CN-121582926-B - Cell segmentation method, device, storage medium and program product based on space histology sequencing

CN121582926BCN 121582926 BCN121582926 BCN 121582926BCN-121582926-B

Abstract

The invention discloses a cell segmentation method, equipment, a storage medium and a program product based on space histology sequencing, wherein the method comprises the steps of obtaining an acquired full-slice image, dividing the full-slice image into a plurality of image blocks, identifying whether heterogeneous areas exist in the image blocks to obtain an identification result, preprocessing each image block according to the identification result to obtain preprocessed image data corresponding to each image block, forming input of a deep learning segmentation model based on preset processed image data corresponding to each image block, outputting segmentation results of each image block through the deep learning segmentation model, and obtaining each cell area in the full-slice image based on the segmentation results of each image block.

Inventors

XU ZAOXU
ZHAO TIANHAO
LIANG HAN
LIAO RENJIE
BAO YUANYE
WANG GUFENG
ZHAO LUYANG

Assignees

上海赛陆生命科学有限公司

Dates

Publication Date: 20260505
Application Date: 20260126

Claims (9)

1.A method of cell segmentation based on spatial histology sequencing, comprising: acquiring an acquired full-slice image, the full-slice image being indicative of a distribution characteristic of a spatial histology signal of a tissue slice; partitioning the full-slice image to obtain a plurality of image blocks, and identifying whether heterogeneous areas exist in the image blocks to obtain an identification result; determining preprocessing parameter data according to the identification result; Preprocessing each image block according to the preprocessing parameter data to obtain preprocessed image data corresponding to each image block, wherein the preprocessing parameter data comprises heterogeneous preprocessing parameters corresponding to each heterogeneous type, preprocessing each image block according to the preprocessing parameter data to obtain preprocessed image data corresponding to each image block, and preprocessing each image block respectively to obtain preprocessed image blocks corresponding to each group of heterogeneous preprocessing parameters, wherein the preprocessed image blocks corresponding to each group of heterogeneous preprocessing parameters are used as preprocessed image data corresponding to the image blocks; Forming input of a deep learning segmentation model based on preset processing image data corresponding to each image block, and outputting segmentation results of each image block through the deep learning segmentation model; And obtaining each cell area in the whole slice image based on the segmentation result of each image block.
2. The method of spatially-histologic sequencing-based cell segmentation of claim 1, wherein determining pretreatment parameter data based on the recognition result comprises: when the identification result indicates that a plurality of image blocks do not have heterogeneous areas, determining a group of uniform preprocessing parameters; and when the identification result indicates that one or more image blocks have heterogeneous areas, determining heterogeneous preprocessing parameters corresponding to each heterogeneous type in the identification result.
3. The method of claim 2, wherein determining heterogeneous pretreatment parameters corresponding to each heterogeneous type in the recognition result comprises: Screening representative image blocks of each heterogeneous type from a plurality of image blocks; for any representative image block, based on the current preprocessing parameters, gradually adjusting the current preprocessing parameters in the current iteration process to obtain preprocessing adjustment parameters; Dividing the representative image block based on the preprocessing adjustment parameters to obtain cell division numbers of the preprocessing adjustment parameters, taking the preprocessing adjustment parameters as the current preprocessing parameters, continuing iteration until iteration termination conditions are reached, and taking the preprocessing parameters corresponding to the maximum cell division numbers as heterogeneous preprocessing parameters of heterogeneous types corresponding to the representative image block.
4. The method of spatially sequencing-based cell segmentation of claim 1, wherein the obtaining of each cell region in the whole-slice image based on the segmentation result of each image block comprises: for any image block, the segmentation result of the image block comprises preprocessed image blocks corresponding to heterogeneous preprocessing parameters, each preprocessed image block corresponds to a segmentation layer, and image fusion is performed according to overlapping data of cell prediction outlines in each segmentation layer to obtain a target cell segmentation result of the image block; And obtaining each cell area in the whole slice image based on the target cell segmentation result of each image block.
5. The method of claim 4, wherein performing image fusion based on overlapping data of predicted contours of cells in each segmented layer to obtain a target cell segmentation result of the image block comprises: For any cell prediction contour in any segmentation layer, when the cell prediction contour is not overlapped with the cell prediction contour of a comparison layer, determining the cell prediction contour as a target cell contour; When the cell prediction contour is overlapped with the comparison cell contour of the comparison layer, calculating the mass center of the cell prediction contour and the mass center of the comparison cell contour, and determining the target cell contour based on the mass center of the cell prediction contour and the mass center of the comparison cell contour.
6. The method of spatially-histologic sequencing-based cell segmentation of claim 5, wherein the determining a target cell profile based on the centroid of the cell prediction profile and the centroid of the comparison cell profile comprises: When the centroid of the cell prediction profile is within the region of the comparative cell profile and the centroid of the comparative cell profile is within the region of the cell prediction profile, fusing the cell prediction profile and the comparative cell profile into a target cell profile; And when the centroid of the cell prediction outline is not in the area of the comparison cell outline and/or the centroid of the comparison cell outline is in the area of the cell prediction outline, creating a dividing line based on the intersection point of the cell prediction outline and the comparison cell outline, so as to obtain two target cell outlines.
7. A computing device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method of any of claims 1 to 6.
8. A computer readable storage medium storing a computer program, which when executed by a processor causes the processor to perform the steps of the method according to any one of claims 1 to 6.
9. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6.

