CN109741395B

CN109741395B - Dual-chamber quantification method and device, electronic equipment and storage medium

Info

Publication number: CN109741395B
Application number: CN201811534455.6A
Authority: CN
Inventors: 王文集; 胡志强; 李嘉辉; 闫桢楠
Original assignee: Beijing Sensetime Technology Development Co Ltd
Current assignee: Beijing Sensetime Technology Development Co Ltd
Priority date: 2018-12-14
Filing date: 2018-12-14
Publication date: 2021-07-23
Anticipated expiration: 2038-12-14
Also published as: CN113674338A; CN109741395A

Abstract

The embodiment of the application provides a biventricular quantization method, a biventricular quantization device, an electronic device and a storage medium, wherein the method comprises the following steps: acquiring a target original image, wherein the target original image is a nuclear magnetic resonance image of a heart; segmenting the target original image by adopting a preset segmentation network to obtain a target multichannel probability image; the target multichannel probability image is input into a preset depth regression network for calculation, so that target comprehensive quantitative data of the heart are obtained, and the target comprehensive quantitative data comprise heart cavity diameter, myocardial wall thickness and ventricular area.

Description

Dual-chamber quantification method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of image processing technologies, and in particular, to a biventricular quantization method and apparatus, an electronic device, and a storage medium.

Background

Magnetic resonance imaging is one of the important imaging modalities for detecting heart disease. Quantification of the ventricles from the magnetic resonance images aids in the early diagnosis and treatment of heart disease. In recent years, the academia has developed a great deal of research primarily in left ventricular quantification.

Since the current work is mainly focused on the left ventricle, and comprehensive and accurate computer-aided diagnosis is desired, it is most desirable to quantify the biventricles. The existing scheme for quantifying the biventricular volume is four-ventricular volume calculation, but the quantification comprehensiveness is low in the aspect of comprehensive quantification, so that the quantification effect is reduced.

Disclosure of Invention

The embodiment of the application provides a biventricular quantization method and device, electronic equipment and a storage medium, which can improve comprehensiveness of ventricular quantization.

A first aspect of an embodiment of the present application provides a bi-ventricular quantification method, wherein the method includes:

acquiring a target original image, wherein the target original image is a nuclear magnetic resonance image of a heart;

segmenting the target original image by adopting a preset segmentation network to obtain a target multichannel probability image;

and inputting the target multichannel probability image into a preset depth regression network for calculation so as to obtain target comprehensive quantitative data of the heart, wherein the target comprehensive quantitative data comprises the heart cavity diameter, the myocardial wall thickness and the ventricular area.

Optionally, the preset depth regression network includes an N-layer convolutional neural network, and the inputting the target multichannel probability image into the preset depth regression network for calculation to obtain target comprehensive quantitative data of the heart includes:

inputting the target multichannel probability image into a first layer of the N-layer convolutional neural network for calculation to obtain a first calculation result, wherein N is a positive integer greater than 1;

and inputting the first calculation result into a second layer of the N-layer convolutional neural network for calculation to obtain a second calculation result, and inputting the N-1 settlement result into an Nth layer of the N-layer convolutional neural network for calculation to obtain target comprehensive quantitative data of the heart.

Optionally, the method further includes obtaining a preset depth regression network, where obtaining the preset depth regression network includes:

step 1: acquiring a sample image, wherein the sample image is a nuclear magnetic resonance image of a heart, and preprocessing the sample image to obtain a preprocessed sample image;

step 2: adopting a first segmentation network to segment the preprocessed sample image to obtain a sample four-channel probability image;

and step 3: inputting the sample four-channel probability image into the N-layer convolutional neural network for calculation to obtain a predicted value;

and 4, step 4: performing mean square error operation on the predicted value and a target true value to obtain a corrected value, wherein the target true value is a true value determined by a preset true value generating function, and the target true value corresponds to the predicted value;

and 5: correcting the first segmentation network through the correction value to obtain a second segmentation network, and replacing the first segmentation network with the second segmentation network;

and repeating the step 2 to the step 5 for M times, and in the process of repeating the step 2 to the step 5 for M times, if a preset parameter is in a preset parameter range, taking the N layers of convolutional neural networks as the preset depth regression network.

