathological diagnosis quality control report template, which includes a first visualization layer for carrying image annotation information, a second visualization layer for carrying an evolutionary feature chain diagram, and a text description area for carrying diagnostic conclusion description text.

[0082] Obtain a pre-defined pathology diagnosis quality control report template. This template is structurally divided into three independent areas: the first visualization layer overlays image annotation information, such as boundary contours and marker points; the second visualization layer is used to draw trend graphs and feature chain diagrams; and the text description area is used to fill in structured diagnostic conclusions.

[0083] Step S152: Map the set of boundary contour coordinate points in the boundary contour chain of the lesion area to the pixel coordinate system of the first visualization layer, generate a closed polygon vector path according to the connection relationship of adjacent coordinate points in the boundary contour coordinate point set, and render the closed polygon vector path using the preset boundary line style and boundary line color parameters of the lesion area.

[0084] Obtain the lesion region boundary contour chain generated in step S145, which contains a set of coordinate points for the boundary contour on each microscopic imaging unit. Map these coordinate points from the original image coordinate system to the pixel coordinate system of the first visualization layer. For each closed boundary contour, connect adjacent points in the coordinate point set in order to form a line, and connect the first and last points to form a closed polygon vector path. Render the closed polygon vector path using a preset lesion region boundary line style and boundary line color parameters, for example, using a solid red line with a line width of 2 pixels.

[0085] Step S153: Extract the spatial position index field of the microscopic imaging image unit corresponding to each set of boundary contour coordinate points in the boundary contour chain of the lesion area, generate a position index identifier at the corresponding coordinate position of the first visualization layer based on the spatial position index field, and establish a visualization association relationship between the position index identifier and the closed polygon vector path.

[0086] For each boundary contour in the lesion region boundary contour chain, obtain the field of view location index field of its corresponding microscopic imaging unit, such as row number and column number. Convert this field of view location index field into a text string. In the first visualization layer, generate a text label near the geometric center of the boundary contour to display the field of view location index field. Establish a visual association between this text label and the corresponding closed polygon vector path using leaders or grouping, so that when viewing the report, it is clear which field of view location each boundary contour originates from.

[0087] Step S154: Establish a mapping relationship between the offset direction angle sequence in the chromatin feature offset vector path and the spatial position index field of the microscopic imaging image unit. In the second visualization layer, draw the offset vector path curve with the spatial position index field as the horizontal axis and the offset direction angle as the vertical axis. The offset vector path curve is drawn using a preset vector path line style and vector path color parameters.

[0088] Obtain the chromatin feature offset vector path generated in step S148, which contains the spatial sequence index and corresponding offset direction angle value of each microscopic imaging unit. Replace the spatial sequence index with the corresponding field of view position index field value. In the second visualization layer, plot the data points using the field of view position index field value as the horizontal axis coordinate and the offset direction angle value as the vertical axis coordinate. Connect adjacent data points with line segments using preset vector path line style and vector path color parameters to form the offset vector path curve.

[0089] Step S155: Draw a diagram of the node connection relationship of the cell atypia evolution feature chain in the second visualization layer. Draw a scatter plot of feature nodes with the spatial position index field of the microscopic imaging unit as the horizontal axis coordinate and the chromatin edge continuity measure as the vertical axis coordinate. Draw connecting line segments between nodes according to the connection relationship between the strengthening connection edge and the weakening connection edge in the cell atypia evolution feature chain.

[0090] Obtain the chromatin edge continuity metric value and the field of view index field of the corresponding microscopic imaging unit for each feature node in the cellular atypia evolution feature chain. In the second visualization layer, plot a scatter plot with the field of view index field as the horizontal axis and the chromatin edge continuity metric value as the vertical axis, where each scatter point represents a feature node. Based on the node connection relationship established in step S135, connect the two scatter points with a thick solid line for adjacent nodes with strengthening connection edges; connect the two scatter points with a thin dashed line for adjacent nodes with weakening connection edges.

[0091] Step S156: Extract the atypia evolution amplitude parameter and the atypia evolution trend consistency parameter of the cell atypia evolution feature chain, and convert the atypia evolution amplitude parameter and the atypia evolution trend consistency parameter into a qualitative descriptive text fragment that conforms to the description standard of pathological diagnosis report. The qualitative descriptive text fragment includes level words used to describe the degree of cell atypia and trend words used to describe the stability of the evolution trend.

[0092] Obtain the anomaly evolution amplitude parameter value calculated in step S136 and the anomaly evolution trend consistency parameter value calculated in step S137. Pre-define an amplitude-level mapping table to map the anomaly evolution amplitude parameter value to corresponding level terms, such as "mild anomaly," "moderate anomaly," and "severe anomaly." Pre-define a consistency-trend mapping table to map the anomaly evolution trend consistency parameter value to corresponding trend terms, such as "stable trend," "fluctuating trend," and "consistent trend." Combine the above level terms and trend terms into a complete qualitative descriptive text fragment, such as "exhibits moderate anomaly, with an overall consistent evolution trend."

[0093] Step S157: Based on the marking symbols used to indicate the production quality level in the quality control audit mark set, retrieve the production quality evaluation description text fragment corresponding to the marking symbol from the preset production quality evaluation text library, and fill the production quality evaluation description text fragment into the first text paragraph position of the text description area.

[0094] Extract the marker symbols used to indicate the slide preparation quality level from the quality control audit mark set generated in step S138. A slide preparation quality evaluation text library is pre-defined, where each quality level marker symbol corresponds to a standardized evaluation description text. Based on the extracted marker symbols, retrieve the corresponding slide preparation quality evaluation description text fragment from the text library, such as "excellent slide preparation quality, uniform cell distribution, and clear staining." Fill this text fragment into the first text paragraph position in the text description area of ​​the pathology diagnosis quality control report template.

[0095] Step S158: Based on the spatial extension length parameter of the boundary contour chain of the lesion area and the average area parameter of the area enclosed by the boundary contour, generate a quantitative descriptive text fragment to describe the lesion range and lesion degree, and fill the quantitative descriptive text fragment into the second text paragraph position of the text description area.

[0096] Calculate the spatial extension length parameter of the lesion region boundary contour chain, which is the spatial distance from the first microscopic imaging unit containing the lesion contour to the last microscopic imaging unit containing the lesion contour. Calculate the arithmetic mean of the area parameters of the regions enclosed by all boundary contours. Convert the spatial extension length parameter and the average area parameter into a quantitative descriptive text fragment, such as "The lesion region extends approximately X millimeters along the tissue section direction, and the average lesion area is approximately Y square micrometers." Fill this text fragment into the second text paragraph position in the text description area of ​​the pathology diagnostic quality control report template.

[0097] Step S159: The first and second visualization layers that have been rendered are overlaid and merged, and the merged visualization layer is combined with the filled text description area to form a pathology quality control review report page. The pathology quality control review report page is encoded into a pathology quality control review report instruction that conforms to the transmission format requirements of the medical information system.

[0098] The first visualization layer rendered in steps S152 and S153 is overlaid and merged with the second visualization layer rendered in steps S154 and S155, so that the lesion boundary outline and feature chain diagram are presented on the same image canvas. The merged visualization layer is combined with the text description area filled in steps S156, S157, and S158 to form a complete pathology quality control review report page. The data on this report page is encoded according to the requirements of the medical information system transmission format, such as encoding it into a document in the medical data exchange standard format, to generate the final pathology quality control review report instruction for remote pathology diagnosis terminal to receive and display.

[0099] Step S160: Obtain the chromatin edge continuity measure of all feature nodes in the cell atypia evolution feature chain, and arrange the chromatin edge continuity measure into a one-dimensional feature measure sequence along the extension direction of the tissue slice according to the spatial continuity relationship of the field of view position index field corresponding to the microscopic imaging image unit. Perform local trend decomposition processing based on sliding window on the one-dimensional feature measure sequence, set the window length parameter and the window sliding step parameter, and calculate the local linear fitting slope parameter and the local fitting residual fluctuation amplitude parameter of the chromatin edge continuity measure in each sliding window.

[0100] Obtain the continuous chromatin edge measurement values ​​for all feature nodes in the cellular atypia evolution feature chain. Based on the field-of-view index field of the microscopic imaging unit corresponding to each feature node, arrange these continuous chromatin edge measurement values ​​into a one-dimensional feature measurement sequence according to the tissue section extension direction. Set a window length parameter, for example, to cover the span of 5 consecutive microscopic imaging units, and a window sliding step size parameter, for example, to move one microscopic imaging unit at a time. Starting from the beginning of the sequence, take a continuous subsequence within one window length each time. For each subsequence, fit a linear regression line using the least squares method to obtain the local linear fitting slope parameter for that subsequence. Simultaneously, calculate the residual of each data point in the subsequence relative to the fitted line, and take the average of the absolute values ​​of all residuals as the local fitting residual fluctuation amplitude parameter. Move the sliding window forward according to the step size, repeating the above calculation until the entire sequence is covered.

[0101] Step S161: Based on the positive and negative polarity change characteristics of the local linear fitting slope parameter, the one-dimensional feature measurement sequence is divided into an upward trend segment, a stable fluctuation segment, and a downward trend segment, and a segment trend type identifier is assigned to each segment.

