CN114494252A

CN114494252A - Heart image processing method, device, equipment and storage medium

Info

Publication number: CN114494252A
Application number: CN202210365400.7A
Authority: CN
Inventors: 贺毅; 宋现涛; 田晋帆; 曹佳鑫; 孔慧慧; 雍婧雯
Original assignee: Beijing Friendship Hospital
Current assignee: Beijing Friendship Hospital
Priority date: 2022-04-02
Filing date: 2022-04-02
Publication date: 2022-05-13

Abstract

The application discloses a heart image processing method, a heart image processing device, a heart image processing apparatus and a storage medium. The method comprises the following steps: acquiring cardiac magnetic resonance film sequence images; the cardiac magnetic resonance film sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial sub-regions; and identifying the cardiac magnetic resonance film sequence images by using a trained machine learning model based on the cardiac layers contained in the cardiac magnetic resonance film sequence images so as to determine the state identification result of the myocardial subregion in each cardiac layer. According to the scheme, the target three-dimensional characteristics with time dimension, myocardium dimension and myocardium subregion dimension are obtained by performing characteristic processing on the acquired cardiac magnetic resonance film sequence image. The three-dimensional characteristics of the target are input into the machine learning model, so that accurate identification of the cardiac magnetic resonance film sequence images can be realized under the condition that the cardiac information is complete and continuous, and the state information of each myocardial subregion film sequence image in each myocardial layer can be accurately identified.

Description

Heart image processing method, device, equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a storage medium for processing a cardiac image.

Background

Cardiac magnetic resonance is an important tool for the imaging diagnosis of cardiovascular diseases. Stress myocardial perfusion imaging (stress myocardial perfusion) has obvious advantages in assessing problems such as myocardial ischemia and myocardial reserve function. The principle is that the coronary artery is dilated to the maximum extent by intravenous injection of vasodilator, and if the coronary artery has a stenosis degree of more than or equal to 50%, the myocardium in the blood supply area can have ischemic change, which is manifested as perfusion defect or local wall motion abnormality. However, in practical applications, there is a certain unpredictable risk of loading magnetic resonance myocardial perfusion, which is mainly reflected in that when a subject receives the loading magnetic resonance myocardial perfusion, a dangerous situation such as sudden cardiac arrest may occur due to the injection of a loading drug. Furthermore, performing stress magnetic resonance myocardial perfusion is an invasive diagnostic modality that is subject to additional physical trauma to the subject.

Disclosure of Invention

In view of the above, the present application provides a cardiac magnetic resonance cine image processing method, apparatus, device and storage medium that solve or partially solve the above-mentioned problems.

The embodiment of the application provides a heart image processing method, which comprises the following steps:

acquiring a cardiac magnetic resonance film sequence image; the cardiac magnetic resonance film sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial subregions;

and identifying the cardiac magnetic resonance film sequence images by using a trained machine learning model based on cardiac layers contained in the cardiac magnetic resonance film sequence images so as to determine a state identification result of the myocardial subregion in each cardiac layer.

Optionally, the acquiring cardiac magnetic resonance cine sequence images includes:

acquiring a layer film sequence image corresponding to each heart layer according to the time sequence of the cardiac cycle;

and obtaining the cardiac magnetic resonance film sequence image containing a plurality of film sequence images of the heart layers based on the corresponding hierarchical order of the heart layers.

Optionally, after obtaining the cardiac magnetic resonance cine sequence images including a plurality of the slice cine sequence images, further comprising:

performing region division on each layer of the cardiac magnetic resonance film sequence image to generate a myocardial subregion film sequence image;

and generating a time sequence feature vector corresponding to the myocardial subregion film sequence image according to the layer film sequence image corresponding to the heart layer, wherein each frame in the myocardial subregion film sequence image corresponds to one time point in the time sequence feature vector.

Optionally, the generating a time-series feature vector corresponding to the cine sequence images of the myocardial sub-region includes:

acquiring a preset mask image;

processing the cardiac magnetic resonance film sequence image by using the preset mask image, and extracting a film sequence image of a target myocardial subregion;

and generating a time sequence feature vector corresponding to the target myocardial subregion film sequence image according to the time sequence.

