CN-122004930-A - Fetal four-cavity heart measuring device and method based on ellipse detection network

CN122004930ACN 122004930 ACN122004930 ACN 122004930ACN-122004930-A

Abstract

The invention provides a fetal four-cavity heart measuring device and method based on an ellipse detection network, comprising the steps of obtaining a four-cavity heart tangential plane image; marking the outline of a heart and a chest on a four-cavity heart tangential image by using an ellipse to obtain a marked four-cavity heart tangential image, establishing a plane rectangular coordinate system on the marked four-cavity heart tangential image to obtain an ellipse equation, obtaining ellipse parameters to be predicted according to the ellipse equation, using the marked four-cavity heart tangential image as a data set, carrying out regression on the ellipse parameters to be predicted by using an ellipse detection network to obtain a trained ellipse detection network, and inputting the current four-cavity heart tangential image into the trained ellipse detection network to obtain a heart-chest ratio and a mandrel included angle. The fetal four-cavity heart-chest ratio and mandrel measuring device based on the ellipse detection network can greatly improve the accuracy of measuring the heart-chest ratio and mandrel clamp angle.

Inventors

HE YIHUA
WANG JINGYI
ZHANG YINGYING
ZHANG TONG
LIU RUNZE
HAO XIAOYAN
MAN TINGTING
ZHANG YE
ZHENG SHUAI

Assignees

首都医科大学附属北京安贞医院

Dates

Publication Date: 20260512
Application Date: 20251204

Claims (10)