Optionally, the method further includes obtaining a preset true value generating function, where obtaining the preset true value generating function includes:

acquiring an original image with a heart contour label, wherein the original image with the heart contour label comprises a plurality of contour points;

processing the original image with the heart contour mark to obtain a mask image;

acquiring barycentric coordinates of the heart from the mask image by adopting a preset acquisition mode, wherein the barycentric coordinates comprise barycentric coordinates of a left ventricle and barycentric coordinates of a right ventricle;

acquiring the distance and the angle between each contour point in the plurality of contour points and the corresponding barycentric coordinate;

and determining the preset truth value generating function by adopting an interpolation method according to the distance and the angle between each contour point and the corresponding gravity center coordinate, wherein the preset truth value generating function is a function between the angle and the distance.

Optionally, the obtaining the barycentric coordinates of the heart from the mask image by using a preset obtaining manner includes:

determining to acquire a contour point set from the mask image through a first preset function;

and determining the barycentric coordinate by adopting a second preset function according to the contour point set.

Optionally, the target multichannel probability image includes: background region images, left/right heart chamber images, and left myocardium images.

Optionally, the method further includes:

resampling the target original image to obtain a resampled image;

and carrying out normalization processing of 0 mean value 1 variance on the resampled image to obtain a preprocessed image.

A second aspect of an embodiment of the present application provides a biventricular quantification apparatus, wherein the apparatus includes:

the system comprises a first acquisition unit, a second acquisition unit and a third acquisition unit, wherein the first acquisition unit is used for acquiring a target original image which is a nuclear magnetic resonance image of a heart;

the segmentation unit is used for segmenting the target original image by adopting a preset segmentation network to obtain a target multichannel probability image;

and the calculation unit is used for inputting the target multichannel probability image into a preset depth regression network for calculation so as to obtain target comprehensive quantitative data of the heart, wherein the target comprehensive quantitative data comprises the heart cavity diameter, the myocardial wall thickness and the ventricle area.

Optionally, the preset depth regression network includes an N-layer convolutional neural network, and in the aspect that the target multichannel probability image is input into the preset depth regression network for calculation to obtain the target comprehensive quantitative data of the heart, the calculation unit is specifically configured to:

Optionally, the apparatus further includes a second obtaining unit, where the second obtaining unit is configured to obtain a preset deep regression network, and in terms of obtaining the preset deep regression network, the second obtaining unit is specifically configured to:

and 5: correcting the preset segmentation network through the correction value to obtain a second segmentation network, and replacing the first segmentation network with the second segmentation network;

Optionally, the apparatus further includes a third obtaining unit, where the third obtaining unit is configured to obtain a preset true value generating function, and in terms of obtaining the preset true value generating function, the third obtaining unit is specifically configured to:

Optionally, in terms of acquiring barycentric coordinates of the heart from the mask map by using the preset acquisition mode, the third acquisition unit is specifically configured to:

Optionally, the apparatus is further specifically configured to:

resampling the target original image to obtain a resampled image;

A third aspect of the embodiments of the present application further provides an electronic device, including:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to perform the method described above.

The fourth aspect of the embodiments of the present application also provides a computer-readable storage medium, on which computer program instructions are stored, and the computer program instructions, when executed by a processor, implement the method described above.

