CN-121998926-A - Digestive tract tumor recognition method, digestive tract tumor recognition system and digestive tract tumor recognition medium

CN121998926ACN 121998926 ACN121998926 ACN 121998926ACN-121998926-A

Abstract

The invention discloses a method, a system and a medium for identifying digestive tract tumor, which relate to the technical field of medical image processing, dynamically optimize the step length and the layer number of Z-axis sampling according to the calculated bevel angle, on the premise of ensuring the acquisition of the complete gland three-dimensional structure, the problems of local defocusing and morphological distortion caused by tissue beveling are effectively avoided, and the definition of the image and the interlayer registration accuracy are remarkably improved. In addition, a virtual perpendicular tangent plane reconstruction strategy is introduced, so that geometric deformation can be corrected in a three-dimensional voxel space, and an observation view conforming to a standard anatomical view angle is provided, thereby eliminating the interference of the film making artifact on diagnosis. And by distinguishing the mucous membrane layer from the non-mucous membrane layer and applying a differentiated scanning strategy, the scanning time and the data redundancy of the non-diagnosis area are greatly reduced while the data quality of the key area is improved, the double optimization of the scanning precision and the efficiency is realized, and the accuracy and the robustness of the automatic identification of the digestive tract tumor are remarkably improved.

Inventors

REN YUAN
Duan Xunmei
LEI YU

Assignees

四川省肿瘤医院

Dates

Publication Date: 20260508
Application Date: 20260120

Claims (10)

1. A method for identifying digestive tract tumors comprises the steps of obtaining a low-power pre-scan image of a digestive tract pathology slide, identifying a tissue area in the low-power pre-scan image, extracting glandular tube structural features in the tissue area, carrying out ellipse fitting on candidate glandular tube units, calculating morphological parameters of glandular tube cross sections, constructing a section gradient estimation model, calculating a section gradient angle of a physical section relative to a vertical axis of a mucous membrane based on the morphological parameters, generating an adaptive Z-axis scanning strategy based on the section gradient angle, wherein the strategy comprises the steps of adjusting a Z-axis sampling step size and the sampling layer number, carrying out high-power scanning according to the adaptive Z-axis scanning strategy, generating full-slide image data containing depth information, extracting tumor morphological features based on the full-slide image data, and identifying digestive tract tumor areas.
2. The method for identifying the digestive tract tumor according to claim 1, wherein the method for identifying the digestive tract tumor is characterized by specifically further comprising the steps of separating a low-power pre-scan image into a hematoxylin channel and an eosin channel based on color deconvolution, detecting a ring-shaped or tubular closed region formed by cell nucleus arrangement in the hematoxylin channel image to define a candidate glandular tube unit, performing least square ellipse fitting on the candidate glandular tube unit to obtain a fitting ellipse minor axis length and a fitting ellipse major axis length, and the morphological parameters comprise the ratio of the fitting ellipse minor axis length to the fitting ellipse major axis length.
3. The method for recognizing tumor in digestive tract according to claim 2, wherein the constructing the model for estimating inclination of the tangent plane comprises setting a cross section of the gland tube in an ideal vertical tangent plane to be circular, a ratio of a short axis length to a long axis length of the gland tube to be 1, calculating an inclination angle of the tangent plane, determining the region as a beveled region when the inclination angle of the tangent plane is greater than a preset first inclination angle threshold, and determining the inclination angle of the tangent plane Calculated based on the following formula: Wherein, the To fit the length of the minor axis of the ellipse, To fit the length of the major axis of the ellipse.
4. The method for identifying tumor in digestive tract according to claim 2, wherein when calculating the inclination angle of the physical slice with respect to the vertical axis of the mucosa, the method specifically further comprises the steps of calculating the alignment direction consistency of all candidate glandular tube units in a local area with a preset size, determining that the embedding is systematic if the included angle of the long axis direction of the candidate glandular tube units exceeding a preset proportion in the local area is smaller than a preset angle threshold and the variance of the morphological parameter is smaller than a preset variance threshold, and applying the adaptive Z-axis scanning strategy by using the uniform inclination angle of the slice to the local area determined as the embedding is systematic.
5. The method for identifying the tumor of the digestive tract according to claim 1, wherein the generation of the self-adaptive Z-axis scanning strategy specifically comprises the steps of setting a basic Z-axis sampling step size and a basic sampling layer number, reducing the Z-axis sampling step size and increasing the basic sampling layer number along with the increase of the inclination angle of the tangent plane, enabling the product of the adjusted sampling step size and the sampling layer number to be larger than or equal to the product of the average diameter of the gland duct and the tangent value of the inclination angle of the tangent plane, and increasing the distribution density of automatic focusing points in the current view field when the inclination angle of the tangent plane is larger than a second inclination angle threshold value, and unevenly arranging focusing points along the inclination gradient direction according to the inclination direction of the long axis of the gland duct.
6. The method for recognizing digestive tract tumor according to claim 5, wherein the step of performing high-power scanning comprises obtaining multiple frames of images of the same view field at different Z-axis heights, calculating an interlayer registration offset based on the tilt angle of the tangential plane, spatially correcting the multiple frames of images, extracting high-frequency clear textures in each layer of images by using a wavelet transform fusion algorithm, and synthesizing an extended depth-of-field image to eliminate local defocus blur caused by beveling.
7. The method for identifying the digestive tract tumor according to claim 1, further comprising the steps of three-dimensional voxel reconstruction, for the region where the inclination angle of the tangent plane exceeds the correction threshold value, performing three-dimensional voxel reconstruction based on the acquired image data containing depth information, re-performing digital slicing in a direction perpendicular to the mucosal layer in a three-dimensional voxel space to generate a virtual vertical tangent plane image, and establishing an associated index of the original chamfer image and the virtual vertical tangent plane image in the generated full slide image so as to facilitate view switching during film reading.
8. The method for identifying the digestive tract tumor according to claim 1, wherein the identifying the tissue region in the low-power pre-scan image specifically comprises the steps of dividing the tissue region into a mucosal layer region and a non-mucosal layer region based on the texture feature region, executing the glandular tube structural feature extraction and adaptive Z-axis scanning strategy only in the mucosal layer region, and maintaining a default single-layer planar scanning strategy for the non-mucosal layer region.
9. A digestive tract tumor recognition system is characterized by comprising a double-rate optical imaging unit, a Z-axis precise driving execution unit, a microscopic morphology analysis unit, an adaptive scanning control unit and an image processing and recognizing unit, wherein the double-rate optical imaging unit is used for executing low-power pre-scanning and high-power fine scanning and outputting digital image signals, the Z-axis precise driving execution unit is used for bearing pathological slides and executing longitudinal displacement of step sizes and longitudinal displacement of layer numbers according to received motion control signals, the microscopic morphology analysis unit is connected with the optical imaging unit and used for recognizing gland structures in low-power pre-scanning images, fitting cross section morphologies of gland ducts and resolving pathological section inclined angles, the adaptive scanning control unit is connected with the microscopic morphology analysis unit and the Z-axis precise driving execution unit and used for generating Z-axis motion control signals containing step size instructions and layer number instructions according to the section inclined angles, and the image processing and recognizing unit is connected with the optical imaging unit and used for generating full slide image data and recognizing tumor areas.
10. A computer-readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements a method for identifying a digestive tract tumor according to any one of claims 1 to 8.