Optionally, before generating the time-series feature vector corresponding to the layer movie sequence image, the method further includes:

determining a plurality of myocardium corresponding to the target myocardial subregion film sequence image;

generating a target three-dimensional feature based on the hierarchical order, the myocardium and the temporal order.

Optionally, the generation manner of the preset mask image includes:

acquiring an original image of a mask;

labeling the film sequence images of the myocardial subregions of each mask in the original mask image according to the time sequence and the level sequence;

and generating the preset mask image.

Optionally, the machine learning model training mode includes:

acquiring a training sample image;

based on the training sample images, performing feature extraction processing according to the time sequence of the cardiac cycle to generate sample statistical features corresponding to the cine sequence images of the myocardial central muscle subregions;

marking the ischemia state corresponding to the statistical characteristics of the sample by using a sample label;

inputting the sample statistical features and the corresponding sample labels into the machine learning model for training.

The embodiment of the application provides a heart image processing device, which comprises:

the acquisition module is used for acquiring cardiac magnetic resonance film sequence images; the cardiac magnetic resonance sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial subregions;

and the identification module is used for identifying the cardiac magnetic resonance film sequence images by using a trained machine learning model based on the cardiac layers contained in the cardiac magnetic resonance film sequence images so as to determine the state identification result of the myocardial subareas in each cardiac layer.

An embodiment of the present application provides an electronic device, including: a memory and a processor; wherein the content of the first and second substances,

the memory is used for storing programs;

the processor, coupled with the memory, to execute the program stored in the memory to:

An embodiment of the present application provides a computer storage medium for storing a computer program, where the computer program enables a computer to implement the following method when executed:

According to the scheme provided by the embodiment of the application, a cardiac magnetic resonance film sequence image is acquired; the cardiac magnetic resonance film sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial subregions; and identifying the cardiac magnetic resonance film sequence images by using a trained machine learning model based on cardiac layers contained in the cardiac magnetic resonance film sequence images so as to determine a state identification result of the myocardial subregion in each cardiac layer. According to the scheme, the three-dimensional target features with time dimension, different anatomical structures and different feature types are obtained by performing feature processing on the acquired cardiac magnetic resonance sequence images. The target three-dimensional characteristics are input into a trained machine learning model, accurate identification of the cardiac magnetic resonance sequence images can be achieved under the condition that cardiac information is complete and continuous, and particularly state information of each myocardial subregion film sequence image in each myocardial layer can be accurately identified.

Drawings

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

fig. 1 is a schematic structural diagram of a cardiac image processing apparatus according to an exemplary embodiment of the present disclosure;

fig. 2 is a schematic flowchart of a cardiac image processing method according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram of image partitioning according to an embodiment of the present disclosure;

FIG. 4 is a schematic diagram of a process for masking a cine sequence of images of a myocardial subregion according to an embodiment of the present disclosure;

fig. 5 is a schematic structural diagram of a cardiac image processing apparatus according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

Before the technical solutions provided by the embodiments of the present application are described, a brief description of specific terms in this document will be provided.

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, 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 some embodiments of the present application, but not all 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 terminology used in the embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the examples of this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise, and "a" and "an" typically include at least two, but do not exclude the presence of at least one.

It should be understood that the term "and/or" as used herein is merely one type of association that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.

It is also noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a good or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such good or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a commodity or system that includes the element.

It should be noted that stress myocardial perfusion imaging (stress myocardial perfusion) has a significant advantage in assessing problems such as myocardial ischemia and myocardial reserve function. The principle is that the coronary artery is dilated to the maximum extent by intravenous injection of vasodilator, and if the coronary artery has a stenosis degree of more than or equal to 50%, the myocardium in the blood supply area can have ischemic change, which is manifested as perfusion defect or local wall motion abnormality. However, in practical applications, there is a certain unpredictable risk of loading magnetic resonance myocardial perfusion, which is mainly reflected in that when a subject receives the loading magnetic resonance myocardial perfusion, a dangerous situation such as sudden cardiac arrest may occur due to the injection of a loading drug. Furthermore, performing stress magnetic resonance myocardial perfusion is an invasive diagnostic modality that is subject to additional physical trauma to the subject. Therefore, a technical solution for accurately classifying and identifying the cardiac magnetic resonance image without invasive image acquisition methods such as myocardial perfusion is needed.