The application has at least the following beneficial effects:

according to the embodiment of the application, the target original image is obtained and is a nuclear magnetic resonance image of the heart, the preset segmentation network is adopted to segment the target original image to obtain the target multichannel probability image, the target multichannel probability image is input into the preset depth regression network to be calculated to obtain the target comprehensive quantization data of the heart, the target comprehensive quantization data comprises the heart cavity diameter, the myocardial wall thickness and the ventricular area, therefore, the target original image is segmented to obtain the four-channel probability image, the characteristics of the original image can be accurately reflected through image segmentation, then the four-channel probability image is input into the depth regression network to be calculated to obtain the comprehensive quantization data, and compared with the existing scheme, the volume of the four ventricles can only be calculated, the quantitative data of the heart can be acquired more comprehensively, so that the comprehensiveness of the heart during quantification is improved to a certain extent.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 provides a schematic view of a bi-ventricular quantification system according to an embodiment of the present application;

FIG. 2A is a schematic flow chart of a biventricular quantification method according to an embodiment of the present application;

FIG. 2B is a schematic diagram of segmentation of a preprocessed image according to an embodiment of the present disclosure;

fig. 2C is a schematic structural diagram of a preset deep regression network according to an embodiment of the present disclosure;

FIG. 3 provides a schematic view of another bi-ventricular quantification method according to an embodiment of the present application;

FIG. 4 is a schematic flow chart of another bi-ventricular quantification method;

fig. 5 is a schematic structural diagram of a terminal according to an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a biventricular quantification apparatus according to an embodiment of the present disclosure;

FIG. 7 is a schematic structural diagram of another dual ventricular quantification apparatus according to an embodiment of the present application;

fig. 8 is a schematic structural diagram of another biventricular quantification apparatus according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

The terms "first," "second," and the like in the description and claims of the present application and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus.

Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the specification. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.

The electronic device according to the embodiments of the present application may include various handheld devices, vehicle-mounted devices, wearable devices, computing devices or other processing devices connected to a wireless modem, and various forms of User Equipment (UE), Mobile Stations (MS), terminal equipment (terminal device), and so on. For convenience of description, the above-mentioned apparatuses are collectively referred to as electronic devices.

In order to better understand the biventricular quantification method provided in the embodiments of the present application, a brief description of the biventricular quantification method is first provided below. Referring to fig. 1, fig. 1 is a schematic diagram of a biventricular quantization system according to an embodiment of the present application. As shown in fig. 1, the biventricular quantization system includes a preprocessing network 101, a segmentation network 102 and a depth regression network 103, first the biventricular quantization system obtains an original image, the original image is a nuclear magnetic resonance image of a heart, then the preprocessing network 101 preprocesses the original image to obtain a preprocessed image, then the preprocessed image is input to the preset segmentation network 102, the preprocessed image is segmented to obtain a target multichannel probability image, and finally the target multichannel probability image is input to the preset depth regression network 103 for calculation to obtain target comprehensive quantized data, the target comprehensive quantized data includes a heart diameter, a myocardial wall thickness and a ventricular area, in the present application scheme, the original image is preprocessed, the preprocessed image is segmented to obtain a four-channel probability image, and by image segmentation, the characteristics of an original image can be accurately reflected, then the four-channel probability image is input into the depth regression network for calculation, so that comprehensive quantitative data are obtained, compared with the existing scheme, the method can only calculate the volume of four ventricles, the quantitative data of the heart can be more comprehensively obtained, and the comprehensiveness of the heart in the quantitative process is improved to a certain extent.

Referring to fig. 2A, fig. 2A is a schematic flow chart of a biventricular quantization method according to an embodiment of the present application. As shown in FIG. 2A, the biventricular quantification method comprises steps 201-204 as follows:

201. and acquiring a target original image, wherein the target original image is a nuclear magnetic resonance image of the heart.

The original image can be input by a user, or the nuclear magnetic resonance image generated by a nuclear magnetic resonance instrument can be directly transmitted to a double-ventricle quantification system. The original image may also be an image marked by a contour mark, where the contour mark is to mark a contour of a heart in the original image, and the mark is composed of a plurality of feature points.

Optionally, after the target original image is obtained, in order to improve the segmentation effect, the target original image may be preprocessed.