Description

Digestive tract tumor recognition method, digestive tract tumor recognition system and digestive tract tumor recognition medium Technical Field The invention belongs to the technical field of medical image processing, and particularly relates to a digestive tract tumor recognition method, a digestive tract tumor recognition system and a digestive tract tumor recognition medium. Background Early diagnosis and classification of digestive tract tumors is highly dependent on microstructure analysis of histopathological sections, where the arrangement of the gland ducts, morphology regularity and polarity of the nuclei are the core basis for the pathologist to determine benign and malignant. The existing digital pathology flow mainly focuses on the color reduction degree and the plane definition of the image, so that the visual effect of the tissue slice under a microscope is truly reduced, and a digital data base is provided for subsequent pathology analysis. However, when the existing digital slice scanning technology is used for treating the digestive tract tissues, due to the limitation of pathological tissue embedding technology and the angle deviation of physical cutting of a slicing machine, the physical slice obtained in practice often cannot be guaranteed to be completely perpendicular to the surface of a mucous membrane, and the non-perpendicular beveling can cause the originally circular gland duct cross section to be in an oval shape or other stretching shape on the image, so that judgment on the tumor gland duct abnormal shape is interfered. In addition, the traditional full-slide scanner generally adopts a single focal plane or a Z-axis stacking scanning strategy with a preset fixed layer number, and the scanning mode cannot sense the local space inclination state of tissues, so that the deep cell structure of a beveling area is easy to generate defocusing blurring due to exceeding the depth of field range, key texture information is lost, a large amount of invalid blank data or redundant layers are generated, the utilization rate of scanning efficiency and storage space is reduced, and the accurate tumor identification capability in a three-dimensional microenvironment is limited. Disclosure of Invention Aiming at the defects in the prior art, the invention provides a digestive tract tumor recognition method, a digestive tract tumor recognition system and a digestive tract tumor recognition medium, so as to solve the technical problems. A method for identifying a tumor in the digestive tract, comprising the steps of: Acquiring a low-power pre-scan image of a pathological slide of the digestive tract, and identifying a tissue region in the low-power pre-scan image; extracting the structural characteristics of the gland duct in the tissue area, carrying out ellipse fitting on candidate gland duct units, and calculating the morphological parameters of the cross section of the gland duct; Constructing a section gradient estimation model, and calculating a section gradient angle of the physical section relative to a mucosa vertical axis based on the morphological parameters; generating a self-adaptive Z-axis scanning strategy based on the section inclination angle, wherein the strategy comprises the step of adjusting Z-axis sampling step length and sampling layer number; performing high-power scanning according to the self-adaptive Z-axis scanning strategy to generate full-slide image data containing depth information; And extracting tumor morphological characteristics based on the whole-slide image data, and identifying a digestive tract tumor area. Preferably, the method for extracting the structural characteristics of the gland duct specifically further comprises the following steps: Separating the low-power pre-scan image into a hematoxylin channel and an eosin channel based on color deconvolution; detecting a ring-shaped or tubular closed region formed by cell nucleus arrangement in a hematoxylin channel image, wherein the ring-shaped or tubular closed region is defined as a candidate gland duct unit; carrying out least square ellipse fitting on the candidate gland duct units to obtain the length of the short axis of the fitting ellipse and the length of the long axis of the fitting ellipse; the morphological parameters include a ratio of a length of a minor axis of the fitted ellipse to a length of a major axis of the fitted ellipse. Preferably, the constructing the section inclination estimation model specifically includes the following steps: setting the cross section of the gland tube under an ideal vertical section to be circular, wherein the ratio of the length of the short axis to the length of the long axis is 1; Calculating a section inclined angle, and judging the area as a beveling area when the section inclined angle is larger than a preset first inclined angle threshold value; The angle of inclination of the tangent plane Calculated based on the following formula: Wherein, the To fit