1. A fetal four-chamber heart measurement device based on an ellipse detection network, characterized by comprising: The four-cavity heart tangential plane image acquisition module is used for acquiring four-cavity heart tangential plane images; The marking module is used for marking the outlines of the heart and the chest on the four-cavity heart tangential plane image by using an ellipse to obtain a marked four-cavity heart tangential plane image; the elliptic equation construction module is used for establishing a plane rectangular coordinate system on the marked four-cavity center tangential plane image to obtain an elliptic equation; the ellipse parameter calculation module is used for obtaining ellipse parameters to be predicted according to the ellipse equation; The ellipse detection training module is used for taking the marked four-cavity heart tangential plane image as a data set, and carrying out regression on the ellipse parameters to be predicted by using an ellipse detection network to obtain a trained ellipse detection network; and the heart-chest ratio and mandrel included angle calculation module is used for inputting the current four-cavity heart tangential plane image into the training-completed ellipse detection network to obtain the heart-chest ratio and the mandrel included angle.
2. The fetal four-chamber heart measurement device based on an ellipse detection network as defined in claim 1, wherein the ellipse equation construction module establishes a plane rectangular coordinate system on the noted four-chamber heart tangential image to obtain an ellipse equation by: the formula is adopted: An elliptic equation is obtained, wherein a represents the length of a long half shaft, b represents the length of a short half shaft, The angle of the ellipse is indicated and, The abscissa of the center point is indicated, Representing the ordinate of the center point.
3. The fetal four-chamber heart measurement device based on an ellipse detection network of claim 2, wherein an ellipse parameter calculation module obtains an ellipse parameter to be predicted from the ellipse equation by: deriving the elliptic equation to obtain a derived elliptic equation; obtaining the intercept of the horizontal tangent of the ellipse and the y axis and the intercept of the vertical tangent of the ellipse and the x axis according to the derived elliptic equation; Obtaining an intermediate predicted value according to the intercept of the horizontal tangent line and the y axis and the intercept of the vertical tangent line and the x axis; Expanding the ellipse to form an expanded square; taking the intermediate predicted value as the side length of the square after expansion; constructing a relation between an ellipse parameter to be predicted and an expanded square parameter according to the position relation between the expanded square and the ellipse; and obtaining the ellipse parameters to be predicted according to the relational expression.
4. A fetal four-chamber heart measurement apparatus based on an ellipse detection network as defined in claim 3, wherein an ellipse parameter calculation module performs the following operations when deriving the ellipse equation to obtain a derived ellipse equation: the formula is adopted: Deriving the elliptic equation to obtain a derived elliptic equation, wherein a represents the length of a major half shaft, b represents the length of a minor half shaft, Representing the elliptical angle.
5. The fetal four-chamber heart measurement apparatus based on an ellipse detection network of claim 4, wherein the ellipse parameter calculation module performs the following operations when deriving an intercept of a horizontal tangent of the ellipse to the y-axis and an intercept of a vertical tangent of the ellipse to the x-axis from the derived ellipse equation: the formula is adopted: Obtaining the intercept of the horizontal tangent of the ellipse with the y-axis and the intercept of the vertical tangent of the ellipse with the x-axis, wherein, Representing the intercept of the horizontal tangent to the y-axis, Representing the intercept of the vertical tangent to the x-axis.
6. The fetal four-chamber heart measurement apparatus based on an ellipse detection network of claim 5, wherein the ellipse parameter calculation module performs the following operations when obtaining an intermediate predicted value from the intercept of the horizontal tangent to the y-axis and the intercept of the vertical tangent to the x-axis: the formula is adopted: Obtaining an intermediate predicted value, wherein, Representing the intermediate predictor.
7. The fetal four-chamber heart measurement device based on an ellipse detection network of claim 6, wherein the ellipse parameter calculation module performs the following operations when constructing a relation between an ellipse parameter to be predicted and an expanded square parameter according to the expanded square and ellipse positional relation: the formula is adopted: constructing a relation between the ellipse parameters to be predicted and the square parameters after expansion, wherein, Representing the absolute offset of the center point in the x-axis, Representing the absolute offset of the center point in the y-axis, Representing the absolute offset of the long half-shaft, Representing the absolute offset of the minor half axis, Representing the absolute offset of the elliptical angle, The abscissa representing the center of the expanded square, Representing the ordinate of the center of the expanded square, Represents the side length of the square after expansion, Representing the abscissa of the center of the ellipse to be predicted, Representing the ordinate of the centre of the ellipse to be predicted, Indicating the length of the major half axis of the ellipse to be predicted, Representing the length of the shorter half axis of the ellipse to be predicted, Representing the elliptical angle to be predicted.
8. The fetal four-chamber heart measurement device of claim 1, wherein the ellipse detection network is configured to perform the following functions: performing feature extraction on the marked four-cavity center tangent plane image to obtain a feature image; Inputting the characteristic image into a square detection module to obtain the position of the expanded square and the side length of the expanded square; Inputting the characteristic image into an ellipse regression module to obtain the ellipse long half-axis length, the ellipse short half-axis length and the rotation angle of the ellipse; And obtaining a heart-chest ratio and a mandrel included angle according to the position of the expanded square, the side length of the expanded square, the length of the elliptic long half-axis, the length of the elliptic short half-axis and the rotation angle of the ellipse.
9. The fetal four-chamber heart measurement method based on the ellipse detection network is characterized by comprising the following steps of: acquiring a four-cavity heart tangential plane image; Marking the outline of the heart and the chest on the four-cavity heart tangential plane image by using an ellipse to obtain a marked four-cavity heart tangential plane image; Establishing a plane rectangular coordinate system on the marked four-cavity center tangent plane image to obtain an elliptic equation; Obtaining ellipse parameters to be predicted according to the ellipse equation; The marked four-cavity center tangent plane image is used as a data set, and an ellipse detection network is utilized to carry out regression on the ellipse parameters to be predicted to obtain a trained ellipse detection network; And inputting the current four-cavity heart tangential plane image into the trained elliptical detection network to obtain the heart-chest ratio and the included angle of the mandrel.
10. The method for fetal four-chamber heart measurement based on an elliptical detection network of claim 9, Establishing a plane rectangular coordinate system on the marked four-cavity center tangent plane image to obtain an elliptic equation, wherein the method comprises the following steps of: the formula is adopted: An elliptic equation is obtained, wherein a represents the length of a long half shaft, b represents the length of a short half shaft, The angle of the ellipse is indicated and, The abscissa of the center point is indicated, An ordinate representing a center point; the formula is adopted: Deriving the elliptic equation to obtain a derived elliptic equation, wherein a represents the length of a major half shaft, b represents the length of a minor half shaft, Representing an elliptical angle; obtaining an intermediate predicted value according to the intercept of the horizontal tangent line and the y axis and the intercept of the vertical tangent line and the x axis, wherein the formula is adopted: Obtaining the intercept of the horizontal tangent of the ellipse with the y-axis and the intercept of the vertical tangent of the ellipse with the x-axis, wherein, Representing the intercept of the horizontal tangent to the y-axis, Representing the intercept of the vertical tangent to the x-axis; the formula is adopted: Obtaining an intermediate predicted value, wherein, Representing the intermediate predicted value; Expanding the ellipse to form an expanded square; taking the intermediate predicted value as the side length of the square after expansion; Constructing a relation between the ellipse parameters to be predicted and the expanded square parameters according to the position relation between the expanded square and the ellipse, wherein the formula is adopted: constructing a relation between the ellipse parameters to be predicted and the square parameters after expansion, wherein, Representing the absolute offset of the center point in the x-axis, Representing the absolute offset of the center point in the y-axis, Representing the absolute offset of the long half-shaft, Representing the absolute offset of the minor half axis, Representing the absolute offset of the elliptical angle, The abscissa representing the center of the expanded square, Representing the ordinate of the center of the expanded square, Represents the side length of the square after expansion, Representing the abscissa of the center of the ellipse to be predicted, Representing the ordinate of the centre of the ellipse to be predicted, Indicating the length of the major half axis of the ellipse to be predicted, Representing the length of the shorter half axis of the ellipse to be predicted, Representing an elliptical angle to be predicted; and obtaining the ellipse parameters to be predicted according to the relational expression.