One possible method for processing an original image to obtain a preprocessed image includes steps a1-a2, which are as follows:

a1, resampling the target original image to obtain a resampled image;

the resampling method includes nearest neighbor interpolation (nearest neighbor interpolation), bilinear interpolation (bilinear interpolation), and cubic convolution interpolation (cubic convolution interpolation). The scheme can adopt any one of the methods to perform resampling to obtain a resampled image.

And A2, carrying out normalization processing of 0 mean value 1 variance on the resampled image to obtain the preprocessed image.

In the normalization processing of the variance of 0 mean value 1, the 0 mean value can be understood as that the sum of the gray values of each pixel point in the resampled picture is 0; 1 variance can be understood as the variance of the gray value of each pixel point in the resampled picture is 1, and the data (gray value) of the resampled picture after normalization processing is mapped to a value between [0,1 ].

202. And segmenting the preprocessed image by adopting a preset segmentation network to obtain a target multichannel probability image.

Optionally, the preset segmentation network includes M convolution blocks, the preprocessed image is input into the segmentation network, convolution operation, sampling operation, iterative depth aggregation and hierarchical depth aggregation operation are performed, and finally a target four-channel probability map is output, where the sampling operation is to increase a sampling rate by one time. A schematic diagram of segmenting a preprocessed image is shown in fig. 2B, where a preset segmentation network includes 26 convolution blocks, 4 iterative depth aggregation networks, 3 hierarchical depth aggregation, 5 sampling operations, and finally a target four-channel probability map is output, where 1 to 26 represent convolution block 1 to convolution block 26, respectively.

Wherein the target multi-channel probability image comprises: the image processing device comprises a background region image, a left/right heart cavity image and a left heart muscle image, wherein the background region image, the left heart muscle image and the left heart cavity image or the background region image, the right heart muscle image and the right heart cavity image are specific.

203. And inputting the target multichannel probability image into a preset depth regression network for calculation so as to obtain target comprehensive quantitative data of the heart, wherein the target comprehensive quantitative data comprises the heart cavity diameter, the myocardial wall thickness and the ventricular area.

Referring to fig. 2C, fig. 2C is a schematic structural diagram of a preset deep regression network according to an embodiment of the present disclosure. As shown in fig. 2C, the deep regression network includes 11 layers of convolutional neural networks, specifically: conv1, relu layer; 4 max, pool layers; conv2, relu layer; conv3, relu layer; conv4, relu layer; conv5, relu layer; fc, relu layer, Fc, sigmoid layer.

Optionally, the depth regression network includes an N-layer convolutional neural network, and a method of inputting the target four-channel probability map into the depth regression network for calculation to obtain comprehensive quantized data includes steps B1-B2, which are specifically as follows:

b1, inputting the target multichannel probability image into a first layer of the N-layer convolutional neural network for calculation to obtain a first calculation result, wherein N is a positive integer greater than 1;

optionally, taking the depth regression network provided in fig. 2C as an example for explanation, inputting the target multichannel probability image into the conv1 and relu layers, and performing calculation through the conv1 and the relu layers to obtain a first calculation result; and inputting the first calculation result into a second layer (max, pool layer), and calculating to obtain a second calculation result.

And B2, inputting the first calculation result into a second layer of the N-layer convolutional neural network for calculation to obtain a second calculation result, and inputting the N-1 th calculation result into an Nth layer of the N-layer convolutional neural network for calculation to obtain target comprehensive quantitative data of the heart.

Optionally, in the present application, the tenth calculation result is input to fc, and the sigmoid layer performs calculation, so as to obtain the target comprehensive quantized data.

Optionally, an embodiment of the present application further provides a method for obtaining a preset deep regression network, where the method includes the following steps:

the method for preprocessing the sample image to obtain the preprocessed sample image may refer to the method for preprocessing the original image in steps a1-a 2.

the first split network is a preset split network, and the preset split network is the split network shown in fig. 2B.