The execution main body of the method provided by the embodiment of the application can be one device or a plurality of devices. The device may be, but is not limited to, a device integrated on any terminal device such as a smart phone, a tablet computer, a PDA (Personal Digital Assistant), a smart television, a laptop computer, a desktop computer, and an intelligent wearable device. The cardiac image processing apparatus is not particularly limited, but refers to an apparatus that can achieve accurate classification recognition of cardiac images without causing visible wounds on individuals. For example, fig. 1 is a schematic structural diagram of a cardiac image processing apparatus according to an embodiment of the present application. As can be seen from fig. 1, the cardiac image processing device comprises a display, a processor, and a magnetic resonance device. The receiver can carry out non-invasive image acquisition through nuclear magnetic resonance equipment, sends the heart magnetic resonance film sequence image that gathers through nuclear magnetic resonance equipment to the processor, and machine learning model that trains in this processor carries out classification recognition to heart magnetic resonance film sequence image to confirm the state recognition result of each heart subregion of receiver heart. A specific method of performing cardiac image processing based on this cardiac image processing apparatus will be specifically exemplified in each of the embodiments described below.

Fig. 2 is a schematic flowchart of a cardiac image processing method according to an embodiment of the present disclosure. As can be seen from fig. 2, the method comprises the following steps:

201: acquiring a cardiac magnetic resonance film sequence image; the cardiac magnetic resonance film sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial subareas.

202: and identifying the cardiac magnetic resonance film sequence images by using a trained machine learning model based on cardiac layers contained in the cardiac magnetic resonance film sequence images so as to determine a state identification result of the myocardial subregion in each cardiac layer.

In practical applications, when performing the mri-related image acquisition, the heart maintains periodic diastole and systole, a diastolic and systolic heart beat cycle being referred to as the cardiac cycle in this application. In order to obtain more comprehensive heart-related image information, image acquisition is performed according to the cardiac cycle of the heart, so that cardiac magnetic resonance cine sequence images can be acquired. In addition, in order to more accurately distinguish the heart images, the heart is subjected to layered image acquisition to acquire a plurality of heart layers.

The state recognition result here includes a state recognition result in which the myocardial region is in an ischemic state and a state recognition result in which the myocardial region is in a non-ischemic state. The identification and segmentation of the myocardial sub-region in the cardiac magnetic resonance cine sequence images may be performed by a machine learning model, or may be performed by a worker.

Specifically, a cardiac segment acquisition technique or a real-time imaging technique (such as cardiac magnetic resonance) can be used to acquire bright blood images of multiple phases of the same slice continuously in one cardiac cycle, and magnetic resonance cine sequence images of the heart undergoing rhythmic contraction and relaxation processes are displayed. Can be used for evaluating cardiac function and myocardial motion state. The cardiac magnetic resonance film sequence image is a sequence image of a plurality of layers and a plurality of frames which are acquired by the cardiac magnetic resonance imaging equipment and contain the systolic phase and the diastolic phase after multi-azimuth scanning positioning.

Specifically, the number of heart layers may vary from 9 to 11, depending on the selected layer thickness and the size of the heart of the subject. It should be noted that regardless of how many slices are acquired, it is ensured that the acquired sequence images cover the heart region from the base to the apex. The number of frames acquired per heart layer is typically 25 frames. In an implementable approach, to ensure the accuracy of the acquired cardiac magnetic resonance sequence images, the cardiac magnetic resonance imaging apparatus automatically acquires images at 25 time points (i.e., 25 frames) of the cardiac cycle based on the electrocardiogram of the subject.

The machine learning model is used for classifying and identifying whether the myocardial subregion is ischemic or not by analyzing the cardiac magnetic resonance film sequence images, and the machine learning model can adopt algorithms such as a support vector machine, a decision tree analysis algorithm, a random forest algorithm and the like.