[0102] The local linear fitting slope parameter calculated for each sliding window in step S160 is iterated. When the local linear fitting slope parameter is positive and its absolute value is greater than a preset trend threshold, the interval corresponding to that window is marked as an upward trend interval. When the local linear fitting slope parameter is negative and its absolute value is greater than the preset trend threshold, the interval corresponding to that window is marked as a downward trend interval. When the absolute value of the local linear fitting slope parameter is less than or equal to the preset trend threshold, the interval corresponding to that window is marked as a stable fluctuation interval. Adjacent intervals with the same trend label are merged into a continuous segment. For each merged segment, a corresponding segment trend type identifier is assigned according to its trend type. For example, +1 represents an upward trend segment, 0 represents a stable fluctuation segment, and -1 represents a downward trend segment.

[0103] Step S162: Extract the sequence values ​​of the local fitting residual fluctuation amplitude parameter within each stationary fluctuation segment, calculate the sequence autocorrelation function of the local fitting residual fluctuation amplitude parameter, and determine whether there is a periodic fluctuation pattern in the chromatin edge continuity measurement within the stationary fluctuation segment based on the decay characteristics of the sequence autocorrelation function. When a periodic fluctuation pattern exists within the stationary fluctuation segment, calculate the fluctuation period length parameter and fluctuation amplitude parameter of the periodic fluctuation pattern, and use the fluctuation period length parameter and fluctuation amplitude parameter as the basis for evaluating the slide quality stability of the microscopic imaging unit region corresponding to the stationary fluctuation segment.

[0104] For each interval marked as a stationary fluctuation segment, the local fitting residual fluctuation amplitude parameter corresponding to all windows within that segment is extracted to form a subsequence. The autocorrelation function of this subsequence is calculated, i.e., the correlation coefficient between the subsequence and itself at different time lags is calculated. The decay characteristics of the autocorrelation function are analyzed: if the autocorrelation function exhibits periodic oscillations with increasing lag and decays slowly, then a periodic fluctuation pattern is determined to exist within the stationary fluctuation segment. If a periodic fluctuation pattern exists, the lag value corresponding to the first peak of the autocorrelation function is taken as the fluctuation period length parameter, and the height of this peak is taken as the fluctuation amplitude parameter. The fluctuation period length parameter and fluctuation amplitude parameter are used as the basis for evaluating the slide quality stability of the microscopic imaging unit region covered by the stationary fluctuation segment. The shorter the period length and the larger the amplitude, the more likely there are periodic staining or focusing instabilities during the slide preparation process.

[0105] Step S163: Based on the magnitude of the slope parameter of the local linear fitting within the upward trend segment and the segment duration parameter, determine the degree of cell atypia aggravation corresponding to the upward trend segment, and establish a correlation between the degree of cell atypia aggravation and the index range of the visual field location covered by the upward trend segment.

[0106] For each interval marked as an upward trend segment, the local linear fit slope parameters of all windows within that segment are obtained, and the arithmetic mean of these slope parameters is calculated as the average upward slope of the segment. Simultaneously, the length of the visual field location index range covered by the segment is calculated, i.e., the distance between the segment's starting and ending visual field indices, as the segment duration parameter. A slope-level mapping table and a length-level mapping table are pre-defined. The slope contribution level is obtained by looking up the table based on the numerical interval of the average upward slope, and the length contribution level is obtained by looking up the table based on the numerical interval of the segment duration parameter. The two levels are then weighted and summed to obtain the degree of cellular atypia aggravation level. This aggravation level is then correlated with the visual field location index range covered by the upward trend segment, for example, recorded as "within the range of FOV_010 to FOV_025, the degree of atypia aggravation is moderate."

[0107] Step S164: Load the fragment trend type identifier, the degree of cell atypia aggravation, and the evaluation criteria for slide quality stability into the supplementary analysis description area of ​​the pathological diagnosis quality control report template to generate an analysis description text paragraph that describes the stage characteristics of lesion evolution and the characteristics of slide quality stability.

[0108] Combine the fragment trend type identifier obtained in step S161, the degree of cellular atypia aggravation obtained in step S163, and the evaluation criteria for slide preparation quality stability obtained in step S162 (including the fluctuation period length parameter and fluctuation amplitude parameter). Convert the above data into natural language descriptive text, such as "The first half of the tissue section shows an upward trend, with moderately aggravated cellular atypia; the second half shows stable fluctuations with a periodic fluctuation pattern, the period length is approximately X fields, and the amplitude is Y, suggesting that there may be periodic vibrations during slide preparation." Load this analytical descriptive text paragraph into the supplementary analysis description area of ​​the pathology diagnosis quality control report template.

[0109] Step S170: Extract the perimeter parameter and the area parameter of the region enclosed by the contour of each lesion core region in the lesion region boundary contour chain. Calculate the morphological complexity parameter of the lesion core region boundary contour based on the perimeter parameter and the area parameter of the region enclosed by the contour. The morphological complexity parameter is the ratio of the square of the perimeter parameter to the area parameter of the region enclosed by the contour.

[0110] Traverse all lesion core region boundary contours in the lesion region boundary contour chain. For each boundary contour, calculate its contour perimeter parameter, which is the sum of the Euclidean distances between adjacent pixels on the contour line. Calculate the area parameter of the region enclosed by the contour, which is the number of pixels contained within the contour. Calculate the morphological complexity parameter by squaring the contour perimeter parameter and dividing the squared result by the area parameter of the region enclosed by the contour. The ratio obtained is the morphological complexity parameter of the lesion core region boundary contour. The larger the value of this morphological complexity parameter, the more irregular and complex the boundary contour, reflecting the irregular growth characteristics of the lesion region.

[0111] Step S171: Extract the perimeter parameter of the boundary contour of each lesion expansion region in the lesion region boundary contour chain and the area parameter of the region enclosed by the contour. Calculate the area expansion rate parameter of the lesion expansion region based on the change in the area parameter of the region enclosed by the boundary contours of adjacent microscopic imaging units.

[0112] Traverse all lesion extension region boundary contours in the lesion region boundary contour chain. For each boundary contour, calculate its contour perimeter parameter and the area parameter of the region enclosed by the contour according to the method in step S170. For two adjacent units in the spatial sequence of microscopic imaging image units, obtain the area parameter of the region enclosed by the contour of the subsequent unit and the area parameter of the region enclosed by the contour of the preceding unit, and calculate the difference between the two. Divide this difference by the spatial distance parameter between the preceding and subsequent units to obtain the area expansion rate parameter. This area expansion rate parameter reflects how fast the lesion region expands in the spatial extension direction of the tissue section.

[0113] Step S172: Construct a two-dimensional feature vector to describe the growth behavior characteristics of the lesion region based on the morphological complexity parameter and the area expansion rate parameter, map the two-dimensional feature vector to a preset lesion growth pattern classification space, and determine the lesion growth pattern category identifier corresponding to the boundary contour chain of the lesion region.

[0114] The morphological complexity parameter calculated in step S170 and the area expansion rate parameter calculated in step S171 are combined into a two-dimensional feature vector. The first component of this two-dimensional feature vector is the morphological complexity parameter, and the second component is the area expansion rate parameter. A lesion growth pattern classification space is pre-defined. This lesion growth pattern classification space is a two-dimensional planar region division obtained by cluster analysis based on the morphological complexity parameter and area expansion rate parameter of a large number of historical pathological samples. Each region corresponds to a lesion growth pattern category, such as "expansive growth pattern", "infiltrative growth pattern", and "nodular growth pattern". This two-dimensional feature vector is projected into this classification space, and the corresponding lesion growth pattern category identifier is determined according to the region it falls into.

[0115] Step S173: Extract the dominant direction angle of chromatin feature offset corresponding to each microscopic imaging image unit in the chromatin feature offset vector path, calculate the change in angular deflection of the dominant direction angle between adjacent microscopic imaging image units, and determine the overall deflection direction trend of chromatin feature offset based on the cumulative value of the change in angular deflection.

[0116] From the dominant direction angles of chromatin feature shifts of each microscopic imaging unit obtained in step S147, extract the dominant direction angle values ​​of all units to form an angle sequence. For each pair of adjacent units in the sequence, calculate the angle value of the subsequent unit minus the angle value of the preceding unit to obtain the angle deflection change. If the absolute value of the difference is greater than a preset angle threshold, angle correction processing on the direction circle is required, for example, adjusting the deflection greater than 180 degrees to the value after subtracting 360 degrees. Sum all the angle deflection changes to obtain the cumulative deflection. The overall deflection direction trend is determined according to the positive or negative sign of the cumulative deflection: a positive cumulative deflection indicates an overall counterclockwise deflection trend, and a negative cumulative deflection indicates an overall clockwise deflection trend.

[0117] Step S174: Obtain the gradient parameter of the change in chromatin edge continuity between adjacent feature nodes in the cell atypia evolution feature chain, extract the sequence value of the gradient parameter, and calculate the mean and coefficient of variation of the gradient parameter sequence values. Based on the mean and coefficient of variation of the gradient parameter sequence values, determine the degree of evolution severity of the cell atypia evolution feature chain, and combine the degree of evolution severity, the lesion growth pattern category identifier, and the overall deflection direction trend to generate a comprehensive lesion behavior feature descriptor.