And step 3: inputting the sample four-channel probability image into an N-layer convolutional neural network for calculation to obtain a predicted value;

the N-layer convolutional neural network is a network that has not been trained, and the output result is data that has not been classified, that is, a predicted value.

optionally, an embodiment of the present application further provides a method for obtaining a true value generation function, including steps C1-C5, which are specifically as follows:

c1, acquiring an original image with a heart contour label, wherein the original image with the heart contour label comprises a plurality of contour points;

the original image with the heart contour label can be obtained by manual labeling, or by methods such as a machine labeling method. The heart contour labeling comprises the following steps: the left myocardial contour labeling, the left heart chamber contour labeling, the right heart muscle contour labeling, the right heart chamber contour labeling, and the like are described here by taking the labeling of the left myocardium and the left heart chamber as an example, that is, the original image with the heart contour labeling is the image with the left myocardial contour labeling and the left heart chamber contour labeling.

C2, processing the original image with the heart contour label to obtain a mask image;

the method for processing the original image to obtain the mask image comprises the following steps: and marking the background image in the original image as 0, and marking the left myocardial contour and the left heart chamber contour as 1 to obtain a mask image, wherein the mask image is a 0-1mask image.

C3, acquiring barycentric coordinates of the heart from the mask image by adopting a preset acquisition mode, wherein the barycentric coordinates comprise barycentric coordinates of the left ventricle and barycentric coordinates of the right ventricle;

optionally, a possible method for obtaining barycentric coordinates of the heart from the mask map includes steps C31-C32, as follows:

c31, determining to acquire a contour point set from the mask image through a first preset function;

the first preset function is a find _ constraints function in the sketch function packet, and the find _ constraints function can extract a contour from the binary function, so that a contour point set can be obtained.

Optionally, before the mask image is determined according to the first preset function, a first coordinate system is further established, the original image is a rectangular image, an origin of the first rectangular coordinate system is located at a vertex of an upper left corner of the original image, a vertical direction is a positive direction of an x axis, and a horizontal direction is a positive direction of a y axis.

And C32, determining the barycentric coordinate by adopting a second preset function according to the contour point set.

The second preset function is a center _ of _ mass function in the scipy function package, and the barycentric coordinate can be determined directly according to the contour point set through the center _ of _ mass function.

Optionally, after the barycentric coordinate is determined, a second coordinate system is established with the barycentric coordinate as an origin, a vertical direction as an x-axis positive direction and a horizontal direction as a y-axis positive direction, the second coordinate system is a polar coordinate system, and the coordinates of the contour points are converted from the first coordinate system to the coordinates in the second coordinate system.

C4, acquiring the distance and the angle between each contour point in the plurality of contour points and the corresponding barycentric coordinate;

the barycentric coordinates corresponding to each contour point can be understood as that the left myocardium contour point corresponds to the barycentric coordinates of the left ventricle, and the right myocardium contour point corresponds to the barycentric coordinates of the right ventricle.

Optionally, the distance between the contour point and the gravity center coordinate may be directly calculated by a distance calculation formula, or may be directly measured, and the angle between the contour point and the gravity center coordinate may be understood as an included angle between a connection line of the contour point and the gravity center and the positive direction of the y-axis.

And C5, determining the preset truth value generating function by an interpolation method according to the distance and the angle between each contour point and the corresponding barycentric coordinate, wherein the preset truth value generating function is a function between the angle and the distance.

Optionally, the method for determining the preset true value generation function according to the interpolation method includes: establishing a discrete function of the angle and the distance according to the distance and the angle between the contour point and the gravity center; and connecting the discrete functions into a continuous function by utilizing an interpolation method so as to obtain a preset true value generating function.

and correcting the first segmentation network by using the correction value, namely correcting errors generated when the first segmentation network segments the preprocessed picture by using the correction value, and optimizing an output result so as to obtain a corrected output result.