Description

Fetal four-cavity heart measuring device and method based on ellipse detection network Technical Field The invention relates to the technical field of image preprocessing, in particular to a fetal four-cavity heart-chest ratio and mandrel measuring device and method based on an ellipse detection network. Background Fetal Congenital Heart Disease (FCHD) is a first disease of birth defects worldwide and has the characteristics of high composition ratio, high disability and high lethality. Fetal echocardiography is still currently the most effective method of determining fetal heart malformations. Fetal structural heart diseases are various in appearance, particularly complex heart deformity, the heart function state and hemodynamic changes are more and more varied, and the heart function change of the fetus is one of the earliest and most direct indexes for clinically evaluating the prognosis of the fetus, so that prenatal judgment of the heart function of the fetus is particularly important. Most of the existing AI intelligent auxiliary diagnosis methods do the classification of diseases. Most of them use internationally published data sets or report-carrying data collected from hospitals, acquire sheet-level disease labels from reports through natural language processing, train a multi-label classification network model or a plurality of single disease classification models, then predict input images, and convert disease characteristic information learned by a deep convolutional neural network model into a thermodynamic diagram form by adopting a class activation mapping (or a weighted gradient class activation mapping) method, so that a doctor can see about where a disease approximate focus area is on the thermodynamic diagram, and further assist doctor diagnosis. However, for fetal congenital heart diseases, the disease pedigree is particularly wide, the disease pedigree comprises hundreds of types, some patients can combine multiple deformities at the same time, the symptoms are quite different, and the disease is not as obvious as common diseases in focus areas, so that diagnosis can not be made through a simple disease classification or focus detection model. Meanwhile, congenital heart disease is one of the most common congenital deformities, which is related to abnormal anatomical structures caused by the formation disorder or dysplasia of heart and large blood vessels during the embryonic development period, and thus can be diagnosed by measuring heart-related indexes. Commonly used indexes are a heart-chest ratio, which is the ratio of areas of a heart region and a chest region, and an included angle of the spindle, which refers to the angle that the apex of the heart points to and is presented by a line connecting the apex of the heart and the sternum of the spine. When the cardiothoracic ratio shows a significant increase, it is suggested that the heart is significantly dilated and that cardiac function may be impaired. Currently, four-chamber heart ultrasound sectional images are a common tool for examining or diagnosing fetal congenital heart disease. In a trimethyl hospital, the number of fetal ultrasonic images produced by daily life is very large, on one hand, a reader needs to manually mark the inner diameters and directions of a fetal heart and a thoracic cavity on an ultrasonic machine, so that the measurement of the heart-chest ratio and the included angle index of a mandrel is very time-consuming and labor-consuming, and on the other hand, fatigue is gradually accumulated in the long-time film reading process, and the condition of large measurement error caused by inaccurate marking can occur. Disclosure of Invention In order to solve the above problems, an object of an embodiment of the present invention is to provide a method for measuring a fetal four-chamber cardiothoracic ratio and a mandrel based on an ellipse detection network. According to a first aspect of the present invention, there is provided a fetal four-chamber heart measurement device based on an ellipse detection network, comprising: The four-cavity heart tangential plane image acquisition module is used for acquiring four-cavity heart tangential plane images; The marking module is used for marking the outlines of the heart and the chest on the four-cavity heart tangential plane image by using an ellipse to obtain a marked four-cavity heart tangential plane image; the elliptic equation construction module is used for establishing a plane rectangular coordinate system on the marked four-cavity center tangential plane image to obtain an elliptic equation; the ellipse parameter calculation module is used for obtaining ellipse parameters to be predicted according to the ellipse equation; The ellipse detection training module is used for taking the marked four-cavity heart tangential plane image as a data set, and carrying out regression on the ellipse parameters to be predicted by using an ellipse detection network to obtai