When the trained machine learning model is used for recognizing the cardiac magnetic resonance film sequence image, each myocardial subregion obtained after image division needs to be respectively subjected to targeted recognition. Therefore, the identification accuracy can be effectively improved. The acquisition method of the myocardial subregion and the identification method of the myocardial subregion will be specifically exemplified in the following embodiments.

In one or more embodiments of the present application, the acquiring cardiac magnetic resonance cine sequence images includes: and acquiring a layer film sequence image corresponding to each heart layer according to the time sequence of the heart cycle. And generating the cardiac magnetic resonance film sequence image based on the corresponding hierarchical order of the plurality of cardiac layers.

As previously described, each heart layer includes 25 frames of images reflecting the entire cardiac cycle, with the 25 frames of images being taken as a layer cine sequence of images of the heart layer. For subsequent recognition and image processing, each layer cine sequence image may be associated with corresponding cardiac layer identifiers that are labeled in a hierarchical order (e.g., layer 1, layer 2 … … 9). Furthermore, a plurality of heart layers (for example, 9 layers) are combined to obtain a cardiac magnetic resonance cine sequence image having coherence, and the entire condition of the heart in the cardiac cycle can be known from the cardiac magnetic resonance cine sequence image.

In one or more embodiments of the present application, after obtaining the cardiac magnetic resonance cine sequence images, the method further includes: and carrying out region division on the cardiac magnetic resonance film sequence image to generate a myocardial subregion film sequence image. And extracting features from the cine sequence images of the myocardial subareas according to the cine sequence images of the cardiac layer corresponding to the cardiac layer, and constructing and obtaining corresponding time sequence feature vectors.

Fig. 3 is a schematic diagram of image division according to an embodiment of the present application. As can be seen from fig. 3, the images in the cine-mri sequence of cardiac magnetic resonance are divided into regions, so as to obtain 12 cine-mri sequence images of cardiac subregions. Specifically, the sequence images include the anterior medial and lateral sides, the posterior medial and lateral sides, the inferior lateral side and the lateral side, and the total of 12 cine sequences of myocardial subregions. Fig. 2 is a schematic view of a certain heart layer. As described above, each heart layer corresponds to the acquired cine sequence images (25 frames of images), and thus, cine sequence images of sub-regions formed of 25 frames of images of each myocardial sub-region per heart layer can be obtained. When performing classification recognition of cardiac magnetic resonance cine sequence images, it is necessary to perform recognition based on slice cine sequence images of sub-regions corresponding to the respective myocardial sub-regions. In practical applications, it is found that the myocardial ischemia states of the inner and outer regions are different even at the same anterior wall or anterior space, and therefore, in order to more accurately identify the cardiac mri cine sequence images, it is necessary to more finely divide the myocardial sub-regions. After the region division mode is adopted, each myocardial subregion film sequence image in each myocardial layer is obtained, and then feature extraction is carried out on a plurality of frames of images contained in each myocardial subregion film sequence image to obtain a corresponding time sequence feature vector.

In one or more embodiments of the present application, the generating a time-series feature vector corresponding to the cine-sequences of myocardial subregions includes: acquiring a preset mask image; processing the film sequence images of the myocardial subregions by utilizing the preset mask images to extract the film sequence images of the target myocardial subregions; and generating a time sequence feature vector corresponding to the target myocardial subregion film sequence image according to the time sequence.

In practical applications, because the cardiac mri cine sequence images contain a large amount of content, the real region of interest needs to be distinguished in a certain way. For example, the cine sequence images of the cardiac magnetic resonance cine sequence may be masked with mask images to obtain cine sequence images of the myocardial subregion of interest. Fig. 4 is a schematic diagram of a process of masking a cine sequence of myocardial subregions according to an embodiment of the present disclosure. As can be seen from fig. 4, before the masking process is performed, the pixel values of the pixel points in the cine sequence images of the myocardial subregions are obtained. Since some of the regions, for example, blood in the ventricle, do not fall within the observation range, but are also displayed in the image and have a certain pixel value. As can be seen from fig. 4, in the preset mask image, only the myocardial subregion of major interest is marked as 1, and the other regions are all marked as 0, so that after the and operation, the cine sequence images of the target myocardial subregion are obtained.