The preset parameter may be a loss value, a difference between a true value and an index value, a preset parameter range may be set by an empirical value or by historical data, and the loss value may be understood as a correction value.

According to the embodiment of the application, the biventricular quantization system outputs comprehensive quantization data, compared with the previous left ventricular quantization and partial multi-ventricular quantization, the biventricular quantization can reflect the real condition of the heart more objectively and truly, and is beneficial to early diagnosis and treatment of heart diseases, and the biventricular quantization system integrates segmentation and regression networks, so that the accuracy of the comprehensive quantization data can be improved to a certain extent, and by designing a new truth value generation method, compared with truth value generation standards such as 2D center line and the like, the generation process is optimized, and the efficiency of truth value generation is improved.

Referring to fig. 3, fig. 3 is a schematic view of another bi-ventricular quantification method according to an embodiment of the present application. As shown in fig. 3, an original image 301 is preprocessed 302 to obtain a preprocessed image, the preprocessed image is input into a segmentation network 303 to be segmented to obtain a four-channel probability image 304, the four-channel probability image 304 is input into a segmentation network 305 to be calculated to obtain comprehensive quantized data 306, where the comprehensive quantized data includes: left ventricle heart cavity area, left myocardium area, left ventricle heart cavity diameter, left myocardium wall thickness, right ventricle heart cavity area, right ventricle heart cavity diameter, specifically mark in the figure as: a1: right ventricular area; a2: left myocardial area; a3: left heart chamber area; d1-d 3: left heart chamber diameter; rd1-rd 2: the right lumen diameter; A-AS (A, AL, IL, I, IS, AS): left myocardial wall thickness.

Referring to fig. 4, fig. 4 is a schematic flow chart of another dual-chamber quantification method, including

steps

401 and 410, as follows:

401. acquiring a sample image, wherein the sample image is a nuclear magnetic resonance image of a heart, and preprocessing the sample image to obtain a preprocessed sample image;

402. adopting a first segmentation network to segment the preprocessed sample image to obtain a sample four-channel probability image;

403. inputting the sample four-channel probability image into the N-layer convolutional neural network for calculation to obtain a predicted value;

404. performing mean square error operation on the predicted value and a target true value to obtain a corrected value, wherein the target true value is a true value determined by a preset true value generating function, and the target true value corresponds to the predicted value;

405. correcting the first segmentation network through the correction value to obtain a second segmentation network, and replacing the first segmentation network with the second segmentation network;

406. repeating the steps 402 to 405 until the predicted value output by the N-layer convolutional neural network is comprehensive quantized data, and taking the N-layer convolutional neural network as a preset depth regression network;

407. acquiring an original image, wherein the target original image is a nuclear magnetic resonance image of a heart;

408. preprocessing the target original image to obtain a preprocessed image;

409. segmenting the preprocessed image by adopting a preset segmentation network to obtain a target multichannel probability image;

410. and inputting the target multichannel probability image into a preset depth regression network for calculation so as to obtain target comprehensive quantitative data of the heart, wherein the target comprehensive quantitative data comprises the heart cavity diameter, the myocardial wall thickness and the ventricular area.

In the example, the depth regression network is constructed, and the end-to-end connection between the segmentation network and the depth regression network is realized, so that the original image is processed, comprehensive quantitative data is obtained, and only left ventricle quantification and part of multiple ventricles can be quantified and double ventricles can be quantified comprehensively in the existing scheme, so that the real condition of the heart can be reflected more objectively and truly, and the accuracy and the comprehensiveness of data acquisition are improved.