Description

Cell segmentation method, device, storage medium and program product based on space histology sequencing Technical Field The invention relates to the technical field of genes, in particular to a cell segmentation method and system based on space histology sequencing, computing equipment, a computer readable storage medium and a computer program product. Background The space histology technique is to perform large-scale assays on molecules (e.g., RNA, proteins, chromatin accessibility, etc.) in tissues while preserving the original spatial positional information of the cells. Among them sequencing-based techniques, in particular spatial transcriptomics (Spatial Transcriptomics), are the mainstay and hot spot of current development. It skillfully combines high throughput sequencing (NGS) with spatial position coding. Cell segmentation (Cell Segmentation) refers to the process of dividing a cell image into several mutually non-overlapping regions according to features such as gray scale, color, texture, geometry, etc., such that the features exhibit similarity in the same region, but exhibit significant differences between different regions. This is the basis for many biomedical studies, whose accuracy is directly related to the reliability of almost all analyses, such as subsequent cell counting, morphological analysis, quantification of gene expression, etc. Common cell images are taken by microscopy, including fluorescence images and bright field images, such as fluorescent images of nuclear dye ssDNA staining and H & E bright field images of hematoxylin and eosin staining. The combination of cell segmentation and space histology classifies molecules in tissues into single cell areas, and is an important means for improving the resolution of space histology and the accuracy of data quantification. How to increase the recognition rate of cell segmentation is an important issue in the field. The existing cell segmentation technology is not suitable for images produced by different laboratories, the segmentation effect is limited by the influence of dyeing and coloring depth, picture shooting brightness and contrast, and the segmentation precision is low. Disclosure of Invention In order to solve the existing technical problems, the invention provides a cell segmentation method and system based on space histology sequencing, a computing device, a computer readable storage medium and a computer program product, which can improve the accuracy of cell segmentation. The method comprises the steps of obtaining collected full-slice images, dividing the full-slice images into a plurality of image blocks, identifying whether heterogeneous areas exist in the image blocks to obtain identification results, preprocessing the image blocks according to the identification results to obtain preprocessed image data corresponding to the image blocks, forming input of a deep learning segmentation model based on the preset processed image data corresponding to the image blocks, outputting segmentation results of the image blocks through the deep learning segmentation model, and obtaining cell areas in the full-slice images based on the segmentation results of the image blocks. In a second aspect, a computing device is provided, comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of a method for spatially histology sequencing-based cell segmentation provided by an embodiment of the present application. In a third aspect, a computer readable storage medium is provided, storing a computer program, which when executed by a processor, causes the processor to perform the steps of the method for cell segmentation based on spatial histology provided by embodiments of the present application. In a fourth aspect, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method for spatially sequencing-based cell segmentation provided by embodiments of the present application. The method comprises the steps of acquiring a collected full-slice image, blocking the full-slice image to obtain a plurality of image blocks, identifying whether heterogeneous areas exist in the image blocks to obtain an identification result, wherein the identification result comprises whether heterogeneous areas exist in the image blocks or not and heterogeneous types corresponding to the image blocks with the heterogeneous areas, searching pretreatment parameters corresponding to the heterogeneous types according to the heterogeneous types of the heterogeneous areas, carrying out pretreatment on each image block by utilizing the pretreatment parameters corresponding to the heterogeneous types to obtain pretreatment image data corresponding to each image block, forming the input of a deep learning segmentation model based on preset treatment image data corresponding to each image block, o