In one or more embodiments of the present application, before generating the time-series feature vector corresponding to the layer movie sequence image, the method further includes: and determining a plurality of myocardium corresponding to the target myocardial subregion film sequence image. Generating a target three-dimensional feature based on the hierarchical order, the myocardium and the temporal order. Representing different kinds of features computed over different myocardium, over different number of frames (represented by time sequence).

And stacking each extracted feature of all the myocardial sub-regions into a 3-dimensional feature matrix according to dimensions of different anatomical structures, different frames and different feature types, and then forming a vector by using feature values extracted from the same myocardial region on the same layer according to different frames so as to create a time series feature vector.

On the basis of the created time series feature vector, on one hand, time domain operation is performed on the time series feature vector, including but not limited to operation such as averaging, variance, difference (mainly including feature difference values corresponding to the systolic period and the diastolic period) and the like on the time series feature vector, so as to obtain time domain features. On the other hand, the time series feature vector is subjected to Fourier transform to obtain a frequency domain feature signal, and then operations such as mean variance calculation and the like are respectively carried out on the low frequency band and the high frequency band of the frequency domain feature signal to obtain the frequency domain feature.

After the target three-dimensional features are obtained, further screening of the features is needed, for example, for the mean feature set of the time domain features and the frequency domain features, a feature screening method based on statistical analysis, a feature screening method based on elastic network regression analysis, and the like are used to screen key features.

Carrying out feature screening on variance feature sets of the time domain and frequency domain derived features by using methods for evaluating variance such as chi-square test, and then applying a support vector machine to construct a model; and aiming at the differential feature set of the time domain derived features, applying a double differential feature screening method to the differential feature set, and then applying a random forest to construct a model.

In one or more embodiments of the present application, the generating manner of the preset mask image includes: acquiring an original image of a mask; and labeling the film sequence images of the mask myocardial subregions in the original mask image according to the time sequence and the hierarchical sequence. And generating the preset mask image.

It should be noted that the mask for each layer and each frame is obtained by a staff member by using the medical labeling tool ITK-SNAP to delineate according to the myocardial structure. In order to distinguish each segment of the myocardial sub-region, the staff uses 12 colors representing different mask labels to outline the myocardial sub-region, that is, the pixel values of the myocardial sub-region mask include 12, for example, the numbers 1 to 12 are used as the mask pixel values of 12 segments of the myocardial sub-region, and one sub-mask of the pixel values corresponds to one segment of the myocardial sub-region. For easy distinction, the myocardial subareas have strict one-to-one correspondence with the mask label colors.

The original mask image includes a plurality of heart layers, and the heart layers are ordered in a hierarchical order. And each heart layer contains a plurality of frame images, which are ordered in a time sequence. In practical application, in order to obtain a better image classification and identification effect, when the film sequence images of the mask myocardial subregions are labeled, all the mask myocardial subregions can be labeled more comprehensively, so that a preset mask image containing all heart layers and frame images of all myocardial subregions in each heart layer is obtained.

For example, the mask of each frame in the 3-dimensional mask of each layer is divided into 12 3-dimensional sub-masks, each sub-mask has only the cardiac sub-region position pixel on the corresponding frame in the time dimension set to 1, and other positions set to 0, that is, 12 3-dimensional sub-masks can be obtained according to the 3-dimensional mask of each frame in a layer, and if 25 frames of a layer all outline 12 segments of sub-regions, the layer can obtain 25 × 12 3-dimensional sub-masks as described above. And sending the 3-dimensional image and the 3-dimensional sub-mask in the cardiac magnetic resonance film sequence into a feature extractor to obtain the image omics feature of the myocardial subregion.

In one or more embodiments of the present application, the machine learning model training method includes: training sample images are acquired. Based on the training sample images, performing feature processing according to the time sequence of the cardiac cycle to generate sample statistical features corresponding to the cine sequence images of the myocardial central muscle subregions. And marking the ischemia state corresponding to the statistical characteristics of the sample by using a sample label. Inputting the sample statistical features and the corresponding sample labels into the machine learning model for training.