In accordance with the foregoing embodiments, please refer to fig. 5, fig. 5 is a schematic structural diagram of a terminal according to an embodiment of the present application, and as shown in the drawing, the terminal includes a processor, an input device, an output device, and a memory, where the processor, the input device, the output device, and the memory are connected to each other, where the memory is used to store a computer program, the computer program includes program instructions, the processor is configured to call the program instructions, and the program includes instructions for performing the following steps;

In the example, a target original image is obtained, the target original image is a nuclear magnetic resonance image of a heart, a preset segmentation network is adopted to segment the target original image to obtain a target multichannel probability image, the target multichannel probability image is input into a preset depth regression network to be calculated to obtain target comprehensive quantized data of the heart, the target comprehensive quantized data comprises heart chamber diameter, myocardial wall thickness and ventricular area, therefore, in the application, a four-channel probability image is obtained by segmenting the target original image, the characteristics of the original image can be reflected more accurately through image segmentation, then the four-channel probability image is input into the depth regression network to be calculated to obtain the comprehensive quantized data, compared with the existing scheme, the volume of only four ventricles can be calculated, and the quantized data of the heart can be obtained more comprehensively, therefore, the comprehensiveness of quantization is improved to a certain extent.

The above description has introduced the solution of the embodiment of the present application mainly from the perspective of the method-side implementation process. It is understood that the terminal includes corresponding hardware structures and/or software modules for performing the respective functions in order to implement the above-described functions. Those of skill in the art will readily appreciate that the present application is capable of hardware or a combination of hardware and computer software implementing the various illustrative elements and algorithm steps described in connection with the embodiments provided herein. Whether a function is performed as hardware or computer software drives hardware depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

In the embodiment of the present application, the terminal may be divided into the functional units according to the above method example, for example, each functional unit may be divided corresponding to each function, or two or more functions may be integrated into one processing unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit. It should be noted that the division of the unit in the embodiment of the present application is schematic, and is only a logic function division, and there may be another division manner in actual implementation.

In accordance with the above, please refer to fig. 6, fig. 6 is a schematic structural diagram of a dual ventricular quantification apparatus according to an embodiment of the present disclosure. The bi-ventricular quantification apparatus comprises a first acquisition unit 601, a segmentation unit 602 and a calculation unit 603, wherein,

a first obtaining unit 601, configured to obtain a target original image, where the target original image is a nuclear magnetic resonance image of a heart;

a segmentation unit 602, configured to segment the target original image by using a preset segmentation network to obtain a target multi-channel probability image;

a calculating unit 603, configured to input the target multichannel probability image into a preset depth regression network for calculation, so as to obtain target comprehensive quantitative data of the heart, where the target comprehensive quantitative data includes a heart cavity diameter, a myocardial wall thickness, and a ventricular area.

Optionally, the preset depth regression network includes an N-layer convolutional neural network, and in the aspect that the target multichannel probability image is input into the preset depth regression network for calculation to obtain the target comprehensive quantitative data of the heart, the calculating unit 603 is specifically configured to:

Referring to fig. 7, fig. 7 is a schematic structural diagram of another biventricular quantification apparatus according to an embodiment of the present application, where the apparatus further includes a second obtaining unit 604, where the second obtaining unit 604 is configured to obtain a predetermined depth regression network, and in terms of obtaining the predetermined depth regression network, the second obtaining unit 604 is specifically configured to:

and repeating the step 2 to the step 5, and taking the N layers of convolutional neural networks as the preset deep regression network when the predicted value output by the N layers of convolutional neural networks is comprehensive quantized data.

Referring to fig. 8, fig. 8 is a schematic structural diagram of another dual ventricular quantization apparatus according to an embodiment of the present application, the apparatus further includes a third obtaining unit 605, the third obtaining unit 605 is configured to obtain a predetermined true value generating function, and in terms of obtaining the predetermined true value generating function, the third obtaining unit 605 is specifically configured to:

Optionally, in terms of acquiring the barycentric coordinate of the heart from the mask map by using the preset acquisition mode, the third acquisition unit 605 is specifically configured to:

Optionally, the apparatus is further specifically configured to:

resampling the target original image to obtain a resampled image;

Embodiments of the present application also provide a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute part or all of the steps of any one of the dual-chamber quantification methods as described in the above method embodiments.

