CN-122023471-A - Contour registration-based micro-cutting multi-slide ROI mapping method and device

CN122023471ACN 122023471 ACN122023471 ACN 122023471ACN-122023471-A

Abstract

The invention discloses a contour registration-based microscopic cutting multi-slide ROI mapping method and device, wherein the method comprises the steps of firstly obtaining a standard image of a reference slide, a region of interest thermodynamic diagram and a working image of a target slide, carrying out image correction on the working image, and respectively extracting a single standard contour of the standard image and one or more working contours of the corrected working image; and then, carrying out iterative closest point registration on each group of contours and the standard contours to obtain optimal mapping parameters, transforming thermodynamic diagrams according to the parameters and combining to generate a complete target area thermodynamic diagram of the working diagram. The invention adapts to dyeing difference scenes, supports multi-tissue drag-out, has high registration precision and full-process automation, provides accurate ROI basis for micro-cutting, and improves cutting efficiency and accuracy.

Inventors

YE QING
Wen Huer
MENG BO
XU QINGHUA

Assignees

苏州可帮基因科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260416

Claims (10)

1. The contour registration-based micro-cutting multi-slide ROI mapping method is characterized by comprising the following steps of: S1, acquiring a standard diagram of a reference slide and a corresponding region of interest thermodynamic diagram thereof, and acquiring a working diagram of a target slide; S2, performing image correction on the working diagram to eliminate equipment environment interference and obtain a corrected working diagram; s3, performing brightness self-adaptive adjustment and self-adaptive threshold segmentation on the standard graph, and extracting a tissue contour as a standard contour; S4, preprocessing the corrected working graph and performing self-adaptive threshold segmentation, and extracting one or more tissue contours as working contours; Step S5, respectively calculating a first physical area of the standard outline and a second physical area corresponding to the sum of pixel areas of all working outlines according to the micron number of each pixel of the standard diagram and the working diagram; determining the number K of the actual fishing pieces corresponding to the working diagram according to the ratio of the second physical area to the first physical area; S6, calculating the mass center coordinates of each working contour in the working graph, clustering all mass center coordinates according to the number K of the actual fishing pieces, and dividing the working contour of the working graph into K contour groups, wherein each contour group corresponds to an independent tissue sample; step S7, for each contour group, carrying out iterative closest point registration on the working contour of the contour group and the standard contour, and determining the optimal mapping parameters from the standard contour to the working contour of the contour group; Step S8, aiming at the optimal mapping parameters corresponding to each contour group, carrying out space transformation on the region of interest thermodynamic diagram, mapping the region of interest thermodynamic diagram onto the working diagram, and generating a target region thermodynamic diagram corresponding to the contour group; And S9, merging target area thermodynamic diagrams of all profile groups, and generating a complete target area thermodynamic diagram of the working diagram.
2. The method according to claim 1, wherein the step S2 comprises: s2.1, acquiring a noiseless white piece as a basic image, carrying out normalization processing on the noiseless white piece, and separating B, G, R three channels of pixels; s2.2, calculating the pixel mean value of each column of the basic image in each color channel, and performing zero removal treatment on the column mean value; s2.3, carrying out normalization processing on the acquired working diagram, and separating B, G, R three channels of pixels; s2.4, dividing pixel values of each color channel of the working graph by column average values of the corresponding color channels of the basic image column by column; S2.5, carrying out maximum value constraint on the pixel values of each channel after division operation, and restoring the correction result into a preset image data format; And S2.6, merging pixel values of the three channels to generate a corrected operation chart.
3. The method according to claim 1, wherein the step S5 comprises: s5.1, acquiring a first micrometer number per pixel of a standard graph and a second micrometer number per pixel of a working graph; s5.2, calculating the pixel area of the standard contour, and calculating according to the first micrometer number per pixel to obtain a first physical area; S5.3, calculating the sum of pixel areas of all working contours, and calculating according to the second micrometer number per pixel to obtain a second physical area; And S5.4, calculating the ratio of the second physical area to the first physical area, and rounding the ratio to obtain the actual number K of the fishing pieces.
4. The method of claim 3, wherein the specific formula for determining the actual number of fishing pieces in step S5 is: ; Wherein: for the number of the actual fishing pieces, For the second physical area of the object, For the first physical area of the substrate, the first physical area, Rounding operations; the calculation formulas of the first physical area and the second physical area are respectively as follows: ; ; Wherein: Pixel area for standard outline; For a first micron per pixel; pixel area for the ith working contour; N is the total number of working contours; the second microns per pixel.
