CN-122025155-A - High-resolution three-dimensional biventricular image construction method, device and medium

CN122025155ACN 122025155 ACN122025155 ACN 122025155ACN-122025155-A

Abstract

The application relates to the technical field of medical images, and provides a high-resolution three-dimensional biventricular image construction method, equipment and medium. The method comprises the steps of obtaining a two-dimensional original image sequence of delayed enhancement cardiac magnetic resonance LGE-MRI, preprocessing the two-dimensional original image sequence to obtain a target ventricular range image sequence, detecting and screening corner points of the target ventricular range image sequence to obtain a target feature point set, and carrying out three-dimensional interpolation according to the target feature point set and a preset variation implicit function to obtain a high-resolution three-dimensional biventricular image. Based on the scheme of the application, the construction efficiency and quality of the high-resolution three-dimensional double-ventricle image can be improved.

Inventors

WANG ZEFENG
DENG DONGDONG
DONG RUIQING
CHENG LITING
Liu Huanfu
XIA LING
WU YONGQUAN

Assignees

首都医科大学附属北京安贞医院

Dates

Publication Date: 20260512
Application Date: 20251212

Claims (10)

1. A high resolution three-dimensional biventricular image construction method, comprising: Acquiring a two-dimensional original image sequence of delayed and enhanced cardiac magnetic resonance (LGE-MRI), and preprocessing the two-dimensional original image sequence to obtain a target ventricular range image sequence; Performing corner detection and screening on the target ventricular range image sequence to obtain a target feature point set; and carrying out three-dimensional interpolation according to the target feature point set and a preset variation implicit function to obtain a high-resolution three-dimensional biventricular image.
2. The method of claim 1, wherein preprocessing the two-dimensional raw image sequence to obtain a target ventricular extent image sequence comprises: image segmentation is carried out on the two-dimensional original image sequence to obtain an initial range image sequence of the inner wall of the left ventricle, an initial range image sequence of the outer wall of the left ventricle, an initial range image sequence of the inner wall of the right ventricle and an initial range image sequence of the outer wall of the right ventricle; And respectively carrying out gray scale normalization processing on the initial range image sequence of the inner wall of the left ventricle, the initial range image sequence of the outer wall of the left ventricle, the initial range image sequence of the inner wall of the right ventricle and the initial range image sequence of the outer wall of the right ventricle to obtain a target ventricular range image sequence.
3. The method of claim 1, wherein the performing corner detection and screening on the target ventricular range image sequence to obtain a target feature point set includes: based on a preset Harris corner detection algorithm, calculating a corner response value of each frame of image in the target ventricular range image sequence to obtain a corner response value of a pixel point of each frame of image; determining the maximum response value in the corner response values of the pixel points of all frame images in the target ventricular range image sequence; Determining a first threshold according to the maximum response value and a preset first proportion, comparing the angular point response value of the pixel points of each frame of image with the first threshold, and constructing a pixel point with the angular point response value larger than the first threshold to obtain an initial angular point set; and screening the initial angle point set to obtain a target characteristic point set.
4. The method of claim 3, wherein the performing, based on a preset Harris corner detection algorithm, a corner response value calculation on each frame of image in the target ventricular range image sequence to obtain a corner response value of a pixel point of each frame of image includes: for each pixel point of each frame image in the target ventricular range image sequence, calculating the gray level variation of the pixel point through a window function The gray level variation The calculation formula of (2) is as follows: ; Wherein, the Indicating the amount of movement of the window, The corresponding image is represented by a representation of the image, Is a window function; the gray level variation is calculated Taylor expansion is performed, and an approximate expression is obtained as follows: ; Wherein, the In the form of a matrix, the matrix is, The expression of (2) is: ; Wherein, the For the corresponding image The derivative of the direction is used, For the corresponding image Derivative of direction; based on the matrix Calculating the corner response value of the pixel point The corresponding calculation formula is: ; Wherein, the For the matrix Is used for the treatment of the disease of the heart, For the matrix Is used for the track of (a), Is a preset constant.
5. The method of claim 4, wherein the matrix-based Calculating the corner response value of the pixel point Comprising: Order the The matrix is then The expression can be as follows: ; calculating the matrix Is of the determinant type Sum trace The corresponding calculation formulas are respectively as follows: ; ; The determinant obtained by calculation And the trace Substitution into Obtaining the angular point response value of the pixel point 。
6. The method of claim 3, wherein the screening the initial set of corner points to obtain a set of target feature points comprises: Traversing each first corner in the initial set of corner points; If the pixel values of the traversed first corner points in the 8 adjacent areas of the corresponding image are all preset foreground gray values or all preset background gray values, the traversed first corner points are removed from the initial corner point set; After traversing each first corner in the initial corner set, forming a middle corner set according to the remaining corners in the initial corner set; traversing each second corner in the set of intermediate corners; Under the condition that a third corner exists in the traversed preset size range of the second corner, if the pixel value of the third corner is a preset background gray value, the third corner is removed from the middle corner set; if the pixel value of the third corner is not the background gray value, comparing the traversed corner response value of the second corner with the corner response value of the third corner, and removing one corner with a lower corner response value from the middle corner set; after traversing each second corner in the middle corner set, forming a target feature point set according to the remaining corners in the middle corner set.
7. The method according to claim 1, wherein the performing three-dimensional interpolation according to the target feature point set and a preset variational implicit function to obtain a high-resolution three-dimensional biventricular image includes: According to the three-dimensional coordinates and the corresponding height values of each feature point in the target feature point set, solving coefficients of a preset variation implicit function to construct and obtain an interpolation function; Determining a maximum coordinate value and a minimum coordinate value of the target feature point set in a three-dimensional coordinate according to a preset target resolution and the target feature point set, and expanding to obtain a reconstructed three-dimensional coordinate space according to the maximum coordinate value, the minimum coordinate value and a preset boundary correction coefficient, wherein the resolution of the three-dimensional coordinate space is determined by the target resolution; based on the interpolation function, performing three-dimensional interpolation calculation on points in the three-dimensional coordinate space to respectively obtain a three-dimensional interpolation image of the left ventricle inner membrane, a three-dimensional interpolation image of the left ventricle outer membrane, a three-dimensional interpolation image of the right ventricle inner membrane and a three-dimensional interpolation image of the right ventricle outer membrane; Combining the high three-dimensional interpolation image of the left ventricular epicardium and the three-dimensional interpolation image of the left ventricular epicardium to obtain a left ventricular myocardial image; combining the three-dimensional interpolation image of the right ventricle inner membrane and the three-dimensional interpolation image of the right ventricle outer membrane to obtain a right ventricle myocardial image; and superposing the left ventricular myocardial image on the right ventricular myocardial image, and reserving a part, positioned outside the left ventricle, of the right ventricular myocardial image to obtain a high-resolution three-dimensional biventricular image.
8. The method of claim 7, wherein the expression of the variational implicit function is: ; Wherein, the As a function of the said interpolation function, For any point in the three-dimensional coordinate space, Set the first target feature point The position coordinates of the individual feature points, Is the first to The weights corresponding to the individual feature points are used, As a function of the radial basis function, Is a first order polynomial; The radial basis function The expression of (2) is: ; based on the position coordinates of each feature point in the target feature point set Height value corresponding to the height value Establishing the following linear equation system to solve the weight And the one-time polynomial Coefficients of (2): ; Wherein, the 。
9. An electronic device comprising a processor and a memory storing a computer program, characterized in that the processor implements the steps of the high resolution three-dimensional bi-ventricular image construction method according to any of claims 1 to 8 when the computer program is executed.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the steps of the high resolution three-dimensional bi-ventricular image construction method according to any one of claims 1 to 8.