[0118] Obtain the gradient parameters of change between all adjacent feature nodes calculated in step S132, forming a gradient parameter sequence. Calculate the arithmetic mean of this gradient parameter sequence to obtain the mean value of the gradient parameter sequence. Calculate the standard deviation of this gradient parameter sequence, divide the standard deviation by the mean, and obtain the coefficient of variation parameter. Pre-set a mean-level mapping table and a coefficient of variation-level mapping table. Look up the mean contribution level in the table based on the numerical interval of the mean, and look up the variation contribution level in the table based on the numerical interval of the coefficient of variation. Sum the two levels with weights to obtain the degree of evolution severity level. Combine the degree of evolution severity level, the lesion growth pattern category identifier obtained in step S172, and the overall deflection direction trend obtained in step S173 into a data structure, namely, the lesion comprehensive behavior feature descriptor.

[0119] Step S175: Encode the comprehensive behavioral feature descriptor of the lesion into a structured supplementary diagnostic information data segment, and attach the supplementary diagnostic information data segment to the extended data area of ​​the pathology quality control review report instruction for reference by the remote pathology diagnostic terminal.

[0120] The comprehensive behavioral characteristic descriptor of the lesion generated in step S174 is structured and encoded according to the medical data exchange standard format, such as encoding it into an entry in Extensible Markup Language (XML) or the medical data exchange standard format. This encoded data segment is then appended as supplementary diagnostic information to the extended data area of ​​the pathology quality control review report instruction generated in step S159. Upon receiving this report instruction, the remote pathology diagnostic terminal can parse the comprehensive behavioral characteristic descriptor of the lesion from the extended data area, providing pathologists with a more in-depth diagnostic reference.

[0121] Step S180: Obtain the overlapping boundary region between the microscopic imaging image units corresponding to the adjacent field of view position index fields in the digital image set of the pathological slide, and extract the pixel value distribution of the hematoxylin staining concentration distribution channel and the pixel value distribution of the eosin staining concentration distribution channel within the overlapping boundary region.

[0122] For each pair of adjacent microscopic imaging units in the field of view location index field of a pathological slide digital image set, such as units adjacent left and right or top and bottom, their spatial overlapping boundary region is determined. This overlapping boundary region is generated due to a fixed overlap ratio during continuous acquisition by the scanning device. Within this overlapping boundary region, the pixel value distribution of the two units in the hematoxylin staining concentration distribution channel is extracted, i.e., the set of hematoxylin staining concentration values ​​of all pixels in the overlapping region. Similarly, the pixel value distribution of the two units in the eosin staining concentration distribution channel is extracted.

[0123] Step S181: Calculate the consistency measure of the concentration distribution of pixel values ​​of the hematoxylin staining concentration distribution channel in the overlapping boundary region between adjacent microscopic imaging image units. The consistency measure of concentration distribution is obtained by comparing the differences in the concentration values ​​of hematoxylin staining at the same pixel position in the overlapping region of adjacent microscopic imaging image units.

[0124] For each pixel location within the overlapping boundary region, obtain the hematoxylin staining concentration values ​​for the first and second microscopic imaging units at that location. Calculate the absolute value of the difference between the two values. Iterate through all pixel locations within the overlapping boundary region, summing the absolute differences across all locations, and then divide by the total number of pixels within the overlapping boundary region to obtain the average absolute difference value. This average absolute difference value is the measure of concentration distribution consistency. The smaller the average absolute difference value, the more consistent the hematoxylin staining concentration distribution between two adjacent units in the overlapping region.

[0125] Step S182: Calculate the texture structure consistency measure of the pixel values ​​of the eosin staining concentration distribution channel in the overlapping boundary region between adjacent microscopic imaging image units. The texture structure consistency measure is obtained by comparing the similarity of the cytoplasmic texture primitive encoding map in the overlapping region of adjacent microscopic imaging image units.

[0126] For each pixel location within the overlapping boundary region, obtain the encoding value of the cytoplasmic texture primitive encoding map generated in step S125 in the first microscopic imaging image unit, and the encoding value in the second microscopic imaging image unit. Determine if the two encoding values ​​are equal. Traverse all pixel locations within the overlapping boundary region, count the number of pixels with equal encoding values, and divide this number by the total number of pixels within the overlapping boundary region to obtain the encoding consistency rate. This encoding consistency rate is the texture structure consistency measure. The higher the encoding consistency rate, the more similar the cytoplasmic texture structure of the two adjacent units in the overlapping region.

[0127] Step S183: Construct an imaging quality continuity parameter for the overlapping boundary region based on the concentration distribution consistency metric and the texture structure consistency metric, and compare the imaging quality continuity parameter with a preset imaging quality continuity evaluation threshold to generate an evaluation result for evaluating the image stitching quality during the pathological slide scanning process.

[0128] The concentration distribution consistency measure obtained in step S181 is denoted as D_color, and the texture structure consistency measure obtained in step S182 is denoted as D_texture. The two measures are weighted and summed to obtain the imaging quality continuity parameter. For example, the weight of the concentration distribution consistency measure is set to 0.4, and the weight of the texture structure consistency measure is set to 0.6. A preset imaging quality continuity evaluation threshold is established. The calculated imaging quality continuity parameter is compared with this evaluation threshold: if the imaging quality continuity parameter is greater than the evaluation threshold, an evaluation result of "image stitching quality is qualified" is generated; if it is less than or equal to the evaluation threshold, an evaluation result of "image stitching quality is unqualified, stitching traces exist" is generated.

[0129] Step S184: Extract the chromatin edge continuity measure from the morphological distribution characteristics of cell nuclei within the overlapping boundary region, calculate the degree of difference in the chromatin edge continuity measure within the overlapping boundary region between adjacent microscopic imaging units, and use the degree of difference as an auxiliary reference for evaluating the staining uniformity during the preparation of pathological slides.

[0130] Within the overlapping boundary region, the chromatin edge continuity measure of all nuclear segmentation regions in the first microscopic imaging unit and the chromatin edge continuity measure of all nuclear segmentation regions in the second microscopic imaging unit are obtained. The arithmetic mean of the chromatin edge continuity measures for both units within the overlapping boundary region is calculated. The absolute value of the difference between the two means is calculated as the degree of difference in chromatin edge continuity measure. This degree of difference reflects the consistency of nuclear chromatin representation between adjacent fields of view; a smaller degree of difference indicates more uniform staining during slide preparation, while a larger degree suggests potential uneven staining.

[0131] Step S185: Based on the difference between the imaging quality continuity parameter and the chromatin edge continuity measure, generate a digital acquisition quality assessment data segment for pathological slides. The digital acquisition quality assessment data segment for pathological slides includes an image stitching quality level identifier and a staining uniformity level identifier.

[0132] The imaging quality continuity parameter generated in step S183 is mapped to image stitching quality level identifiers according to a preset level division interval, such as "stitching quality level A", "stitching quality level B", and "stitching quality level C". The degree of difference in chromatin edge continuity measurement calculated in step S184 is mapped to staining uniformity level identifiers according to a preset degree of difference interval, such as "excellent uniformity", "good uniformity", and "poor uniformity". The above two identifiers are combined into a digital acquisition quality assessment data segment for pathological slides.

[0133] Step S186: Embed the digital acquisition quality assessment data segment of the pathological slide into the report header information area of ​​the pathological quality control review report instruction.

[0134] The pathology slide digitization acquisition quality assessment data segment generated in step S185 is written into the report header information area of ​​the pathology quality control review report instruction generated in step S159. The report header information area is typically used to store metadata and does not affect the main content of the report, but it can be read by the parsing system. When the remote pathology diagnostic terminal receives the report instruction, it can extract this quality assessment data segment from the report header information area to understand the quality status of the digitization acquisition process.

[0135] For example, the method may further include: step S190: extracting nucleolar position offset parameters of all feature nodes in the cell atypia evolution feature chain, arranging the nucleolar position offset parameters along the spatial sequence of the microscopic imaging image unit to generate a nucleolar position offset parameter sequence, performing distribution characteristic analysis processing based on directional statistics on the nucleolar position offset parameter sequence, and calculating the circumferential distribution concentration parameter of the offset direction angle in the nucleolar position offset parameters, wherein the circumferential distribution concentration parameter is used to characterize the degree of consistency of the spatial distribution of the nucleolar position offset direction in the cell nucleus.

[0136] Nucleolar position offset parameters for all feature nodes in the cellular atypia evolution feature chain are obtained. Each nucleolar position offset parameter contains two components: offset direction angle and offset distance. Following the spatial sequence of microscopic imaging units, the offset direction angles of all feature nodes are extracted to form an angle sequence. Orientational statistical analysis is performed on this angle sequence, treating each angle as a unit vector on a circle. The sum of all unit vectors is calculated, i.e., the sum of the horizontal and vertical components of each unit vector. The magnitude of this sum is divided by the total number of angles in the sequence to obtain the circumferential distribution concentration parameter. The circumferential distribution concentration parameter ranges from 0 to 1; a value closer to 1 indicates that all offset directions are more concentrated in the same direction, while a value closer to 0 indicates that the offset directions are more dispersed on the circle.