Embodiments of the present application also provide a computer program product comprising a non-transitory computer readable storage medium storing a computer program, the computer program causing a computer to perform some or all of the steps of any one of the above-described method embodiments of bi-ventricular quantification methods.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present application is not limited by the order of acts described, as some steps may occur in other orders or concurrently depending on the application. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required in this application.

In the foregoing embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus may be implemented in other manners. For example, the above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one type of division of logical functions, and there may be other divisions when actually implementing, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of some interfaces, devices or units, and may be an electric or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may be implemented in the form of a software program module.

The integrated units, if implemented in the form of software program modules and sold or used as stand-alone products, may be stored in a computer readable memory. Based on such understanding, the technical solution of the present application may be substantially implemented or a part of or all or part of the technical solution contributing to the prior art may be embodied in the form of a software product stored in a memory, and including several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned memory comprises: various media capable of storing program codes, such as a usb disk, a read-only memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and the like.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable memory, which may include: flash memory disks, read-only memory, random access memory, magnetic or optical disks, and the like.

The foregoing detailed description of the embodiments of the present application has been presented to illustrate the principles and implementations of the present application, and the above description of the embodiments is only provided to help understand the method and the core concept of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A bi-ventricular quantification method, characterized in that the method comprises:

segmenting the target original image by adopting a preset segmentation network to obtain a target multichannel probability image, wherein the preset depth regression network comprises N layers of convolutional neural networks;

inputting the first calculation result into a second layer of the N-layer convolutional neural network for calculation to obtain a second calculation result, and inputting the calculation result of the (N-1) th layer into an Nth layer of the N-layer convolutional neural network for calculation to obtain target comprehensive quantitative data of the heart, wherein the target comprehensive quantitative data comprises the heart cavity diameter, the myocardial wall thickness and the ventricular area;

the method further comprises the step of obtaining a preset depth regression network, wherein the step of obtaining the preset depth regression network comprises the following steps:

2. The method of claim 1, further comprising obtaining a preset truth generating function, wherein obtaining the preset truth generating function comprises:

3. The method according to claim 2, wherein the obtaining the barycentric coordinates of the heart from the mask image by using a preset obtaining manner comprises:

acquiring a contour point set from the mask image through a first preset function;

4. The method of any one of claims 1 to 3, wherein the target multi-channel probability image comprises: background region images, left/right heart chamber images, and left myocardium images.

5. The method of claim 4, further comprising:

resampling the target original image to obtain a resampled image;

6. A bi-ventricular quantification device, the device comprising:

the segmentation unit is used for segmenting the target original image by adopting a preset segmentation network to obtain a target multichannel probability image, and the preset depth regression network comprises an N-layer convolutional neural network;

the calculation unit is used for inputting the target multichannel probability image into a first layer of the N-layer convolutional neural network for calculation to obtain a first calculation result, wherein N is a positive integer greater than 1;

the device further comprises a second obtaining unit, wherein the second obtaining unit is configured to obtain a preset depth regression network, and in terms of obtaining the preset depth regression network, the second obtaining unit is specifically configured to:

7. The apparatus according to claim 6, further comprising a third obtaining unit, configured to obtain a preset true value generating function, where in obtaining the preset true value generating function, the third obtaining unit is specifically configured to:

8. The apparatus according to claim 7, wherein in the acquiring of the barycentric coordinates of the heart from the mask map by using the preset acquisition mode, the third acquisition unit is specifically configured to:

9. The apparatus of any one of claims 6 to 8, wherein the target multi-channel probability image comprises: background region images, left/right heart chamber images, and left myocardium images.

10. The apparatus of claim 9, wherein the apparatus is further specifically configured to:

resampling the target original image to obtain a resampled image;

11. An electronic device, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to: performing the method of any one of claims 1 to 5.

12. A computer readable storage medium having computer program instructions stored thereon, which when executed by a processor implement the method of any one of claims 1 to 5.