Description

High-resolution three-dimensional biventricular image construction method, device and medium Technical Field The invention relates to the technical field of medical images, in particular to a high-resolution three-dimensional bi-ventricular image construction method, equipment and medium. Background The high-resolution three-dimensional biventricular image plays an important role in diagnosis, operation planning and curative effect evaluation of cardiovascular diseases, and particularly, a series of two-dimensional images are generally constructed to obtain a three-dimensional ventricular image capable of visualizing the anatomical structure and functional state of the heart. The related art has limited accuracy and smoothness when dealing with structures having complex curved surfaces such as ventricles. In order to improve the construction quality, a large number of constraint points are needed to be selected manually to serve as a modeling basis in the technical scheme, so that the process is time-consuming and labor-consuming, high in subjectivity and difficult to guarantee the consistency and the optimality of selection, and the construction efficiency and the quality of the high-resolution three-dimensional biventricular image are limited. Therefore, how to improve the construction efficiency and quality of the high-resolution three-dimensional bi-ventricular image has become a technical problem to be solved. Disclosure of Invention The embodiment of the application provides a high-resolution three-dimensional biventricular image construction method, equipment and medium, which are used for solving the technical problem of how to improve the construction efficiency and quality of the high-resolution three-dimensional biventricular image. In a first aspect, an embodiment of the present application provides a method for constructing a high-resolution three-dimensional biventricular image, including: acquiring a two-dimensional original image sequence of delayed enhancement cardiac magnetic resonance (LGE-MRI), and preprocessing the two-dimensional original image sequence to obtain a target ventricular range image sequence; performing corner detection and screening on the target ventricular range image sequence to obtain a target feature point set; And carrying out three-dimensional interpolation according to the target feature point set and a preset variation implicit function to obtain a high-resolution three-dimensional biventricular image. In a second aspect, an embodiment of the present application provides an electronic device, including a processor and a memory storing a computer program, where the processor implements the steps of the high resolution three-dimensional biventricular image construction method of the first aspect when the processor executes the program. In a third aspect, embodiments of the present application provide a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the high resolution three-dimensional biventricular image construction method of the first aspect. According to the high-resolution three-dimensional biventricular image construction method, device and medium, firstly, a two-dimensional original image sequence of delayed enhancement cardiac magnetic resonance LGE-MRI is obtained, the two-dimensional original image sequence is preprocessed to obtain a target ventricular range image sequence, then, angular point detection and screening are carried out on the target ventricular range image sequence, so that a target characteristic point set is obtained, finally, three-dimensional interpolation is carried out by utilizing a preset variation implicit function based on the target characteristic point set, so that a high-resolution three-dimensional biventricular image is constructed, a standardized data basis is provided for subsequent processing through the process, the angular point detection and screening carried out on the basis can automatically and accurately extract the target characteristic point set meeting the subsequent interpolation requirement, the high-quality target characteristic point set is used as input of the preset variation implicit function, the accuracy and smoothness of the three-dimensional interpolation process are guaranteed, and finally, the automatic construction from the two-dimensional original image sequence of the delayed enhancement cardiac magnetic resonance LGE-MRI to the high-resolution three-dimensional biventricular image is realized, and the high-resolution biventricular image construction efficiency and high-quality biventricular image construction efficiency are improved. Drawings In order to more clearly illustrate the application or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in