CN-121982304-A - Automatic processing method of thyroid tissue digital pathological image

CN121982304ACN 121982304 ACN121982304 ACN 121982304ACN-121982304-A

Abstract

The invention discloses an automatic processing method of thyroid tissue digital pathology images, which comprises the steps of obtaining thyroid cancer digital pathology images and LAB images, dividing the LAB images into a plurality of areas in the LAB images corresponding to a plurality of superpixel blocks, obtaining gradient characteristic values of each pixel point, obtaining position parameters according to gradient directions of gradient characteristic values of edge points in each area, determining enclosed bubble areas and cell nucleus areas according to the position parameters of each area, fusing the enclosed areas with the corresponding enclosed areas to obtain initial target areas, determining initial target superpixel blocks according to the gradient characteristic values, fusing the initial target superpixel blocks with color distances between adjacent superpixel blocks to obtain target superpixel blocks, and marking the areas corresponding to the target superpixel blocks of the LAB images in the pathology images as target areas.

Inventors

QIAN CHEN
YANG CHENGGUANG

Assignees

上海市同仁医院

Dates

Publication Date: 20260505
Application Date: 20260108

Claims (10)

1. An automatic processing method of thyroid tissue digital pathological images is characterized in that: acquiring a thyroid cancer digital pathological image and an LAB image of the pathological image, and dividing the LAB image into a plurality of super-pixel blocks by utilizing super-pixel segmentation, wherein each super-pixel block corresponds to one region in the LAB image; Determining a gradient characteristic value of each pixel point according to the distribution difference of the channel values of the pixel points in different super pixel blocks on the same channel and the gradient value of the channel values of the pixel points on each channel; Determining the position parameter of each region according to the gradient direction of the gradient characteristic value of the edge point in each region; Determining a surrounded bubble area, a cell nucleus area, a corresponding cytoplasm area and a corresponding tissue fluid area according to the position parameters of each area, and respectively fusing the bubble area and the cell nucleus area in the image with the corresponding tissue fluid area and the corresponding cytoplasm area to obtain a plurality of initial target areas; Determining an initial target super-pixel block according to the gradient characteristic value of the super-pixel block corresponding to each initial target area; Fusing the initial target super-pixel blocks and the adjacent super-pixel blocks according to the color distance between the initial target super-pixel blocks and the adjacent super-pixel blocks according to the channel values of the central points of each initial target super-pixel block and the adjacent super-pixel blocks and the gradient feature values to obtain target super-pixel blocks; and taking the corresponding region of the target super-pixel block in the thyroid cancer digital pathological image as a target region.
2. The method of claim 1, wherein the step of dividing the LAB image into a plurality of super-pixel blocks using super-pixel segmentation comprises: graying treatment is carried out on the thyroid cancer digital pathological image to obtain a gray image, and the number of super-pixel blocks is determined according to the clustering result of the gray values of the pixel points in the gray image; the LAB image is segmented into a plurality of super-pixel blocks using a super-pixel segmentation algorithm based on the number of super-pixel blocks.
3. The automatic processing method of thyroid tissue digital pathological image according to claim 2, wherein the method for determining the number of super pixel blocks is as follows: and taking the gray values of all the pixel points in the gray image as the input of an AP clustering algorithm, dividing all the pixel points in the gray image into different clusters by using the AP clustering algorithm, and counting the number of the clusters as the number of super-pixel blocks, wherein the measurement distance in clustering is the difference value between the gray values of the two pixel points.
4. The automatic processing method of thyroid tissue digital pathological image according to claim 1, wherein the method for determining the gradient eigenvalue of each pixel point is: determining a weighting weight of each channel based on the distribution similarity of pixel point channel values among different super pixel blocks; for any channel in the LAB image, sorting the channel values of all the pixel points on each channel according to the positions of the pixel points to obtain a channel value matrix of each channel, and obtaining the gradient value of the channel value of each pixel point by using a Sobel operator; and linearly weighting the gradient value of the channel value of each pixel point on each channel based on the weighting weight of each channel, wherein the linear weighting result is the gradient characteristic value of each pixel point.
5. The method for automatically processing digital pathological images of thyroid tissue according to claim 4, wherein the method for determining the weighting weight of each channel is as follows: counting the distribution histograms of the channel values of all pixel points in each super pixel block on each channel, and calculating the Pasteur distance between the distribution histograms of the channel values of any two different super pixel blocks on the same channel; Taking the normalization result of each Pasteur distance as the distribution distance of two corresponding super pixel blocks on the same channel; And taking the average value of the accumulated results of the ratio of the distribution distance of the two super pixel blocks on each channel to the sum of the distribution distances of the two super pixel blocks on the three channels as the weighting weight of each channel.
6. The automatic processing method of thyroid tissue digital pathological image according to claim 1, wherein the method for determining the position parameter of each region is as follows: setting the sign function value of the edge pixel points, of which the gradient directions point to the outside of each region, of the gradient characteristic values in each region to be 1; Setting the sign function value of the edge pixel point, which points the gradient direction of the gradient characteristic value in each region to the inside of each region, as-1; And taking the average value of the sign function values of all the edge pixel points in each region as the position parameter of each region.
7. The method of claim 1, wherein the step of determining the enclosed bubble and nucleus regions and the corresponding cytoplasmic and interstitial fluid regions comprises: If the position parameter of the region is 1, the region is taken as a surrounded bubble region or a cell nucleus region; the area adjacent to the bubble area is a tissue fluid area; The region adjacent to the nuclear region is the cytoplasmic region.
8. The method of claim 1, wherein the step of obtaining a plurality of initial target areas comprises: the bubble area is surrounded by a plurality of tissue fluid areas, and any one of the tissue fluid areas is selected to be fused with the corresponding bubble area; The cell nucleus region is surrounded by a plurality of cytoplasmic regions, and any one of the cytoplasmic regions is selected to be fused with the corresponding cell nucleus region.
9. The automatic processing method of thyroid tissue digital pathological image according to claim 1, wherein the method for obtaining the color distance between the initial target superpixel block and the adjacent superpixel block is as follows: And calculating the square value of the difference value between the central point of the initial target super pixel block and the channel value of the central point of the adjacent super pixel block on each channel, and taking the sum of the accumulated result of the square value of the difference value on three channels and the square value of the difference value between the central point of the initial target super pixel block and the gradient characteristic value of the central point of the adjacent super pixel block as the color distance between the initial target super pixel block and the adjacent super pixel block.
10. The automatic processing method of thyroid tissue digital pathological image according to claim 1, wherein the method for fusing the initial target superpixel block and the adjacent superpixel block according to the color distance to obtain the target superpixel block is as follows: counting the color distances between all initial target super-pixel blocks and adjacent super-pixel blocks, and acquiring the segmentation threshold values of all the color distances as a color distance threshold value by using an Ojin threshold algorithm; And when the color distance is smaller than the color distance threshold, fusing the initial target super-pixel block with the adjacent super-pixel block, and marking the fused super-pixel block as the target super-pixel block.