[0137] Step S191: Calculate the statistical distribution characteristics of the offset distance values ​​in the nucleolus position offset parameters. The statistical distribution characteristics include the distribution center position parameter and the distribution discrete width parameter of the offset distance values. Construct a nucleolus spatial offset feature vector based on the circumferential distribution concentration parameter and the distribution center position parameter. Input the nucleolus spatial offset feature vector into a preset nucleolus offset pattern discrimination function to obtain the nucleolus offset pattern classification result corresponding to the pathological slide digital image set.

[0138] Obtain all offset distance values ​​corresponding to the nucleus position offset parameter sequence in step S190, forming an offset distance value sequence. Calculate the arithmetic mean of this offset distance value sequence as the distribution center position parameter of the offset distance values. Calculate the standard deviation of this offset distance value sequence as the distribution dispersion width parameter of the offset distance values. Combine the circumferential distribution concentration parameter calculated in step S190 and the distribution center position parameter of the offset distance values ​​into a two-dimensional nucleus spatial offset feature vector. Preset a nucleus offset pattern discrimination function, which can be a classifier trained based on historical samples, such as a linear discriminant analysis classifier or a support vector machine classifier. Input the nucleus spatial offset feature vector into this discrimination function, and output a nucleus offset pattern classification result, such as "central offset pattern", "eccentric offset pattern", or "random offset pattern".

[0139] Step S192: Extract the cytoplasmic fiber orientation consistency parameter from the cytoplasmic texture topological structure features, arrange the cytoplasmic fiber orientation consistency parameter along the spatial sequence of the microscopic imaging image unit to generate a cytoplasmic fiber orientation consistency parameter sequence, and calculate the sequence correlation degree measure between the cytoplasmic fiber orientation consistency parameter sequence and the offset distance value sequence in the nucleolar position offset parameter sequence. The sequence correlation degree measure is used to characterize the close correlation between the orderly arrangement of cytoplasmic fibers and the degree of abnormality in the spatial position of the nucleolar.

[0140] Cytoplasmic fiber orientation consistency parameters are extracted from the cytoplasmic texture topology features of each microscopic imaging unit obtained in step S127. These cytoplasmic fiber orientation consistency parameters are arranged into a sequence according to the spatial sequence of the microscopic imaging units. Simultaneously, the offset distance numerical portion is extracted from the nucleolar position offset parameter sequence generated in step S190, forming an offset distance numerical sequence. The Pearson correlation coefficient between the two sequences is calculated, i.e., the product of the covariance of the two sequences and their respective standard deviations, to obtain a measure of sequence correlation. This measure of sequence correlation ranges from -1 to +1; a positive value indicates that higher cytoplasmic fiber orientation consistency results in a larger nucleolar offset distance, while a negative value indicates an inverse relationship between the two.

[0141] Step S193: Based on the nucleolar offset pattern classification results and the sequence correlation degree measure, generate an association anomaly analysis descriptor for cell nucleoli and cytoplasmic texture, and use the association anomaly analysis descriptor as supplementary cytological anomaly evaluation content in the pathology quality control review report instruction.

[0142] The nucleolar shift pattern classification result obtained in step S191 and the sequence correlation measure obtained in step S192 are combined into an association anomaly analysis descriptor. For example, if the nucleolar shift pattern classification result is "eccentric shift pattern" and the sequence correlation measure is greater than a preset positive correlation threshold, the descriptor "eccentric nucleolar shift is highly correlated with cytoplasmic fiber orientation, suggesting that cytoskeleton remodeling may drive nucleolar displacement" is generated. If the nucleolar shift pattern classification result is "random shift pattern" and the absolute value of the sequence correlation measure is less than a preset correlation threshold, the descriptor "random nucleolar shift direction, with no obvious correlation to cytoplasmic fiber arrangement" is generated. This association anomaly analysis descriptor is added as supplementary cytological anomaly evaluation content to the pathology quality control review report instructions.

[0143] Step S194: Extract the nucleolar position offset parameter of the nucleus morphology distribution feature within the microscopic imaging image unit corresponding to each feature node in the cell atypia evolution feature chain. Based on the spatial continuity relationship of the field of view position index field corresponding to the microscopic imaging image unit, arrange the offset direction angle of the nucleolar position offset parameter along the extension direction of the tissue slice into an offset direction angle sequence, and arrange the offset distance value of the nucleolar position offset parameter along the extension direction of the tissue slice into an offset distance value sequence.

[0144] For each feature node in the cellular atypia evolution feature chain, the nucleolar position offset parameters of all segmented nuclei within the corresponding microscopic imaging unit are obtained. For each microscopic imaging unit, the arithmetic mean of the offset direction angles and the arithmetic mean of the offset distance values ​​of all nucleolar position offset parameters within that unit are calculated. Following the spatial sequence of the microscopic imaging units, the average offset direction angles of each unit are arranged into an offset direction angle sequence, and the average offset distance values ​​of each unit are arranged into an offset distance value sequence.

[0145] Step S195: Perform directional stability analysis processing based on a circular data sliding window on the offset direction angle sequence, set the directional analysis window length parameter and the directional analysis window sliding step size parameter, and calculate the circumferential mean direction and circumferential direction concentration parameter of the offset direction angle in each directional analysis window. The circumferential direction concentration parameter is the ratio of the magnitude of the resultant vector of the offset direction angle on the circumference in the window to the number of offset direction angles in the window.

[0146] Define a direction analysis window length parameter, for example, a span covering 5 consecutive microscopic imaging units, and a direction analysis window sliding step size parameter, for example, a step size of 1 microscopic imaging unit per movement. Starting from the beginning of the offset direction angle sequence, take a continuous subsequence within one window length each time. For each subsequence, treat each angle as a unit vector on the circumference, calculate the vector sum of all unit vectors, and use the direction angle of this vector sum as the circumferential mean direction of the window. Divide the magnitude of this vector sum by the number of angles within the window to obtain the circumferential direction concentration parameter. The sliding window moves backward according to the step size, repeating the above calculation until the entire sequence is covered.

[0147] Step S196: Based on the change curve of the circumferential concentration parameter along the extension direction of the tissue section, the continuous segments in the change curve where the circumferential concentration parameter is lower than the preset directional concentration attenuation threshold are marked as nucleolar polarity disordered segments. The nucleolar polarity disordered segments reflect the range of tissue section locations where the polarity arrangement of the cell nucleus is significantly abnormal.

[0148] The circumferential concentration parameters calculated for each window in step S195 are arranged according to the spatial position corresponding to the window center, resulting in a variation curve. A preset directional concentration attenuation threshold is established. The variation curve is scanned to identify all windows where the circumferential concentration parameters are lower than the preset directional concentration attenuation threshold. The spatial position intervals corresponding to the above windows are merged to obtain continuous spatial segments. These segments are marked as nucleolar polarity disordered segments. These segments reflect that within the location of the tissue section, the offset direction of the nucleus and nucleolus is highly dispersed, and the polarity arrangement pattern is significantly abnormal.

[0149] Step S197: Extract the field of view position index field range corresponding to the microscopic imaging image unit covered by the nucleolar polarity disordered segment, extract the cytoplasmic fiber orientation consistency parameter in the cytoplasmic texture topology features of each microscopic imaging image unit within the field of view position index field range, and correlate the cytoplasmic fiber orientation consistency parameter with the circumferential direction concentration parameter within the nucleolar polarity disordered segment.

[0150] Obtain the start and end values ​​of the field-of-view location index field covered by each nucleolar polarity disordered segment marked in step S196. Within the range of this field-of-view location index field, extract the cytoplasmic fiber orientation consistency parameter for each microscopic imaging unit. Simultaneously, obtain the circumferential concentration parameter corresponding to each window within this range. Correspond to both according to the same spatial location, that is, record the cytoplasmic fiber orientation consistency parameter and the circumferential concentration parameter simultaneously on the same spatial coordinate.

[0151] Step S198: Calculate the local cross-correlation function between the circumferential concentration parameter and the cytoplasmic fiber orientation consistency parameter within the nucleolar polarity disordered segment. Based on the peak position of the local cross-correlation function, determine the spatial lag distance parameter between the nucleolar polarity disorder and the abnormal cytoplasmic fiber arrangement. The lag distance parameter represents the sequential relationship of the propagation of nucleolar polarity changes and cytoplasmic texture changes in the tissue section space.

[0152] Within the nucleolar polarity disorder region, the circumferential concentration parameter sequence is used as the first sequence, and the cytoplasmic fiber alignment consistency parameter sequence is used as the second sequence. The cross-correlation function between the two sequences is calculated; that is, for different spatial lag offsets, the sum of the dot products of the first sequence and the translated second sequence is calculated. The lag offset corresponding to the maximum value of the cross-correlation function is found; this lag offset is the lag distance parameter. A positive lag distance parameter indicates that abnormal cytoplasmic fiber arrangement lags behind nucleolar polarity disorder, while a negative lag distance parameter indicates that nucleolar polarity disorder lags behind abnormal cytoplasmic fiber arrangement.