5. The method according to claim 1, wherein the step S7 includes: Step S7.1, calculating a scaling factor according to the first micron number per pixel of the standard graph and the second micron number per pixel of the working graph, and carrying out size scaling on the standard contour according to the scaling factor so as to enable the physical size of the standard contour to be matched with the physical size of the working contour of the contour group to be registered; s7.2, respectively rotating the working contours of the contour groups to be registered by a plurality of preset angles to form a plurality of rotated contour group working contours; Step S7.3, performing iterative closest point registration on the scaled standard contour and the working contours of the plurality of rotated contour groups respectively to obtain a plurality of groups of registration results and registration errors corresponding to the registration results; and S7.4, selecting a group with the smallest registration error as a final registration result, and recording corresponding mapping parameters.
6. The method according to claim 1, wherein in the step S8, for the j-th profile group, the transformation formula is: ; Wherein: And A rotation matrix and a translation vector of the j-th profile group respectively; coordinates of any pixel point on a thermodynamic diagram of a region of interest of the standard graph; coordinates of corresponding pixel points on the working graph after transformation; And (3) resampling pixel values by bilinear interpolation in the affine transformation process, and generating a target area thermodynamic diagram corresponding to the contour group.
7. The method according to claim 1, further comprising, after said step S9, a step S10 of checking the mapping result: S10.1, calculating the pixel area of the generated complete target area thermodynamic diagram, and calculating according to the micron number of each pixel of the working diagram to obtain a third physical area of the complete target area thermodynamic diagram; Step S10.2, calculating the difference ratio of the third physical area and the first physical area according to the following formula ; ; Wherein: For the first physical area of the substrate, the first physical area, Is a third physical area; and S10.3, judging the difference proportion and a preset threshold value, and judging that the mapping fails if the difference proportion exceeds the preset threshold value.
8. A contour registration-based microdissection multi-slide ROI mapping apparatus comprising: the image acquisition module is used for acquiring a standard image of the reference slide and a corresponding region of interest thermodynamic diagram thereof and acquiring a working image of the target slide; the image correction module is used for carrying out image correction on the working graph so as to eliminate the environmental interference of equipment and obtain a corrected working graph; The standard contour extraction module is used for carrying out brightness self-adaptive adjustment and self-adaptive threshold segmentation on the standard graph, and extracting a tissue contour as a standard contour; The working contour extraction module is used for preprocessing the corrected working graph and performing self-adaptive threshold segmentation to extract one or more tissue contours as working contours; The drag-out judging module is used for respectively calculating a first physical area of the standard outline and a second physical area corresponding to the sum of pixel areas of all working outlines according to the micrometer number of each pixel of the standard diagram and the working diagram; determining the number K of the actual fishing pieces corresponding to the working diagram according to the ratio of the second physical area to the first physical area; the contour clustering module is used for calculating the barycenter coordinates of each working contour in the working graph, clustering all barycenter coordinates according to the number K of the actual fishing pieces, dividing the working contour of the working graph into K contour groups, and each contour group corresponds to an independent tissue sample; the mapping parameter acquisition module is used for carrying out iterative closest point registration on the working contour of each contour group and the standard contour, and determining the optimal mapping parameters from the standard contour to the working contour of the contour group; The registration mapping module is used for carrying out space transformation on the region of interest thermodynamic diagram aiming at the optimal mapping parameter corresponding to each contour group, mapping the region of interest thermodynamic diagram onto the working diagram and generating a target region thermodynamic diagram corresponding to the contour group; and the thermodynamic diagram merging module is used for merging the thermodynamic diagrams of the target areas of all the contour groups and generating the complete target area thermodynamic diagram of the working diagram.
9. A computing device, comprising: one or more processors; a memory; And one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions of the contour registration-based microdissection multi-slide ROI mapping method of any one of claims 1-7.
10. A storage medium, characterized in that the storage medium stores one or more computer readable programs, the one or more programs comprising instructions adapted to be loaded by a memory and to perform the contour registration based microdissection multi-slide ROI mapping method of any one of the preceding claims 1-7.