In practical application, the training samples are generated as follows: cardiac magnetic resonance cine sequence images of a plurality of specimen providers are acquired having 6-9 slices of the heart, and then 12 different myocardial subregions are segmented by the staff for each cardiac slice. And then taking the cardiac magnetic resonance film sequence image corresponding to each myocardial subregion and a label of whether the myocardial subregion is ischemic or not as a training sample. For example, the label of the ischemic myocardial subregion is 1, and the label of the ischemic myocardial subregion is 0, and of course, the label corresponding value may be adjusted as needed in actual application, which is only used for illustration and does not limit the technical solution of the present application.

In practical application, the sample statistical characteristics of the training sample are as follows: extracting the iconomics characteristics of the 12 myocardial subregions of each layer and each frame by using the preset mask image which is drawn according to professional knowledge and comprises the 12 myocardial subregions; and selecting and processing the characteristics of the image omics according to the preselected characteristic types. And creating the selected features into a time sequence feature vector according to a time sequence, and further processing the time sequence feature vector to obtain the statistical features of each segment of the myocardial subregion of each layer. The label of the training sample is given by the staff according to the myocardial perfusion image and the professional knowledge, and the label corresponding to the label of each segment of the myocardial subregion of each layer of the corresponding cardiac magnetic resonance film sequence image (for example, ischemia: 1, and non-ischemia: 0) is given. And during training, training the initial machine learning model by using the statistical characteristics and the labels, and obtaining the final trained machine learning model when the classification effect meets the set requirement. In an embodiment of the present application, the machine learning model may be a classification model.

The initial classification model gives a prediction label according to the input statistical characteristics of the myocardial subregion, and performs difference analysis on the prediction label and a real label of a training sample by combining the real label of the myocardial ischemia state given by a worker according to the myocardial perfusion image, and feeds an evaluation result back to the initial classification model, and the model updates model parameters according to the feedback so as to train the machine learning model. The initial classification model improves the discrimination capability, namely improves the classification accuracy, and is continuously updated iteratively in such a way until the difference analysis result meets the set requirement, so that a final classification model (namely a pre-trained model) is obtained. Therefore, a classification state recognition result of the classification model which can judge whether a certain myocardial subregion is ischemic or not according to the cardiac magnetic resonance film sequence image is obtained.

Based on the embodiment, the cardiac magnetic resonance film sequence image is obtained; the cardiac magnetic resonance film sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial subregions; and identifying the cardiac magnetic resonance film sequence images by using a trained machine learning model based on cardiac layers contained in the cardiac magnetic resonance film sequence images so as to determine a state identification result of the myocardial subregion in each cardiac layer. According to the scheme, the target three-dimensional characteristics with anatomical structure dimension (such as myocardium dimension), time dimension and characteristic type dimension (such as myocardium sub-region dimension) are obtained by performing characteristic processing on the acquired cardiac magnetic resonance film sequence images. The target three-dimensional characteristics are input into a trained machine learning model, accurate identification of the cardiac magnetic resonance film sequence images can be achieved under the condition that the cardiac information is complete and continuous, and particularly state information of each myocardial subregion film sequence image in each myocardial layer can be accurately identified.

When the cardiac magnetic resonance film sequence images are classified and identified, the classified identification is respectively carried out according to a plurality of myocardial subareas contained in each heart layer obtained by division. Moreover, in order to obtain a more accurate identification effect, the cine sequence images of the myocardial sub-regions corresponding to each myocardial sub-region in each heart layer all include a plurality of frames of images, and the classification result (for example, whether ischemia exists) of the current myocardial sub-region is identified based on the plurality of frames of images.

Based on the same idea, the embodiment of the present application further provides a cardiac image processing apparatus. Fig. 5 is a schematic structural diagram of a cardiac image processing apparatus according to an embodiment of the present application. As can be seen from fig. 5, the apparatus comprises:

an acquisition module 51, configured to acquire cardiac magnetic resonance cine sequence images; the cardiac magnetic resonance film sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial subareas.