[0153] Step S199: Determine the evolutionary driving relationship identifier between nucleolar polarity disorder and cytoplasmic fiber arrangement abnormality based on the positive or negative sign of the hysteresis distance parameter. When the hysteresis distance parameter is positive, it indicates that the cytoplasmic fiber arrangement abnormality lags behind the nucleolar polarity disorder, generating a driving relationship identifier for cytoplasmic texture reconstruction driven by nucleolar polarity change. When the hysteresis distance parameter is negative, it indicates that the nucleolar polarity disorder lags behind the cytoplasmic fiber arrangement abnormality, generating a driving relationship identifier for nucleolar polarity shift driven by changes in the cytoplasmic microenvironment.

[0154] Determine the sign of the hysteresis distance parameter calculated in step S198. If the hysteresis distance parameter is positive, it indicates that nucleolar polarity disorder occurred first during spatial propagation, followed by abnormal cytoplasmic fiber arrangement, thus generating an evolutionary driving relationship identifier of "nucleolus driving cytoplasm". If the hysteresis distance parameter is negative, it indicates that abnormal cytoplasmic fiber arrangement occurred first, followed by nucleolar polarity disorder, thus generating an evolutionary driving relationship identifier of "cytoplasm driving nucleolus".

[0155] Step S200: Combine the field of view location index field range of the nucleolar polarity disordered segment, the change curve of the circumferential direction concentration parameter, the hysteresis distance parameter, and the evolution driving relationship identifier into a nucleolar polarity dynamic evolution feature descriptor, and encode the nucleolar polarity dynamic evolution feature descriptor into a structured data block and append it to the extended data area of ​​the pathology quality control review report instruction.

[0156] The visual field location index field range of the nucleolar polarity disordered segment obtained in step S196, the change curve of the circumferential concentration parameter obtained in step S195, the hysteresis distance parameter obtained in step S198, and the evolution driving relationship identifier obtained in step S199 are combined into a nucleolar polarity dynamic evolution feature descriptor. This descriptor is encoded into a structured data block according to the medical data exchange standard format, such as encoded as nested entries in Extensible Markup Language (XML) format. This data block is appended to the extended data area of ​​the pathology quality control review report instruction for advanced analysis and diagnostic reference by remote pathology diagnostic terminals.

[0157] Step S210: Obtain the chromatin edge continuity measure and the cytoplasmic vacuoles spatial distribution measure in the cell nucleus morphology distribution features of each microscopic imaging image unit in the microscopic imaging image unit spatial sequence. Based on the field of view position index field corresponding to the microscopic imaging image unit, arrange the chromatin edge continuity measure and the cytoplasmic vacuoles spatial distribution measure along the extension direction of the tissue section as the chromatin edge continuity measure sequence and the cytoplasmic vacuoles spatial distribution measure sequence, respectively.

[0158] For each microscopic imaging unit in the spatial sequence of microscopic imaging units, a continuous chromatin edge metric is extracted from the nuclear morphology distribution features obtained in step S127, and a spatial distribution metric of cytoplasmic vacuoles is extracted from the cytoplasmic texture topology features. Following the order of the spatial sequence of microscopic imaging units, the continuous chromatin edge metrics are arranged into a continuous chromatin edge metric sequence, and the number of cytoplasmic vacuoles per unit area in the spatial distribution metric is used as a representative value to form a spatial distribution metric sequence of cytoplasmic vacuoles.

[0159] Step S211: Perform nonlinear dynamic analysis based on phase space reconstruction on the chromatin edge continuous measurement sequence and the cytoplasmic vacuoles spatial distribution measurement sequence, set the embedding dimension parameter and the delay time parameter, and reconstruct the chromatin edge continuous measurement sequence and the cytoplasmic vacuoles spatial distribution measurement sequence into a high-dimensional phase space according to the embedding dimension parameter and the delay time parameter, respectively, to generate the phase space trajectory of the chromatin edge continuous measurement and the phase space trajectory of the cytoplasmic vacuoles spatial distribution measurement.

[0160] Let m be an embedding dimension parameter and τ be a delay time parameter. For a continuous chromatin edge measurement sequence, each data point x(t) is expanded into an m-dimensional vector: X(t) = [x(t), x(t+τ), x(t+2τ), ..., x(t+(m-1)τ)]. Traversing the entire sequence yields a series of m-dimensional vectors, whose distribution in the high-dimensional phase space constitutes the phase space trajectory of the continuous chromatin edge measurement. Using the same method, the same phase space reconstruction process is performed on the cytoplasmic vacuolar spatial distribution measurement sequence to generate the phase space trajectory of the cytoplasmic vacuolar spatial distribution measurement.

[0161] Step S212: Calculate the correlation dimension parameter of the phase space trajectory of the chromatin edge continuous measurement. The correlation dimension parameter is obtained by statistically analyzing the scaling relationship of the number of point pairs in the phase space trajectory whose distance is less than a preset distance threshold as a function of the preset distance threshold. The correlation dimension parameter is used to characterize the dynamic complexity of the chromatin edge continuous measurement sequence in phase space.

[0162] For all point pairs in the phase space trajectory of a continuous chromatin edge measure, calculate the Euclidean distance between each pair. Set a distance threshold variable *r*, and count the proportion of point pairs with a distance less than *r* out of the total number of point pairs, obtaining the correlation integral function *C(r)*. Plot the curve of *logC(r)* versus *logr* in a log-log coordinate system. Select a linear scaling interval on the curve and calculate the slope of *logC(r)* relative to *logr* within that interval; this slope is the correlation dimension parameter. The correlation dimension parameter reflects the geometric complexity of the phase space trajectory; a larger value indicates a more complex dynamic behavior of the system.

[0163] Step S213: Calculate the maximum Lyapunov exponent parameter of the phase space trajectory of the cytoplasmic vacuolar spatial distribution measure. The maximum Lyapunov exponent parameter is obtained by tracking the exponential divergence rate of the distance between adjacent orbital point pairs in the phase space trajectory. The maximum Lyapunov exponent parameter is used to characterize the sensitivity dependence of the cytoplasmic vacuolar spatial distribution measure sequence on the initial state.

[0164] For the phase space trajectory measuring the spatial distribution of cytoplasmic vacuoles, find the nearest neighbor X_nearest(t) for each point X(t) and calculate the distance L(t) between them. Track the evolution of these two points in phase space and calculate the distance L'(t) after a time step Δt. For multiple distinct t values, calculate log(L'(t) / L(t)), then average over all t values ​​and divide by Δt to obtain an estimate of the maximum Lyapunov exponent parameter. A positive value of the maximum Lyapunov exponent parameter indicates that the system is sensitive to initial conditions and exhibits chaotic characteristics; a negative value indicates that the system tends to be stable; and a value of zero indicates that the system is in a periodic motion state.

[0165] Step S214: Construct a two-dimensional chaotic feature vector to describe the dynamic characteristics of cell atypia evolution based on the correlation dimension parameter and the maximum Lyapunov exponent parameter. Map the two-dimensional chaotic feature vector to a preset lesion evolution dynamics pattern classification space. Determine the lesion evolution dynamics pattern category identifier corresponding to the set of digital images of pathological slides. The lesion evolution dynamics pattern category identifier includes a deterministic chaotic evolution pattern identifier, a random fluctuation evolution pattern identifier, and a critical phase transition evolution pattern identifier.

[0166] The correlation dimension parameter calculated in step S212 and the maximum Lyapunov exponent parameter calculated in step S213 are combined into a two-dimensional chaotic feature vector. A disease evolution dynamics pattern classification space is pre-defined, which is obtained through cluster analysis based on the correlation dimension parameter and the maximum Lyapunov exponent parameter of a large number of historical pathological samples. The two-dimensional chaotic feature vector is projected into this classification space: if the correlation dimension parameter is non-integer and the maximum Lyapunov exponent parameter is positive, it is determined as a deterministic chaotic evolution pattern, and a corresponding deterministic chaotic evolution pattern identifier is generated; if the correlation dimension parameter is close to zero and the maximum Lyapunov exponent parameter is negative, it is determined as a random fluctuation evolution pattern, and a corresponding random fluctuation evolution pattern identifier is generated; if the correlation dimension parameter is near a critical value and the maximum Lyapunov exponent parameter is close to zero, it is determined as a critical phase transition evolution pattern, and a corresponding critical phase transition evolution pattern identifier is generated.

[0167] Step S215: When the lesion evolution dynamics mode category identifier is the critical phase transition evolution mode identifier, extract the change gradient parameter sequence of chromatin edge continuous measurement between adjacent microscopic imaging image units in the chromatin edge continuous measurement sequence, perform detrending fluctuation analysis on the change gradient parameter sequence, and calculate the Hearst exponent parameter of the change gradient parameter sequence. The Hearst exponent parameter is used to characterize the long-term memory characteristics and trend persistence of the change gradient parameter sequence.

[0168] When the lesion evolution dynamics mode category identifier output in step S214 is the critical phase transition evolution mode identifier, the change gradient parameter sequence calculated in step S132 is obtained. Detrending fluctuation analysis is performed on this change gradient parameter sequence: the change gradient parameter sequence is divided into multiple windows of equal length, and the local trend is fitted using the least squares method within each window. After removing the trend, the fluctuation function of the remaining sequence is calculated. In a double logarithmic coordinate system, a curve of the fluctuation function changing with the window length is plotted, and the slope of the curve is calculated; this slope is the Hearst exponent parameter. A Hearst exponent parameter greater than 0.5 indicates that the sequence has long-term memory and trend persistence, while a value less than 0.5 indicates anti-persistence.