Description

Automatic processing method of thyroid tissue digital pathological image Technical Field The invention relates to the technical field of image segmentation, in particular to an automatic processing method of thyroid tissue digital pathological images. Background The incidence rate of thyroid cancer is gradually increased in the global scope, the ranking of thyroid cancer is gradually increased in malignant tumor harm, the influence brought by the thyroid cancer is larger and larger, the thyroid cancer can be subjected to pathological diagnosis according to thyroid cancer cells, digital pathological images in pathological diagnosis are usually formed by scanning and fusing pathological tissues through microscopic scanning equipment after operations such as dehydration embedding and slicing, unlike the traditional microscope for observing pathological tissues, digital case images are independent of an optical microscope, can be directly displayed and observed on any equipment, can be amplified at a local high magnification, and can help doctors to observe tissue cells more clearly. The digital pathological image is higher in resolution, the whole image is larger, the number of pixel points is large, a single pixel point does not have any medical characteristics, so that the digital pathological image can be divided into a plurality of super pixel blocks by utilizing super pixel division, the medical characteristics are displayed according to the super pixel blocks, however, in the digital pathological image, the color of air bubbles and tissue fluid in follicular cancer cells of thyroid cancer is different, the air bubbles and tissue fluid of the follicular cancer cells cannot be divided into the same super pixel block during super pixel division, the pathological characteristics cannot be displayed, so that a cancerous region is difficult to be identified by the super pixel division directly, the papillary cancer cells of thyroid cancer are not accurate and reliable due to the fact that the color of cell nuclei and cytoplasm parts of the papillary cancer cells of thyroid cancer is different and are irregular, and the dividing effect on the digital pathological image of thyroid cancer is poor. Disclosure of Invention The invention provides an automatic processing method of thyroid tissue digital pathological images, which aims to solve the existing problems. The invention relates to an automatic processing method of thyroid tissue digital pathological images, which adopts the following technical scheme: acquiring a thyroid cancer digital pathological image and an LAB image of the pathological image, and dividing the LAB image into a plurality of super-pixel blocks by utilizing super-pixel segmentation, wherein each super-pixel block corresponds to one region in the LAB image; Determining a gradient characteristic value of each pixel point according to the distribution difference of the channel values of the pixel points in different super pixel blocks on the same channel and the gradient value of the channel values of the pixel points on each channel; Determining the position parameter of each region according to the gradient direction of the gradient characteristic value of the edge point in each region; Determining a surrounded bubble area, a cell nucleus area, a corresponding cytoplasm area and a corresponding tissue fluid area according to the position parameters of each area, and respectively fusing the bubble area and the cell nucleus area in the image with the corresponding tissue fluid area and the corresponding cytoplasm area to obtain a plurality of initial target areas; Determining an initial target super-pixel block according to the gradient characteristic value of the super-pixel block corresponding to each initial target area; Fusing the initial target super-pixel blocks and the adjacent super-pixel blocks according to the color distance between the initial target super-pixel blocks and the adjacent super-pixel blocks according to the channel values of the central points of each initial target super-pixel block and the adjacent super-pixel blocks and the gradient feature values to obtain target super-pixel blocks; and taking the corresponding region of the target super-pixel block in the thyroid cancer digital pathological image as a target region. Preferably, the step of dividing the LAB image into a plurality of super-pixel blocks using super-pixel division comprises: graying treatment is carried out on the thyroid cancer digital pathological image to obtain a gray image, and the number of super-pixel blocks is determined according to the clustering result of the gray values of the pixel points in the gray image; the LAB image is segmented into a plurality of super-pixel blocks using a super-pixel segmentation algorithm based on the number of super-pixel blocks. Preferably, the method for determining the number of super pixel blocks is as follows: and taking the gray values of all the pixel po