Description

Contour registration-based micro-cutting multi-slide ROI mapping method and device Technical Field The invention belongs to the technical field of medical image processing, the technical field of micro-cutting automation and the technical field of computer vision registration, and particularly relates to a micro-cutting multi-slide ROI mapping method and device based on contour registration. Background In tumor pathology research and clinical diagnosis, microdissection techniques are widely used to acquire specific regions (i.e., regions of interest, ROIs) from tissue sections for subsequent molecular detection, such as NGS sequencing, PCR analysis, and the like. In practical applications, multiple slides of the same case often need to be operated, and typical scenarios include: (1) Serial section scene, in which a plurality of continuous slides (interval is 3-5 microns) are cut from the same tissue block, HE staining slides are used for pathological diagnosis and ROI identification, and adjacent sections are used for molecular experiment cutting; (2) Different staining controls HE stained slides were used for tumor area identification, special stained slides (e.g. immunohistochemistry) were used for validation, white slides were used for actual cutting; (3) Multiple tissue scooping, namely, simultaneously scooping multiple tissue samples on the same slide, and respectively identifying and mapping the ROI of each tissue. In a multi-slide processing scenario, precisely mapping the ROI on a reference slide (e.g., HE staining standard chart) onto a target slide (e.g., white-slide worksheet) faces a series of technical challenges: First, the difference in staining results in the image features being quite different. HE-stained images have rich color features (nuclear bluish purple, cytoplasmic pink), while white-patch images have little color information, and traditional gray-scale or feature point (e.g., SIFT, SURF) -based image registration methods are difficult to apply in this scenario. Second, tissue morphology changes and rotates. Tissue morphology changes exist between successive slices and the direction of placement may be different (e.g., 0 °, 90 °, 180 °, or 270 ° rotation) when the slices are fished, which requires that the registration method must be rotationally invariant. Again, the grouping of the multi-tissue-salvaged pieces is identified. Multiple tissue samples may be present on the same slide, requiring accurate identification of the individual contours of each tissue and separate registration. Finally, environmental interference factors. The working environment of the micro-cutting device has interference factors such as a slide base, an illumination environment and the like, and the image quality and the contour extraction precision are affected. Aiming at the requirement of the multi-slide ROI mapping, the prior art provides various solutions, but all have obvious defects, and cannot meet the actual requirement of clinical microscopic cutting: 1) The method based on mechanical coordinates has extremely high requirements on the precision of the placement position of the slide, and cannot adapt to the morphological change and rotation of any angle among tissue slices. 2) The feature point registration-based method relies on image features (such as SIFT and SURF), and the success rate of feature matching is extremely low under the scene of great difference of imaging conditions of HE staining images and white-patch images. 3) The method based on intensity correlation requires that the images to be registered have similar gray level distribution, and cannot cope with the situation of large dyeing difference. 4) The method lacks multiple tissue processing capability, can not automatically identify and process multiple independent tissues on the same slide, and is difficult to meet the requirement of batched multiple tissue processing. Disclosure of Invention In order to solve the technical problems, the invention provides a method and a device for mapping a micro-cut multi-slide ROI based on contour registration. In order to achieve the above purpose, the technical scheme of the invention is as follows: in a first aspect, the invention discloses a contour registration-based micro-cut multi-slide ROI mapping method, which comprises the following steps: S1, acquiring a standard diagram of a reference slide and a corresponding region of interest thermodynamic diagram thereof, and acquiring a working diagram of a target slide; s2, performing image correction on the working diagram to eliminate equipment environment interference and obtain a corrected working diagram; Step S3, performing brightness self-adaptive adjustment and self-adaptive threshold segmentation on the standard graph, and extracting a tissue contour as a standard contour; S4, preprocessing and self-adaptive threshold segmentation are carried out on the corrected working graph, and one or more tissue contours are extracted to serve