[0169] Step S216: Based on the deviation of the value of the Hearst exponent parameter from the preset critical Hearst threshold, a critical phase transition warning sensitivity parameter is generated. The critical phase transition warning sensitivity parameter reflects the degree to which the tissue lesion area represented by the set of digital images of the pathological slides approaches the critical state of qualitative change.

[0170] A critical Hearst threshold is preset, typically set to 0.5. The difference between the Hearst exponent parameter and this critical Hearst threshold is calculated, and the absolute value of the difference is taken as the degree of deviation. The degree of deviation is mapped to a range of 0 to 1 to generate a critical phase transition early warning sensitivity parameter. The larger the degree of deviation, the further the system is from the critical state; the smaller the degree of deviation, the closer the system is to the critical phase transition state. For example, the closer the Hearst exponent parameter is to 0.5, the closer the critical phase transition early warning sensitivity parameter is to 0, indicating that the lesion area of ​​the tissue is closer to the critical state of qualitative change.

[0171] Step S217: Combine the correlation dimension parameter, the maximum Lyapunov index parameter, the lesion evolution dynamics pattern category identifier, and the critical phase transition early warning sensitivity parameter into a lesion evolution dynamics feature descriptor, and encode the lesion evolution dynamics feature descriptor into a structured data block and append it to the extended data area of ​​the pathology quality control review report instruction.

[0172] The correlation dimension parameter calculated in step S212, the maximum Lyapunov exponent parameter calculated in step S213, the lesion evolution dynamics pattern category identifier determined in step S214, and the critical phase transition early warning sensitivity parameter generated in step S216 are combined into a lesion evolution dynamics feature descriptor. This descriptor is encoded into a structured data block according to the medical data exchange standard format, for example, encoded as an extended entry in the medical data exchange standard format. This data block is appended to the extended data area of ​​the pathology quality control review report instruction, so that the remote pathology diagnosis terminal can not only obtain static quality control review results, but also obtain lesion evolution trend prediction reference data based on nonlinear dynamics analysis, assisting pathologists in judging the progression stage and future evolution direction of the lesion.

[0173] In one exemplary embodiment, an online quality control and review system for remote pathology diagnosis is provided. This system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2As shown, the online quality control and auditing system for remote pathology diagnosis includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements an online quality control and auditing method for remote pathology diagnosis. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the shell of an online quality control and auditing system for remote pathological diagnosis, or an external keyboard, touchpad, or mouse, etc.

[0174] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. An online quality control review method for remote pathological diagnosis, characterized in that, The method includes: Receive a set of digital images of pathological slides uploaded by a remote pathological diagnosis terminal. The set of digital images of pathological slides includes microscopic imaging image units with depth-of-field markers that are continuously acquired under multiple fields of view. The pathological morphological features of the digital image set of the pathological slides are analyzed to obtain the morphological distribution features of cell nuclei and the topological features of cytoplasmic texture in the microscopic imaging image unit. The morphological distribution features of cell nuclei include nucleolar position offset parameters and chromatin edge continuity measures. The topological features of cytoplasmic texture include cytoplasmic fiber orientation consistency parameters and cytoplasmic vacuolar spatial distribution measures. Based on the morphological distribution characteristics of the cell nuclei and the topological structure characteristics of the cytoplasm texture, a cell atypia evolution feature chain based on the continuous measurement of chromatin edges is constructed, and a quality control review mark set for the set of digital images of the pathological slides is generated based on the cell atypia evolution feature chain. Based on the quality control audit mark set, determine the lesion region boundary contour chain and chromatin feature offset vector path of the pathological slide digital image set; The boundary contour chain of the lesion area and the chromatin feature offset vector path are loaded into a preset pathological diagnosis quality control report template to generate a pathological quality control review report instruction containing the mapping relationship between the cell atypia evolution feature chain and the boundary contour chain of the lesion area.

2. The method for online quality control review for telepathology according to claim 1, wherein, The set of digital images of pathological slides uploaded by the remote pathological diagnosis terminal includes: Receive the original scan data stream of pathological slides uploaded by the remote pathological diagnosis terminal through the medical digital imaging transmission protocol. The original scan data stream of pathological slides includes panoramic scan data of slides continuously acquired under different objective magnifications and with a unified spatial reference system. The original scan data stream of the pathological slide is parsed to extract the depth level marker field and field of view position index field from the packet header information. The original scan data stream of the pathological slide is then segmented according to the depth level marker field and field of view position index field to obtain a set of original image data segments with depth level attributes. For each original image data segment in the set of original image data segments, the spatial hierarchy order of the original image data segment in the direction of the vertical optical axis is determined according to the value of its corresponding depth level marker field, and the original image data segments are stacked according to the spatial hierarchy order to generate a multi-layer image slice data stack with the depth level as the vertical dimension. Obtain the image pixel matrix of each layer of the multi-layer image slice data stack, and perform color space conversion processing on the image pixel matrix to convert the pixel values ​​under the original sensor color space into the color channel component values ​​under the standard pathological color space that is independent of the display device. Based on the converted color channel component values, pixel-level registration is performed on image slices between adjacent depth layers in the multi-layer image slice data stack. By calculating the color gradient difference value at the same pixel position between adjacent layer slices, a displacement correction field is generated to characterize the degree of pixel offset between depth layers. The displacement correction field is used to perform spatial position remapping processing on the image pixel matrix of each layer of the multi-layer image slice data stack to eliminate the image misalignment and offset between layers caused by the micro-vibration of the microscope stage, and obtain the multi-layer image slice data stack with spatial position alignment. Image fusion preprocessing is performed on each layer of the multi-layer image slice data stack after spatial alignment. Based on the rate of change of color channel component values ​​of each pixel in adjacent depth layers, pixels that meet the preset depth focus evaluation conditions are extracted as the focus clear pixels of the corresponding layer, and a focus clear pixel position mask is generated. The focus-clear pixel location mask of each layer is combined with the color channel component values ​​of the slice of that layer in a pixel-by-pixel weighted combination. Based on the weight coefficient marked in the focus-clear pixel location mask, the multi-layer image slice data stack is subjected to depth-of-field expansion synthesis processing to obtain a microscopic imaging image unit with full field-of-field depth of field clarity. The field of view location index field is attached as a spatial location label to the corresponding microscopic imaging image unit, and the microscopic imaging image units with continuous field of view location index fields are combined into a pathological slide digital image set. Each microscopic imaging image unit in the pathological slide digital image set carries a corresponding depth level mark and field of view location index field.

3. The method for online quality control review in telepathology according to claim 1, wherein, The pathological morphological feature analysis processing of the digital image set of the pathological slides yields the cell nucleus morphological distribution features and cytoplasmic texture topological structure features in the microscopic imaging image units, including: The microscopic imaging image units in the pathological slide digital image set are loaded into the pathological morphological feature analysis processor. Color deconvolution separation processing is performed on the microscopic imaging image units. Based on the preset hematoxylin staining spectral absorption characteristic curve and eosin staining spectral absorption characteristic curve, the pixel color information of the microscopic imaging image units is demixed into hematoxylin staining concentration distribution channels and eosin staining concentration distribution channels. In the hematoxylin staining concentration distribution channel, cell nucleus candidate region localization processing based on morphological top cap reconstruction is performed. By constructing a disk-shaped structural element with radius parameter adaptation, grayscale opening operation is performed on the hematoxylin staining concentration distribution channel to suppress background texture components with a size smaller than the radius parameter of the disk-shaped structural element, and a cell nucleus foreground distribution map after background suppression is obtained. In the cell nucleus foreground distribution map, watershed segmentation is performed based on the concentration gradient magnitude and gradient direction of the pixels. Boundary separation is performed on densely populated regions of cell nuclei that are stuck together. Segmentation boundary ridges between cell nucleus regions are generated by tracing the local maximum path of the concentration gradient magnitude, resulting in a set of cell nucleus segmentation regions enclosed by independent cell nucleus contour lines. For each cell nucleus segmentation region in the set of cell nucleus segmentation regions, the geometric morphology description parameters of the cell nucleus segmentation region are extracted. The geometric morphology description parameters include the length of the major axis of the equivalent ellipse of the cell nucleus contour, the length of the minor axis of the equivalent ellipse, and the eccentricity parameter of the equivalent ellipse. Extract the concentration distribution values ​​of hematoxylin staining within each segmented region of the cell nucleus, calculate the mean and local variation range of the concentration distribution values ​​within the segmented region of the cell nucleus, and mark the local variation range of the concentration distribution values ​​as the chromatin edge continuity measure of the segmented region of the cell nucleus. In each segmented region of the cell nucleus, locate the local maximum point of the concentration distribution of hematoxylin staining agent, mark the local maximum point as the nucleolar center position point, and calculate the Euclidean distance offset of the nucleolar center position point relative to the geometric center point of the segmented region of the cell nucleus, and mark the Euclidean distance offset as the nucleolar position offset parameter. In the eosin staining concentration distribution channel, cytoplasmic texture primitive extraction processing based on the local binary mode operator is performed. The eosin staining concentration distribution channel is traversed and scanned with a preset circular neighborhood radius parameter and a neighborhood sampling point number parameter to generate a cytoplasmic texture primitive encoding map that reflects the local texture arrangement pattern of the pixels. The uniformity and consistency measure of the local texture direction is calculated based on the cytoplasmic texture primitive encoding map. By counting the number of bitwise jumps in the texture primitive encoding within the circular neighborhood, the region with the number of jumps below the preset jump number threshold is marked as a cytoplasmic fiber direction consistent region, and the consistency measure value of the cytoplasmic fiber direction consistent region is marked as the cytoplasmic fiber direction consistency parameter. In the eosin staining concentration distribution channel, connected regions with gray values ​​lower than the preset cytoplasmic vacuoles gray value threshold are extracted. The area parameter and roundness parameter of the connected regions are calculated, and the connected regions that meet the area parameter and roundness parameter constraints are marked as cytoplasmic vacuoles. The number of cytoplasmic vacuoles per unit area and the degree of spatial distribution dispersion are counted as the spatial distribution measure of cytoplasmic vacuoles. The nucleolar position offset parameter of all cell nucleus segmentation regions within the same microscopic imaging unit is combined with the chromatin edge continuity measure to form the cell nucleus morphology distribution feature, and the cytoplasmic fiber orientation consistency parameter and the cytoplasmic vacuole spatial distribution measure within the same microscopic imaging unit are combined to form the cytoplasmic texture topology feature.

4. The online quality control and auditing method for remote pathological diagnosis according to claim 1, characterized in that, The process involves constructing a cell atypia evolution feature chain based on chromatin edge continuity measurement according to the cell nuclear morphology distribution characteristics and the cytoplasmic texture topology characteristics, and generating a quality control and audit mark set for the set of digital images of the pathological slides based on the cell atypia evolution feature chain, including: Extract the field of view position index field corresponding to the microscopic imaging image unit, sort all microscopic imaging image units in the pathological slide digital image set according to the spatial continuity relationship of the field of view position index field, and obtain a spatial sequence of microscopic imaging image units along the preset tissue section extension direction. Extract the chromatin edge continuity measure of the cell nucleus morphology distribution features of each microscopic imaging image unit in the spatial sequence of microscopic imaging image units, and use the chromatin edge continuity measure as the feature node value of the microscopic imaging image unit in the cell atypia evolution feature chain. Calculate the gradient parameter of the change in chromatin edge continuity measure between adjacent microscopic imaging image units in the spatial sequence of microscopic imaging image units. The gradient parameter is the ratio of the absolute value of the difference between the chromatin edge continuity measure of the subsequent microscopic imaging image unit and the chromatin edge continuity measure of the preceding microscopic imaging image unit to the spatial distance parameter between adjacent microscopic imaging image units. Based on the positive and negative polarity of the changing gradient parameter, the evolution direction identifier between adjacent feature nodes in the cell atypia evolution feature chain is determined. When the chromatin edge continuity measure of the subsequent microscopic imaging image unit is greater than that of the preceding microscopic imaging image unit, the evolution direction identifier is set to an increasing identifier; otherwise, the evolution direction identifier is set to a decreasing identifier. The spatial distribution measure of cytoplasmic vacuoles in the cytoplasmic texture topology features of each microscopic imaging image unit in the spatial sequence of microscopic imaging image units is obtained. The distribution density change rate of the spatial distribution measure of cytoplasmic vacuoles between adjacent microscopic imaging image units is calculated, and the distribution density change rate is used as an auxiliary evolution constraint parameter of the feature nodes in the cell heterogeneity evolution feature chain. Based on the evolution direction identifier and the auxiliary evolution constraint parameter, a node connection relationship of the cell atypia evolution feature chain is constructed. When the evolution direction identifier is an increasing identifier and the auxiliary evolution constraint parameter is greater than the preset evolution constraint threshold, a strengthening connection edge is established between adjacent feature nodes. When the evolution direction identifier is a decreasing identifier or the auxiliary evolution constraint parameter is less than or equal to the preset evolution constraint threshold, a weakening connection edge is established between adjacent feature nodes. Based on the distribution range of chromatin edge continuity measurement values ​​of all feature nodes in the cell atypia evolution feature chain, the atypia evolution amplitude parameter of the cell atypia evolution feature chain is determined. The atypia evolution amplitude parameter is the difference between the maximum and minimum chromatin edge continuity measurement in the cell atypia evolution feature chain. Based on the number of consecutive identical evolution direction identifiers between adjacent feature nodes in the cell atypia evolution feature chain, the atypia evolution trend consistency parameter of the cell atypia evolution feature chain is determined. The atypia evolution trend consistency parameter is the maximum sequence length value of the evolution direction identifiers that continuously maintain an increasing or decreasing identifier. The parameter of the magnitude of the anomalous evolution and the parameter of the consistency of the anomalous evolution trend are mapped to a preset quality control audit mark mapping table to generate a quality control audit mark set corresponding to the set of digital images of the pathological slides. The quality control audit mark set includes mark symbols for indicating the slide quality level and mark symbols for indicating the risk tendency of the lesion.

5. The method for online quality control auditing of telepathological diagnoses according to claim 1, wherein, The step of determining the lesion region boundary contour chain and chromatin feature offset vector path of the pathological slide digital image set based on the quality control review mark set includes: Extract the marker symbols used to indicate the risk tendency of lesions from the quality control audit mark set, and determine the range of lesion risk areas that need to be processed by boundary tracking in the digital image set of pathological slides based on the preset risk tendency level value corresponding to the marker symbols; In the spatial sequence of microscopic imaging image units in the digital image set of the pathological slide, locate the microscopic imaging image unit corresponding to the feature node in the cell atypia evolution feature chain where the continuous measurement of chromatin edge exceeds the preset atypia determination threshold, and mark the located microscopic imaging image unit as the lesion core region unit. The hematoxylin staining concentration distribution channel of the core lesion region unit is obtained. Boundary evolution tracking processing based on active contour model is performed in the hematoxylin staining concentration distribution channel. The cell nucleus segmentation region with the highest chromatin edge continuity metric in the core lesion region unit is used as the initial contour line. The initial contour line is driven to iteratively deform and evolve towards the true boundary of the lesion region according to the hematoxylin staining concentration gradient field to obtain the boundary contour of the core lesion region. The boundary contour of the lesion core region is spatially propagated to adjacent microscopic imaging image units along the sequence direction of the spatial sequence of the microscopic imaging image units. The boundary contour of the lesion core region determined in the preceding microscopic imaging image unit is used as the initial constraint contour line for boundary tracking of the adjacent subsequent microscopic imaging image units. The boundary evolution direction in the adjacent subsequent microscopic imaging image units is guided by the constraint contour line to obtain the boundary contour of the lesion extension region of the adjacent subsequent microscopic imaging image units. All microscopic imaging image units containing the boundary contour of the lesion core region or the boundary contour of the lesion extension region in the spatial sequence of microscopic imaging image units are connected according to spatial adjacency. The set of boundary contour coordinate points corresponding to each microscopic imaging image unit is extracted, and the set of boundary contour coordinate points between adjacent microscopic imaging image units is spliced ​​end to end to generate a lesion region boundary contour chain that extends continuously along the extension direction of the tissue section. The nucleolar position offset parameter of each feature node in the cell atypia evolution feature chain is extracted, and the nucleolar position offset parameter is converted into a two-dimensional plane offset vector with the geometric center point of the cell nucleus as the starting point. The direction of the two-dimensional plane offset vector is determined by the azimuth angle of the nucleolar center position point relative to the geometric center point of the cell nucleus, and the magnitude of the two-dimensional plane offset vector is determined by the Euclidean distance offset. Based on the spatial positional relationship between adjacent microscopic imaging image units in the spatial sequence of microscopic imaging image units, the two-dimensional planar offset vectors of all cell nucleus regions within each microscopic imaging image unit are subjected to spatial interpolation fitting to generate an offset vector field distribution map covering the entire microscopic imaging image unit region. The regions in the offset vector field distribution map whose offset vector magnitude is greater than a preset offset significance threshold are extracted as chromatin feature offset significant regions. The main angular distribution direction of the offset vector direction within the chromatin feature offset significant regions is extracted and taken as the dominant angular distribution direction of the chromatin feature offset within the microscopic imaging image unit. The dominant direction of chromatin feature offset of each microscopic imaging image unit in the spatial sequence of microscopic imaging image units is connected according to the spatial adjacency order of the microscopic imaging image units to generate a chromatin feature offset vector path with spatial position as the horizontal axis and offset direction angle as the vertical axis.

6. The method for online quality control auditing of telepathological diagnoses according to claim 1, characterized in that, The step of loading the boundary contour chain of the lesion region and the chromatin feature offset vector path into a preset pathological diagnosis quality control report template to generate a pathological quality control review report instruction containing the mapping relationship between the cell atypia evolution feature chain and the boundary contour chain of the lesion region includes: Obtain a preset pathological diagnosis quality control report template, which includes a first visualization layer for carrying image annotation information, a second visualization layer for carrying an illustration of the evolutionary feature chain, and a text description area for carrying diagnostic conclusion description text. The boundary contour coordinate point set in the boundary contour chain of the lesion area is mapped to the pixel coordinate system of the first visualization layer. A closed polygon vector path is generated according to the connection relationship between adjacent coordinate points in the boundary contour coordinate point set. The closed polygon vector path is rendered and drawn using the preset boundary line style and boundary line color parameters of the lesion area. Extract the spatial position index field of the microscopic imaging image unit corresponding to each set of boundary contour coordinate points in the boundary contour chain of the lesion area, generate a position index identifier at the corresponding coordinate position in the first visualization layer based on the spatial position index field, and establish a visualization association relationship between the position index identifier and the closed polygon vector path; A mapping relationship is established between the offset direction angle sequence in the chromatin feature offset vector path and the spatial position index field of the microscopic imaging image unit. In the second visualization layer, the offset vector path curve is drawn with the spatial position index field as the horizontal axis and the offset direction angle as the vertical axis. The offset vector path curve is drawn using a preset vector path line style and vector path color parameters. In the second visualization layer, a diagram of the node connection relationship of the cell atypia evolution feature chain is drawn. A scatter plot of feature nodes is drawn with the spatial position index field of the microscopic imaging unit as the horizontal axis coordinate and the chromatin edge continuity measure as the vertical axis coordinate. Connecting line segments between nodes are drawn according to the connection relationship between the strengthening connection edge and the weakening connection edge in the cell atypia evolution feature chain. Extract the atypia evolution amplitude parameter and the atypia evolution trend consistency parameter of the cell atypia evolution feature chain, and convert the atypia evolution amplitude parameter and the atypia evolution trend consistency parameter into a qualitative descriptive text fragment that conforms to the expression standard of pathological diagnosis report. The qualitative descriptive text fragment contains level words for describing the degree of cell atypia and trend words for describing the stability of evolution trend. Based on the marking symbols used to indicate the production quality level in the quality control audit mark set, retrieve the production quality evaluation description text fragment corresponding to the marking symbol from the preset production quality evaluation text library, and fill the production quality evaluation description text fragment into the first text paragraph position of the text description area; Based on the spatial extension length parameter of the boundary contour chain of the lesion area and the average area parameter of the area enclosed by the boundary contour, a quantitative descriptive text fragment for describing the lesion range and lesion degree is generated, and the quantitative descriptive text fragment is filled into the second text paragraph position of the text description area; The first and second visualization layers, which have been rendered, are overlaid and merged. The merged visualization layer is then combined with the filled text description area to form a pathology quality control review report page. The pathology quality control review report page is then encoded into a pathology quality control review report instruction that conforms to the transmission format requirements of the medical information system.

7. The method for online quality control auditing of telepathological diagnoses according to claim 1, characterized in that, The method further includes: The chromatin edge continuity measure of all feature nodes in the cell atypia evolution feature chain is obtained, and according to the spatial continuity relationship of the field of view position index field corresponding to the microscopic imaging image unit, the chromatin edge continuity measure is arranged into a one-dimensional feature measure sequence along the extension direction of the tissue section. The one-dimensional feature measure sequence is subjected to local trend decomposition processing based on sliding window. The window length parameter and the window sliding step parameter are set. In each sliding window, the local linear fitting slope parameter and the local fitting residual fluctuation amplitude parameter of the chromatin edge continuity measure are calculated. Based on the positive and negative polarity change characteristics of the local linear fitting slope parameter, the one-dimensional feature measurement sequence is divided into an upward trend segment, a stable fluctuation segment, and a downward trend segment, and a segment trend type identifier is assigned to each segment. Extract the sequence values ​​of the local fitting residual fluctuation amplitude parameter within each stationary fluctuation segment, calculate the sequence autocorrelation function of the local fitting residual fluctuation amplitude parameter, and determine whether there is a periodic fluctuation pattern in the chromatin edge continuity measurement within the stationary fluctuation segment based on the decay characteristics of the sequence autocorrelation function. When a periodic fluctuation pattern exists within the stationary fluctuation segment, calculate the fluctuation period length parameter and fluctuation amplitude parameter of the periodic fluctuation pattern, and use the fluctuation period length parameter and fluctuation amplitude parameter as the basis for evaluating the slide quality stability of the microscopic imaging unit region corresponding to the stationary fluctuation segment. Based on the magnitude of the slope parameter of the local linear fitting within the upward trend segment and the segment duration parameter, the degree of cell atypia aggravation corresponding to the upward trend segment is determined, and a correlation is established between the degree of cell atypia aggravation and the index range of the visual field location covered by the upward trend segment. The fragment trend type identifier, the degree of cell atypia aggravation, and the evaluation criteria for slide quality stability are loaded into the supplementary analysis and description area of ​​the pathological diagnosis quality control report template to generate an analysis and description text paragraph that describes the stage characteristics of lesion evolution and the characteristics of slide quality stability.

8. The method for online quality control auditing of telepathological diagnoses according to claim 1, characterized in that, The method further includes: Extract the perimeter parameter and the area parameter of the region enclosed by the contour of each core region of the lesion in the boundary contour chain of the lesion region. Calculate the morphological complexity parameter of the boundary contour of the core region of the lesion based on the perimeter parameter and the area parameter of the region enclosed by the contour. The morphological complexity parameter is the ratio of the square of the perimeter parameter to the area parameter of the region enclosed by the contour. Extract the perimeter parameter and the area parameter of the region enclosed by the contour of each lesion expansion region in the lesion region boundary contour chain, and calculate the area expansion rate parameter of the lesion expansion region based on the change in the area parameter of the region enclosed by the contour of the lesion expansion region between adjacent microscopic imaging image units. A two-dimensional feature vector is constructed based on the morphological complexity parameter and the area expansion rate parameter to describe the growth behavior characteristics of the lesion region. The two-dimensional feature vector is mapped to a preset lesion growth pattern classification space to determine the lesion growth pattern category identifier corresponding to the boundary contour chain of the lesion region. Extract the dominant direction angle of chromatin feature offset for each microscopic imaging unit in the chromatin feature offset vector path, calculate the change in angular deflection of the dominant direction angle between adjacent microscopic imaging units, and determine the overall deflection direction trend of chromatin feature offset based on the cumulative value of the change in angular deflection. The gradient parameter of the change in chromatin edge continuity between adjacent feature nodes in the cell atypia evolution feature chain is obtained. The sequence value of the gradient parameter is extracted, and the mean and coefficient of variation of the gradient parameter sequence value are calculated. Based on the mean and coefficient of variation of the gradient parameter sequence value, the degree of evolution severity of the cell atypia evolution feature chain is determined. The degree of evolution severity, the lesion growth pattern category identifier, and the overall deflection direction trend are combined to generate a comprehensive lesion behavior feature descriptor. The comprehensive behavioral characteristic descriptor of the lesion is encoded into a structured supplementary diagnostic information data segment, and the supplementary diagnostic information data segment is attached to the extended data area of ​​the pathology quality control review report instruction for reference by the remote pathology diagnostic terminal.

9. The method for online quality control auditing of telepathological diagnoses according to claim 1, characterized in that, The method further includes: Obtain the overlapping boundary region between microscopic imaging units corresponding to adjacent field of view position index fields in the digital image set of pathological slides, and extract the pixel value distribution of the hematoxylin staining concentration distribution channel and the pixel value distribution of the eosin staining concentration distribution channel within the overlapping boundary region; The consistency measure of concentration distribution between adjacent microscopic imaging units is calculated by comparing the differences in hematoxylin concentration values ​​at the same pixel position within the overlapping region of adjacent microscopic imaging units. The texture structure consistency measure of the pixel values ​​of the eosin staining concentration distribution channel in the overlapping boundary region is calculated between adjacent microscopic imaging image units. The texture structure consistency measure is obtained by comparing the similarity of the cytoplasmic texture primitive encoding map in the overlapping region of adjacent microscopic imaging image units. Based on the concentration distribution consistency measure and the texture structure consistency measure, an imaging quality continuity parameter for the overlapping boundary region is constructed, and the imaging quality continuity parameter is compared with a preset imaging quality continuity evaluation threshold to generate an evaluation result for evaluating the image stitching quality during the pathological slide scanning process. Extract the chromatin edge continuity measure from the cell nucleus morphology distribution characteristics within the overlapping boundary region, calculate the degree of difference in the chromatin edge continuity measure within the overlapping boundary region between adjacent microscopic imaging units, and use the degree of difference as an auxiliary reference for evaluating the staining uniformity during the preparation of pathological slides. Based on the difference between the imaging quality continuity parameter and the chromatin edge continuity measure, a digital acquisition quality assessment data segment for pathological slides is generated. The digital acquisition quality assessment data segment for pathological slides includes an image stitching quality level identifier and a staining uniformity level identifier. The digital acquisition quality assessment data segment of the pathological slides is embedded into the report header information area of ​​the pathology quality control review report instruction.

10. An online quality control and review system for remote pathological diagnosis, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the online quality control audit method for remote pathological diagnosis as described in any one of claims 1 to 9 by executing the machine-executable instructions.

Images