The identifying module 52 is configured to identify the cardiac magnetic resonance cine sequence images by using a trained machine learning model based on cardiac layers included in the cardiac magnetic resonance cine sequence images, so as to determine a state identification result of the myocardial sub-region in each cardiac layer.

Optionally, the obtaining module 51 is further configured to obtain a slice cine sequence image corresponding to each heart slice according to a time sequence of the cardiac cycle; and obtaining the cardiac magnetic resonance film sequence image containing a plurality of film sequence images of the heart layers based on the corresponding hierarchical order of the heart layers.

Optionally, the obtaining module 51 is further configured to perform region division on each layer of the cardiac magnetic resonance cine sequence images to generate cine sequence images of myocardial subregions;

Optionally, the obtaining module 51 is further configured to obtain a preset mask image;

processing the cardiac magnetic resonance film sequence image by using the preset mask image, and extracting to obtain a film sequence image of a target myocardial subregion;

and generating a time sequence feature vector corresponding to the layer film sequence image according to the time sequence.

Optionally, the obtaining module 51 is further configured to determine a plurality of myocardium corresponding to the cine sequence images of the target myocardial sub-region; generating a target three-dimensional feature based on the hierarchical order, the myocardium and the temporal order.

Optionally, the system further includes a generating module 53, configured to acquire an original mask image;

and generating the preset mask image.

Optionally, a training module 54 is further included for acquiring training sample images;

Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present application. As shown in fig. 6, the electronic apparatus includes: a memory 61 and a processor 62; wherein the content of the first and second substances,

the memory 61 is used for storing programs;

the processor 62, coupled with the memory, is configured to execute the program stored in the memory to:

The memory 61 described above may be configured to store other various data to support operations on the computing device. Examples of such data include instructions for any application or method operating on a computing device. The memory 61 may be implemented by any type or combination of volatile or non-volatile memory devices, such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.

The processor 62 may also implement other functions besides the above functions when executing the program in the memory 61, and specifically refer to the description of the foregoing embodiments.

Further, as shown in fig. 6, the electronic device further includes: a display 63, a power supply component 64, a communication component 65, and the like. Only some of the components are schematically shown in fig. 6, and it is not meant that the electronic device comprises only the components shown in fig. 6.

Accordingly, embodiments of the present application also provide a computer-readable storage medium storing a computer program, which, when executed by a computer, can implement the steps or functions of the cardiac image processing method provided in the foregoing embodiments.

The above-described embodiments of the apparatus are merely illustrative, and 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 position, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. Based on the understanding, the above technical solutions substantially or otherwise contributing to the prior art may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the various embodiments or some parts of the embodiments.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.

Claims

1. A method of cardiac image processing, the method comprising:

2. The method of claim 1, wherein the acquiring cardiac magnetic resonance cine sequence images comprises:

3. The method of claim 2, wherein obtaining the cardiac magnetic resonance cine sequence images comprising a plurality of the slice cine sequence images further comprises:

4. The method of claim 3, wherein generating the time-series feature vectors corresponding to the cine-series of myocardial subregions comprises:

acquiring a preset mask image;

5. The method of claim 4, wherein before generating the time-series feature vectors corresponding to the cine-series images of the target myocardial subregion, further comprising:

6. The method according to claim 4, wherein the preset mask image is generated in a manner comprising:

acquiring an original image of a mask;

and generating the preset mask image.

7. The method of claim 1, wherein the machine learning model training mode comprises:

acquiring a training sample image;

8. An image processing apparatus, characterized in that the apparatus comprises:

the acquisition module is used for acquiring cardiac magnetic resonance film sequence images; the cardiac magnetic resonance film sequence image comprises a plurality of cardiac layers, and each cardiac layer comprises a plurality of myocardial subregions;

9. An electronic device, comprising: a memory and a processor; wherein the content of the first and second substances,

the memory is used for storing programs;

10. A computer storage medium storing a computer program which, when executed by a computer, causes the computer to perform the method of: