Myocardial injury information analysis method and device based on multi-sequence magnetic resonance imaging

Abstract

The invention discloses a myocardial damage information analysis method and device based on multi-sequence magnetic resonance imaging, and the method comprises the steps: obtaining historical multi-sequence magnetic resonance imaging data generated by a myocardial tissue of a to-be-recognized object based on nuclear magnetic resonance, the historical multi-sequence magnetic resonance imaging data comprising an MTT sequence, a VNE image, a CEST sequence and an OS-BOLD sequence; inputting the historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model so as to analyze the historical multi-sequence magnetic resonance imaging data; and outputting multiple types of myocardial tissue regions corresponding to the historical multi-sequence magnetic resonance imaging data, and respectively calculating quantitative parameters of the MTT sequence, the VNE image, the CEST sequence and the OS-BOLD sequence in the multiple types of myocardial tissue regions to obtain myocardial damage information of the to-be-identified object. Therefore, by adopting the embodiment of the invention, the accuracy of the myocardial injury information is improved.

Description

Technical Field

[0001] The present application relates to the fields of smart medical technology and computer technology, and in particular to a method and device for analyzing myocardial injury information based on multi-sequence magnetic resonance imaging. Background Art

[0002] In the diagnosis and treatment of cardiovascular disease, magnetic resonance imaging (MRI) has become an important tool for assessing cardiac structure and function due to its high resolution and superior soft tissue contrast. In particular, in the analysis of myocardial injury, MRI can provide detailed information about the infarcted area.

[0003] Currently, commonly used imaging sequences in cardiac MRI include multi-task tissue characterization (MTT) technology, virtual late gadolinium enhancement (VNE) images, chemical exchange saturation transfer (CEST) sequences, and blood oxygenation level-dependent (OS-BOLD) sequences. These techniques can provide multidimensional information about the myocardium, such as tissue characteristics, metabolic status, and oxygenation function.

[0004] Although existing technologies have made certain progress in cardiac MRI imaging, the information integration between different imaging sequences is insufficient, which reduces the accuracy of myocardial injury information. Summary of the Invention

[0005] The embodiments of this application provide a method and apparatus for analyzing myocardial injury information based on multi-sequence magnetic resonance imaging. To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is provided below. This summary is not intended to be a comprehensive review, identify key or important elements, or delineate the scope of protection for these embodiments. Its sole purpose is to present some concepts in a simplified form, serving as a prelude to the detailed description that follows.

[0006] In a first aspect, an embodiment of the present application provides a method for analyzing myocardial injury information based on multi-sequence magnetic resonance imaging, the method comprising:

[0007] Acquire historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the subject to be identified based on nuclear magnetic resonance, the historical multi-sequence magnetic resonance imaging data including MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence;

[0008] Inputting historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model to analyze the historical multi-sequence magnetic resonance imaging data;

[0009] The multiple myocardial tissue regions corresponding to historical multi-sequence MRI data are output, and the quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in the multiple myocardial tissue regions are calculated to obtain the myocardial damage information of the object to be identified.

[0010] Optional, pre-trained myocardial injury information analysis model includes convolutional neural network, visual transformer, codec structure and feature fusion module, region division and quantitative parameter calculation module; MTT sequence includes T2* image, T1 image, T1ρ image and T2 image;

[0011] Analyze historical multi-sequence MRI data, including:

[0012] The convolutional neural network extracts multi-scale features of preset sequence types for VNE images, T2* images, CEST sequences, and OS-BOLD sequences to obtain multi-scale features of the images.

[0013] The visual transformer performs multi-sequence deep feature fusion on the multi-scale features of the image, and analyzes the correlation and difference of image features between different sequences through the multi-head self-attention mechanism preset in the visual transformer to obtain the main input sequence features;

[0014] The convolutional neural network extracts multi-scale features of preset sequence types from T1 images, T1ρ images, and T2 images to obtain auxiliary input sequence features;

[0015] The encoder-decoder structure fuses and reduces the dimensions of the main input sequence features and the auxiliary input sequence features in the feature fusion module to obtain reduced-dimensional features.

[0016] A convolutional neural network combined with skip connections uses multi-scale image features and dimensionality reduction features to automatically identify subregions of myocardial tissue damage, obtaining regions of interest corresponding to VNE images, CEST sequences, and OS-BOLD sequences.

[0017] The regional division and quantitative parameter calculation module generates multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data based on the region of interest, and calculates the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in multiple types of myocardial tissue regions.

[0018] Optionally, based on the region of interest, multiple types of myocardial tissue regions corresponding to historical multi-sequence MRI data are generated, and quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in the multiple types of myocardial tissue regions are calculated, including:

[0019] Based on the region of interest, the hemorrhage area, infarct area, surrounding border area and normal tissue area were divided as multiple types of myocardial tissue areas;

[0020] The Cr-CEST metabolic value, which reflects the myocardial metabolic state, and the OS-SI value, which reflects the myocardial oxygenation state, were calculated in the hemorrhagic area, infarct area, surrounding border area, and normal tissue area using MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence.

[0021] Cr-CEST metabolic values and OS-SI values were used as quantitative parameters for MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in multiple types of myocardial tissue regions.

[0022] Optionally, the visual transformer includes a sequence embedding layer, a position embedding layer, and multiple self-attention layers;

[0023] Multi-sequence deep feature fusion is performed on the multi-scale features of the image to analyze the correlation and difference of image features between different sequences through the multi-head self-attention mechanism, and the main input sequence features are obtained, including:

[0024] The sequence embedding layer maps the multi-scale features of the image into low-dimensional feature vectors to capture the similarity between pixels and obtain a multi-scale embedding vector;

[0025] The position embedding layer extracts the relative position features of different pixel positions from the multi-scale embedding vector to obtain multi-scale position features;

[0026] The multi-scale position features are input into multiple self-attention layers, and different preset semantic tokens are collected for global classification to obtain the main input sequence features.

[0027] Optionally, before obtaining historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the subject to be identified, the method further includes:

[0028] Acquiring a nuclear magnetic resonance image of the myocardial tissue of the subject to be identified;

[0029] Using an inversion pulse, a T2 preparation pulse, and a T1ρ preparation pulse to generate different T1, T2, and T1ρ contrasts on a nuclear magnetic resonance image, thereby obtaining a first image sequence;

[0030] Reconstructing a multi-echo image using data collected from myocardial tissue during the diastole of a target cardiac cycle, and performing a second image processing on the multi-echo image to obtain a second image sequence; the first image sequence and the second image sequence include images of multiple dimensions of T1 / T2 / T2* / T1ρMapping;

[0031] Use T1 / T2 / T2* / T1ρMapping images in multiple dimensions as MTT sequences;

[0032] The MTT sequence is processed by a preset generative adversarial network to obtain a VNE image;

[0033] By using a short pulse of saturation signal, the preset chemical groups in the myocardial tissue are imaged separately, so as to fit the CEST sequence by measuring the creatine content in the myocardial tissue;

[0034] The signal changes caused by hyperventilation and breath holding of the subject to be identified are monitored through a carbon dioxide mask inhalation device, so as to fit the OS-BOLD sequence through the signal changes.

[0035] Optionally, a pre-trained myocardial injury information analysis model is generated by following the steps below:

[0036] Collecting historical cardiac magnetic resonance data of hemorrhagic myocardial infarction model subjects and non-hemorrhagic myocardial infarction model subjects, wherein the historical cardiac magnetic resonance data are acquired at multiple preset time points;

[0037] Create a network model that combines convolution and transformer structures. The network model includes a convolutional neural network, a visual transformer, an encoder-decoder structure and feature fusion module, and a region division and quantitative parameter calculation module.

[0038] Label the regions of interest (ROIs) of historical cardiac magnetic resonance data to obtain model training samples for segmentation tasks.

[0039] A pre-trained myocardial injury information analysis model is generated based on the model training samples and the network model.

[0040] Optionally, the historical cardiac magnetic resonance data includes historical VNE images, historical T2* images, historical CEST images, and historical OS-BOLD images;

[0041] Regions of interest (ROIs) are annotated on historical cardiac MRI data to obtain model training samples for segmentation tasks, including:

[0042] Regions of interest (ROIs) of the left ventricle, infarct, and hemorrhage areas were annotated on historical VNE images;

[0043] Marking of regions of interest in the hemorrhage area on historical T2* images;

[0044] The regions of interest of the left ventricle and infarct area were marked on historical CEST images;

[0045] Regions of interest (ROIs) of the left ventricle and infarct area were annotated on historical OS-BOLD images;

[0046] The above labeled results are used as model training samples for the segmentation task.

[0047] Optionally, a pre-trained myocardial injury information analysis model is generated based on the model training samples and the network model, including:

[0048] Use the Dice loss function as the loss function for the segmentation task for each preset label category;

[0049] The loss function is integrated into the network model to obtain the myocardial injury information analysis model;

[0050] Input the model training samples into the myocardial injury information analysis model and output the loss value of the model;

[0051] When the loss value reaches the minimum, a pre-trained myocardial injury information analysis model is generated.

[0052] Optionally, the method further includes:

[0053] When the loss value has not reached the minimum, the loss value is back-propagated to update the model parameters of the network model, and the step of inputting the historical multi-sequence magnetic resonance imaging data into the pre-trained myocardial injury information analysis model is continued until the loss value reaches the minimum.

[0054] In a second aspect, an embodiment of the present application provides a myocardial injury information analysis device based on multi-sequence magnetic resonance imaging, the device comprising:

[0055] An image acquisition module is used to acquire historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the object to be identified based on nuclear magnetic resonance, where the historical multi-sequence magnetic resonance imaging data includes MTT sequence, VNE image, CEST sequence and OS-BOLD sequence;

[0056] a data input module, configured to input historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model to analyze the historical multi-sequence magnetic resonance imaging data;

[0057] The result output module is used to output multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data, and to calculate the quantitative parameters of MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in multiple types of myocardial tissue regions to obtain myocardial damage information of the object to be identified.

[0058] The technical solutions provided by the embodiments of the present application may have the following beneficial effects:

[0059] In an embodiment of the present application, by integrating multi-sequence magnetic resonance imaging data of MTT sequence, VNE image, CEST sequence and OS-BOLD sequence, combined with a pre-trained deep learning model, high-precision identification and quantitative analysis of myocardial damage areas are achieved. Since multi-sequence magnetic resonance imaging data provides multi-dimensional information for the analysis of myocardial damage, the model uses a convolutional neural network to extract multi-scale features specific to the sequence type, uses a visual transformer to perform multi-sequence deep feature fusion, and explores the correlation and difference of image features between different sequences through a multi-head self-attention mechanism, thereby improving the accuracy of myocardial damage information and reducing dependence on subjective judgment of experts.

[0060] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0062] Figure 1 This is a schematic diagram of a method flow of a myocardial injury information analysis method based on multi-sequence magnetic resonance imaging provided in an embodiment of the present application;

[0063] Figure 2 This is a schematic diagram of a network architecture of a network model provided in an embodiment of the present application;

[0064] Figure 3 This is a flow chart of a model training method provided in an embodiment of the present application;

[0065] Figure 4 This is a schematic structural diagram of a myocardial injury information analysis device based on multi-sequence magnetic resonance imaging provided by the present application;

[0066] Figure 5 This is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0067] The following description and the drawings sufficiently illustrate specific embodiments of the application to enable those skilled in the art to practice them.

[0068] It should be clear that the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0069] When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of devices and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0070] In the description of this application, it should be understood that the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances. In addition, in the description of this application, unless otherwise specified, "multiple" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the previous and subsequent associated objects are in an "or" relationship.

[0071] The present application provides a method and device for analyzing myocardial injury information based on multi-sequence magnetic resonance imaging to solve the problems existing in the related technical problems. In the embodiment of the present application, by integrating multi-sequence magnetic resonance imaging data of MTT sequence, VNE image, CEST sequence and OS-BOLD sequence, combined with a pre-trained deep learning model, high-precision recognition and quantitative analysis of myocardial injury areas are achieved. Since multi-sequence magnetic resonance imaging data provides multi-dimensional information for the analysis of myocardial injury, the model uses a convolutional neural network to extract multi-scale features specific to the sequence type, uses a visual transformer to perform multi-sequence deep feature fusion, and explores the correlation and difference of image features between different sequences through a multi-head self-attention mechanism, thereby improving the accuracy of myocardial injury information and reducing dependence on subjective judgment of experts. The following exemplary embodiments are used for detailed description.

[0072] The following will be combined with the Figure 1 -Attached Figure 3 This article details the multi-sequence magnetic resonance imaging-based myocardial injury information analysis method provided in the embodiments of this application. This method can be implemented using a computer program and run on a multi-sequence magnetic resonance imaging-based myocardial injury information analysis device based on the von Neumann architecture. This computer program can be integrated into an application or run as a standalone tool application.

[0073] See Figure 1 , provides a flow chart of a method for analyzing myocardial injury information based on multi-sequence magnetic resonance imaging for an embodiment of the present application. Figure 1 As shown, the method of the embodiment of the present application may include the following steps:

[0074] S101, acquiring historical multi-sequence magnetic resonance imaging data of myocardial tissue of a subject to be identified, generated based on nuclear magnetic resonance, the historical multi-sequence magnetic resonance imaging data including an MTT sequence, a VNE image, a CEST sequence, and an OS-BOLD sequence;

[0075] Among them, the MTT sequence is a multi-task tissue feature quantitative imaging (MTT) technology, which can obtain T1 / T2 / T2* / T1ρMapping images in multiple dimensions through a short period of free breathing for quantitative analysis of myocardial tissue. The VNE image is a virtual late gadolinium enhancement (VNE) image, which can provide information on the myocardial infarction area. The chemical exchange saturation transfer (CEST) sequence is a magnetic resonance imaging technology that can non-invasively detect molecules in the tissue body, such as proteins, peptides, creatine, etc., which is of great significance for the early diagnosis and treatment monitoring of various diseases. The oxygen-sensitive blood oxygen level-dependent (OS-BOLD) sequence evaluates myocardial oxygenation function by monitoring the signal changes caused by hyperventilation and breath holding through a carbon dioxide mask inhalation device.

[0076] In an embodiment of the present application, the specific process of generating historical multi-sequence magnetic resonance imaging data includes: obtaining a nuclear magnetic resonance image of the myocardial tissue of the object to be identified; using a flip pulse, a T2 preparation pulse and a T1ρ preparation pulse to generate different T1, T2 and T1ρ contrasts on the nuclear magnetic resonance image to obtain a first image sequence; using data collected from the myocardial tissue during the diastole period of the target cardiac cycle to reconstruct a multi-echo image, and performing a second image processing on the multi-echo image to obtain a second image sequence; the first image sequence and the second image sequence include images of multiple dimensions of T1 / T2 / T2* / T1ρMapping; using the images of multiple dimensions of T1 / T2 / T2* / T1ρMapping as an MTT sequence; processing the MTT sequence through a preset generative adversarial network to obtain a VNE image; using a short pulse saturation signal to separately image the preset chemical groups in the myocardial tissue, so as to fit the CEST sequence by measuring the creatine content in the myocardial tissue; monitoring the signal changes caused by hyperventilation and breath holding of the object to be identified through a carbon dioxide mask inhalation device, so as to fit the OS-BOLD sequence through the signal changes.

[0077] For example, MRI equipment can be used to perform multiple imaging sequences on the subject, including MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence. These sequences will provide detailed information about the characteristics of myocardial tissue, metabolic status, and oxygenation function.

[0078] S102, inputting the historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model to analyze the historical multi-sequence magnetic resonance imaging data;

[0079] Among them, the pre-trained myocardial injury information analysis model is a model trained using machine learning or deep learning technology.

[0080] The MTT sequence includes T2* images, T1 images, T1ρ images, and T2 images. The pre-trained myocardial injury information analysis model includes a convolutional neural network, a visual transformer, an encoder-decoder structure and feature fusion module, and a region division and quantitative parameter calculation module.

[0081] In some embodiments of the present application, the specific process of analyzing historical multi-sequence magnetic resonance imaging data includes: a convolutional neural network performs multi-scale feature extraction of a preset sequence type on VNE images, T2* images, CEST sequences, and OS-BOLD sequences to obtain multi-scale features of the image; a visual transformer performs multi-sequence deep feature fusion on the multi-scale features of the image, and analyzes the correlation and difference of image features between different sequences through the multi-head self-attention mechanism preset inside the visual transformer to obtain the main input sequence features; a convolutional neural network performs multi-scale feature extraction of a preset sequence type on T1 images, T1ρ images, and T2 images to obtain auxiliary input sequences. Features; the encoder-decoder structure fuses and reduces the dimensions of the main input sequence features and the auxiliary input sequence features in the feature fusion module to obtain reduced-dimensional features; the convolutional neural network is combined with skip connections, and uses image multi-scale features and dimensionality reduction features to automatically identify subregions of myocardial tissue damage, and obtain regions of interest corresponding to VNE images, CEST sequences, and OS-BOLD sequences; the region division and quantitative parameter calculation module generates multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data based on the regions of interest, and calculates the quantitative parameters of MTT sequences, VNE images, CEST sequences, and OS-BOLD sequences in multiple types of myocardial tissue regions.

[0082] For example, threshold values set for T2, T2*, and LGE are obtained, and the T2, T2*, and LGE values for each pixel in the medical image sample are compared with the corresponding threshold values. If the values are greater than or equal to the threshold values, the pixel is within the region of interest; if the values are less than the threshold values, the pixel is outside the region of interest. If the preset parameters include MPSI signal intensity, the boundary zone can be an area where the MPSI signal intensity exceeds 40%.

[0083] Myocardial quantitative parameters refer to quantifiable myocardial parameters, such as mean intensity, Cr-CEST (creatine chemical exchange saturation transfer) metabolic value, and OS-SI value. Mean intensity is the mean signal intensity, and OS-SI = (carbon dioxide load - resting) / resting × 100%.

[0084] Specifically, for example Figure 2As shown in Figure 1, the visual transformer consists of a sequence embedding layer, a position embedding layer, and multiple self-attention layers. Each self-attention layer consists of an input unit, a multi-head self-attention mechanism, residual and normalization, an activation function, and an output unit.

[0085] In some embodiments of the present application, multi-sequence deep feature fusion is performed on the multi-scale features of the image to analyze the correlation and difference of image features between different sequences through a multi-head self-attention mechanism, and the specific process of obtaining the main input sequence features includes: the sequence embedding layer maps the multi-scale features of the image into low-dimensional feature vectors to capture the similarity between pixels and obtain a multi-scale embedding vector; the position embedding layer extracts the relative position features of different pixel positions from the multi-scale embedding vector to obtain multi-scale position features; the multi-scale position features are input into multiple self-attention layers, and different preset semantic tokens are collected for global classification to obtain the main input sequence features.

[0086] In some embodiments of the present application, based on the region of interest, multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data are generated, and the specific process of respectively calculating the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in the multiple types of myocardial tissue regions includes: based on the region of interest, dividing the hemorrhage area, infarction area, surrounding boundary area and normal tissue area as the multiple types of myocardial tissue regions; respectively calculating the Cr-CEST metabolic value for reflecting the myocardial metabolic state and the OS-SI value for reflecting the myocardial oxygenation state of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in the hemorrhage area, infarction area, surrounding boundary area and normal tissue area; and using the Cr-CEST metabolic value and the OS-SI value as the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in the multiple types of myocardial tissue regions.

[0087] For example, consider analyzing a patient's myocardial damage. The patient first undergoes a series of MRI scans, including MTT sequences, VNE images, CEST sequences, and OS-BOLD sequences. These scans are performed at different time points to capture the characteristics of myocardial tissue in different states. The physician then inputs these multi-sequence MRI datasets into a pre-trained myocardial damage information analysis model. This model, which may be a deep learning-based convolutional neural network or a visual transformer, has been trained to recognize patterns of myocardial damage. The model analyzes the input data and identifies different regions within the myocardial tissue, such as hemorrhagic areas, infarcted areas, surrounding border areas, and normal tissue areas. The model also calculates quantitative parameters for each region, such as T1, T2, T2* values, Cr-CEST metabolic values, and OS-SI values.

[0088] S103: Output multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data, and calculate quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in the multiple types of myocardial tissue regions to obtain myocardial injury information of the object to be identified.

[0089] In some embodiments of the present application, after the model processing obtains the results, the model can output multiple types of myocardial tissue areas corresponding to historical multi-sequence magnetic resonance imaging data, and calculate the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in multiple types of myocardial tissue areas to obtain the myocardial damage information of the object to be identified.

[0090] In an embodiment of the present application, by integrating multi-sequence magnetic resonance imaging data of MTT sequence, VNE image, CEST sequence and OS-BOLD sequence, combined with a pre-trained deep learning model, high-precision identification and quantitative analysis of myocardial damage areas are achieved. Since multi-sequence magnetic resonance imaging data provides multi-dimensional information for the analysis of myocardial damage, the model uses a convolutional neural network to extract multi-scale features specific to the sequence type, uses a visual transformer to perform multi-sequence deep feature fusion, and explores the correlation and difference of image features between different sequences through a multi-head self-attention mechanism, thereby improving the accuracy of myocardial damage information and reducing dependence on subjective judgment of experts.

[0091] See Figure 3 , provides a flowchart of a damage level analysis model training method according to an embodiment of the present application. Figure 3 As shown, the method of the embodiment of the present application may include the following steps:

[0092] S201, collecting historical cardiac magnetic resonance data of a hemorrhagic myocardial infarction model subject and a non-hemorrhagic myocardial infarction model subject, where the historical cardiac magnetic resonance data is collected at multiple preset time points;

[0093] A hemorrhagic myocardial infarction model refers to a model used to simulate myocardial infarction in experimental or clinical research, accompanied by myocardial hemorrhage. A non-hemorrhagic myocardial infarction model refers to a model that simulates myocardial infarction without myocardial hemorrhage. Pre-setting multiple time points involves performing cardiac magnetic resonance imaging at multiple predetermined time points before and after a myocardial infarction in order to fully understand cardiac changes after a myocardial infarction during research or clinical observation.

[0094] For example, these time points may include before myocardial infarction occurs (baseline), early after myocardial infarction (such as a few hours, a few days), and subsequent recovery period (such as a few weeks, a few months later). This application can be 7 time points, namely baseline before modeling, 0 day, 1 day, 3 days, 5 days, 7 days and 8 weeks after modeling.

[0095] S202, creating a network model combining convolution and transformer structures, the network model including a convolutional neural network, a visual transformer, an encoder-decoder structure and feature fusion module, and a region division and quantitative parameter calculation module;

[0096] S203, annotating regions of interest on historical cardiac magnetic resonance data to obtain model training samples for the segmentation task;

[0097] The historical cardiac magnetic resonance data include historical VNE images, historical T2* images, historical CEST images, and historical OS-BOLD images.

[0098] In some embodiments of the present application, the specific process of labeling the regions of interest of historical cardiac magnetic resonance data to obtain model training samples for the segmentation task includes: labeling the regions of interest of the left ventricle, infarction and hemorrhage areas of historical VNE images; labeling the regions of interest of the hemorrhage area of historical T2* images; labeling the regions of interest of the left ventricle and infarction areas of historical CEST images; labeling the regions of interest of the left ventricle and infarction areas of historical OS-BOLD images; and using the above labeling results as model training samples for the segmentation task.

[0099] S204: Generate a pre-trained myocardial injury information analysis model based on the model training samples and the network model.

[0100] In some embodiments of the present application, the specific process of generating a pre-trained myocardial injury information analysis model based on the model training samples and the network model includes: using the Dice loss function as the loss function of the segmentation task for each preset labeling category; integrating the loss function into the network model to obtain the myocardial injury information analysis model; inputting the model training samples into the myocardial injury information analysis model and outputting the loss value of the model; generating a pre-trained myocardial injury information analysis model when the loss value reaches the minimum.

[0101] In some embodiments of the present application, the method further includes: when the loss value has not reached the minimum, backpropagating the loss value to update the model parameters of the network model, and continuing to execute the step of inputting historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model until the loss value reaches the minimum.

[0102] In an embodiment of the present application, by integrating multi-sequence magnetic resonance imaging data of MTT sequence, VNE image, CEST sequence and OS-BOLD sequence, combined with a pre-trained deep learning model, high-precision identification and quantitative analysis of myocardial damage areas are achieved. Since multi-sequence magnetic resonance imaging data provides multi-dimensional information for the analysis of myocardial damage, the model uses a convolutional neural network to extract multi-scale features specific to the sequence type, uses a visual transformer to perform multi-sequence deep feature fusion, and explores the correlation and difference of image features between different sequences through a multi-head self-attention mechanism, thereby improving the accuracy of myocardial damage information and reducing dependence on subjective judgment of experts.

[0103] The following are device embodiments of the present application, which can be used to implement the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0104] See Figure 4 , which shows a schematic diagram of the structure of a multi-sequence magnetic resonance imaging-based myocardial injury information analysis device, provided by an exemplary embodiment of the present application. This multi-sequence magnetic resonance imaging-based myocardial injury information analysis device can be implemented as all or part of an electronic device through software, hardware, or a combination of both. The device 1 includes an image acquisition module 10, a data input module 20, and a result output module 30.

[0105] An image acquisition module 10 is configured to acquire historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the subject to be identified, generated based on nuclear magnetic resonance, wherein the historical multi-sequence magnetic resonance imaging data includes an MTT sequence, a VNE image, a CEST sequence, and an OS-BOLD sequence;

[0106] A data input module 20 is used to input historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model to analyze the historical multi-sequence magnetic resonance imaging data;

[0107] The result output module 30 is used to output multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data, and to calculate the quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in multiple types of myocardial tissue regions to obtain myocardial damage information of the object to be identified.

[0108] It should be noted that the multi-sequence MRI-based myocardial injury information analysis device provided in the above embodiment, when executing the multi-sequence MRI-based myocardial injury information analysis method, only uses the division of the above-mentioned functional modules as an example. In actual application, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the multi-sequence MRI-based myocardial injury information analysis device provided in the above embodiment and the multi-sequence MRI-based myocardial injury information analysis method embodiment are based on the same concept. The implementation process is detailed in the method embodiment and will not be repeated here.

[0109] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0110] In an embodiment of the present application, by integrating multi-sequence magnetic resonance imaging data of MTT sequence, VNE image, CEST sequence and OS-BOLD sequence, combined with a pre-trained deep learning model, high-precision identification and quantitative analysis of myocardial damage areas are achieved. Since multi-sequence magnetic resonance imaging data provides multi-dimensional information for the analysis of myocardial damage, the model uses a convolutional neural network to extract multi-scale features specific to the sequence type, uses a visual transformer to perform multi-sequence deep feature fusion, and explores the correlation and difference of image features between different sequences through a multi-head self-attention mechanism, thereby improving the accuracy of myocardial damage information and reducing dependence on subjective judgment of experts.

[0111] The present application also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implements the myocardial injury information analysis method based on multi-sequence magnetic resonance imaging provided by the above-mentioned various method embodiments.

[0112] The present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the myocardial injury information analysis method based on multi-sequence magnetic resonance imaging of each of the above method embodiments.

[0113] See Figure 5 , is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 5 As shown, the electronic device 1000 may include: at least one processor 1001 , at least one network interface 1004 , a user interface 1003 , a memory 1005 , and at least one communication bus 1002 .

[0114] The communication bus 1002 is used to implement the connection and communication between these components.

[0115] The user interface 1003 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.

[0116] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0117] The processor 1001 may include one or more processing cores. The processor 1001 utilizes various interfaces and circuits to connect the various components within the entire electronic device 1000. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 1005, and calling data stored in the memory 1005, the processor 1001 performs various functions of the electronic device 1000 and processes data. Optionally, the processor 1001 may be implemented in the form of at least one hardware component selected from the group consisting of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 1001 may integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU primarily processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display; and the modem is responsible for handling wireless communications. It is understood that the modem may not be integrated into the processor 1001 and may be implemented separately on a single chip.

[0118] Among them, the memory 1005 may include a random access memory (RAM) or a read-only memory (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 1005 may also be optionally at least one storage system located away from the aforementioned processor 1001. As Figure 5 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a myocardial injury information analysis application based on multi-sequence magnetic resonance imaging.

[0119] exist Figure 5 In the electronic device 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain user input data; and the processor 1001 can be used to call the myocardial injury information analysis application based on multi-sequence magnetic resonance imaging stored in the memory 1005 and specifically perform the following operations:

[0120] Acquire historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the subject to be identified based on nuclear magnetic resonance, the historical multi-sequence magnetic resonance imaging data including MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence;

[0121] Inputting historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model to analyze the historical multi-sequence magnetic resonance imaging data;

[0122] The multiple myocardial tissue regions corresponding to historical multi-sequence MRI data are output, and the quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in the multiple myocardial tissue regions are calculated to obtain the myocardial damage information of the object to be identified.

[0123] In one embodiment, when the processor 1001 analyzes the historical multi-sequence magnetic resonance imaging data, it specifically performs the following operations:

[0124] The convolutional neural network extracts multi-scale features of preset sequence types for VNE images, T2* images, CEST sequences, and OS-BOLD sequences to obtain multi-scale features of the images.

[0125] The visual transformer performs multi-sequence deep feature fusion on the multi-scale features of the image, and analyzes the correlation and difference of image features between different sequences through the multi-head self-attention mechanism preset in the visual transformer to obtain the main input sequence features;

[0126] The convolutional neural network extracts multi-scale features of preset sequence types from T1 images, T1ρ images, and T2 images to obtain auxiliary input sequence features;

[0127] The encoder-decoder structure fuses and reduces the dimensions of the main input sequence features and the auxiliary input sequence features in the feature fusion module to obtain reduced-dimensional features.

[0128] A convolutional neural network combined with skip connections uses multi-scale image features and dimensionality reduction features to automatically identify subregions of myocardial tissue damage, obtaining regions of interest corresponding to VNE images, CEST sequences, and OS-BOLD sequences.

[0129] The regional division and quantitative parameter calculation module generates multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data based on the region of interest, and calculates the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in multiple types of myocardial tissue regions.

[0130] In one embodiment, when the processor 1001 generates multiple types of myocardial tissue regions corresponding to historical multi-sequence magnetic resonance imaging data based on the region of interest, and respectively calculates quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in the multiple types of myocardial tissue regions, the processor 1001 specifically performs the following operations:

[0131] Based on the region of interest, the hemorrhage area, infarct area, surrounding boundary area and normal tissue area were divided as multiple types of myocardial tissue areas;

[0132] The Cr-CEST metabolic value, which reflects the myocardial metabolic state, and the OS-SI value, which reflects the myocardial oxygenation state, were calculated in the hemorrhagic area, infarct area, surrounding border area, and normal tissue area using MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence.

[0133] Cr-CEST metabolic values and OS-SI values were used as quantitative parameters for MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in multiple types of myocardial tissue regions.

[0134] In one embodiment, when performing multi-sequence deep feature fusion on multi-scale image features to analyze the correlation and difference of image features between different sequences through a multi-head self-attention mechanism to obtain the main input sequence features, the processor 1001 specifically performs the following operations:

[0135] The sequence embedding layer maps the multi-scale features of the image into low-dimensional feature vectors to capture the similarity between pixels and obtain a multi-scale embedding vector;

[0136] The position embedding layer extracts the relative position features of different pixel positions from the multi-scale embedding vector to obtain multi-scale position features;

[0137] The multi-scale position features are input into multiple self-attention layers, and different preset semantic tokens are collected for global classification to obtain the main input sequence features.

[0138] In one embodiment, before executing the process of acquiring historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the subject to be identified, generated based on nuclear magnetic resonance, the processor 1001 further performs the following operations:

[0139] Acquiring a nuclear magnetic resonance image of the myocardial tissue of the subject to be identified;

[0140] Using an inversion pulse, a T2 preparation pulse, and a T1ρ preparation pulse to generate different T1, T2, and T1ρ contrasts on a nuclear magnetic resonance image, thereby obtaining a first image sequence;

[0141] Reconstructing a multi-echo image using data collected from myocardial tissue during the diastole of a target cardiac cycle, and performing a second image processing on the multi-echo image to obtain a second image sequence; the first image sequence and the second image sequence include images of multiple dimensions of T1 / T2 / T2* / T1ρMapping;

[0142] Use T1 / T2 / T2* / T1ρMapping images in multiple dimensions as MTT sequences;

[0143] The MTT sequence is processed by a preset generative adversarial network to obtain a VNE image;

[0144] By using a short pulse of saturation signal, the preset chemical groups in the myocardial tissue are imaged separately, so as to fit the CEST sequence by measuring the creatine content in the myocardial tissue;

[0145] The signal changes caused by hyperventilation and breath holding of the subject to be identified are monitored through a carbon dioxide mask inhalation device, so as to fit the OS-BOLD sequence through the signal changes.

[0146] In one embodiment, when generating a pre-trained myocardial injury information analysis model, the processor 1001 specifically performs the following operations:

[0147] Collecting historical cardiac magnetic resonance data of hemorrhagic myocardial infarction model subjects and non-hemorrhagic myocardial infarction model subjects, wherein the historical cardiac magnetic resonance data are acquired at multiple preset time points;

[0148] Create a network model that combines convolution and transformer structures. The network model includes a convolutional neural network, a visual transformer, an encoder-decoder structure and feature fusion module, and a region division and quantitative parameter calculation module.

[0149] Label the regions of interest (ROIs) of historical cardiac magnetic resonance data to obtain model training samples for segmentation tasks.

[0150] A pre-trained myocardial injury information analysis model is generated based on the model training samples and the network model.

[0151] In one embodiment, when the processor 1001 performs region of interest annotation on historical cardiac magnetic resonance data to obtain model training samples for a segmentation task, the processor 1001 specifically performs the following operations:

[0152] Regions of interest (ROIs) of the left ventricle, infarct, and hemorrhage areas were annotated on historical VNE images;

[0153] Marking of regions of interest in the hemorrhage area on historical T2* images;

[0154] The regions of interest of the left ventricle and infarct area were marked on historical CEST images;

[0155] Regions of interest (ROIs) of the left ventricle and infarct area were annotated on historical OS-BOLD images;

[0156] The above labeled results are used as model training samples for the segmentation task.

[0157] In one embodiment, when the processor 1001 generates a pre-trained myocardial injury information analysis model based on the model training samples and the network model, the processor 1001 specifically performs the following operations:

[0158] Use the Dice loss function as the loss function for the segmentation task for each preset label category;

[0159] The loss function is integrated into the network model to obtain the myocardial injury information analysis model;

[0160] Input the model training samples into the myocardial injury information analysis model and output the loss value of the model;

[0161] When the loss value reaches the minimum, a pre-trained myocardial injury information analysis model is generated.

[0162] In one embodiment, the processor 1001 further performs the following operations:

[0163] When the loss value has not reached the minimum, the loss value is back-propagated to update the model parameters of the network model, and the step of inputting the historical multi-sequence magnetic resonance imaging data into the pre-trained myocardial injury information analysis model is continued until the loss value reaches the minimum.

[0164] In an embodiment of the present application, by integrating multi-sequence magnetic resonance imaging data of MTT sequence, VNE image, CEST sequence and OS-BOLD sequence, combined with a pre-trained deep learning model, high-precision identification and quantitative analysis of myocardial damage areas are achieved. Since multi-sequence magnetic resonance imaging data provides multi-dimensional information for the analysis of myocardial damage, the model uses a convolutional neural network to extract multi-scale features specific to the sequence type, uses a visual transformer to perform multi-sequence deep feature fusion, and explores the correlation and difference of image features between different sequences through a multi-head self-attention mechanism, thereby improving the accuracy of myocardial damage information and reducing dependence on subjective judgment of experts.

[0165] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The program for analyzing myocardial injury information based on multi-sequence MRI can be stored in a computer-readable storage medium. When executed, the program can include the processes in the above-described method embodiments. The storage medium for the program for analyzing myocardial injury information based on multi-sequence MRI can be a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0166] The above disclosure is only a preferred embodiment of the present application, and certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A method for analyzing myocardial injury information based on multi-sequence magnetic resonance imaging, characterized in that: The method comprises: Acquiring historical multi-sequence magnetic resonance imaging data of myocardial tissue of the subject to be identified based on nuclear magnetic resonance, wherein the historical multi-sequence magnetic resonance imaging data includes an MTT sequence, a VNE image, a CEST sequence, and an OS-BOLD sequence; Inputting the historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model to analyze the historical multi-sequence magnetic resonance imaging data; Output multiple types of myocardial tissue regions corresponding to the historical multi-sequence magnetic resonance imaging data, and calculate the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in the multiple types of myocardial tissue regions, to obtain myocardial damage information of the object to be identified.

2. The method according to claim 1, characterized in that The pre-trained myocardial injury information analysis model includes a convolutional neural network, a visual transformer, a codec structure and feature fusion module, and a region division and quantitative parameter calculation module; the MTT sequence includes T2* images, T1 images, T1ρ images, and T2 images; The analyzing the historical multi-sequence magnetic resonance imaging data includes: The convolutional neural network performs multi-scale feature extraction of preset sequence types on the VNE image, T2* image, CEST sequence, and OS-BOLD sequence to obtain multi-scale features of the image; The visual transformer performs multi-sequence deep feature fusion on the multi-scale features of the image, so as to analyze the correlation and difference of image features between different sequences through the multi-head self-attention mechanism preset in the visual transformer, and obtain the main input sequence features; The convolutional neural network performs multi-scale feature extraction of a preset sequence type on the T1 image, T1ρ image, and T2 image to obtain auxiliary input sequence features; The codec structure performs feature fusion and dimensionality reduction on the main input sequence features and the auxiliary input sequence features in the feature fusion module to obtain reduced dimensionality features; The convolutional neural network is combined with skip connections to automatically identify subregions of myocardial tissue damage using the image multi-scale features and the dimensionality reduction features, thereby obtaining regions of interest corresponding to VNE images, CEST sequences, and OS-BOLD sequences; The region division and quantitative parameter calculation module generates multiple types of myocardial tissue regions corresponding to the historical multi-sequence magnetic resonance imaging data based on the region of interest, and calculates the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in the multiple types of myocardial tissue regions.

3. The method according to claim 2, characterized in that Generating multiple types of myocardial tissue regions corresponding to the historical multi-sequence magnetic resonance imaging data according to the region of interest, and respectively calculating quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in the multiple types of myocardial tissue regions, includes: Based on the region of interest, a hemorrhage area, an infarct area, a surrounding boundary area and a normal tissue area are divided as multiple types of myocardial tissue areas; Calculating the Cr-CEST metabolic value for reflecting the myocardial metabolic state and the OS-SI value for reflecting the myocardial oxygenation state in the hemorrhage area, infarct area, surrounding border area and normal tissue area using the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence respectively; The Cr-CEST metabolic value and the OS-SI value are used as quantitative parameters of the MTT sequence, VNE image, CEST sequence, and OS-BOLD sequence in the multiple types of myocardial tissue regions.

4. The method according to claim 2, characterized in that The visual transformer includes a sequence embedding layer, a position embedding layer, and multiple self-attention layers; The multi-scale features of the image are subjected to multi-sequence deep feature fusion to analyze the correlation and difference of image features between different sequences through a multi-head self-attention mechanism to obtain the main input sequence features, including: The sequence embedding layer maps the multi-scale features of the image into a low-dimensional feature vector to capture the similarity between pixels and obtain a multi-scale embedding vector; The position embedding layer extracts relative position features of different pixel positions from the multi-scale embedding vector to obtain multi-scale position features; The multi-scale position features are input into the multiple self-attention layers, and different preset semantic tokens are collected for global classification to obtain the main input sequence features.

5. The method according to claim 1, wherein Before acquiring historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the subject to be identified, generated based on nuclear magnetic resonance, the method further includes: Acquiring a nuclear magnetic resonance image of the myocardial tissue of the subject to be identified; Using an inversion pulse, a T2 preparation pulse, and a T1ρ preparation pulse to generate different T1, T2, and T1ρ contrasts on the nuclear magnetic resonance image to obtain a first image sequence; Reconstructing a multi-echo image using data collected from the myocardial tissue during the diastole of a target cardiac cycle, and performing a second image processing on the multi-echo image to obtain a second image sequence; the first image sequence and the second image sequence include images of multiple dimensions of T1 / T2 / T2* / T1ρMapping; The T1 / T2 / T2* / T1ρMapping multi-dimensional images are used as MTT sequences; Processing the MTT sequence through a preset generative adversarial network to obtain a VNE image; Using a short pulse saturation signal, individually imaging the preset chemical groups in the myocardial tissue, so as to fit a CEST sequence by measuring the creatine content in the myocardial tissue; The signal changes caused by hyperventilation and breath holding of the subject to be identified are monitored through a carbon dioxide mask inhalation device, so as to fit the OS-BOLD sequence through the signal changes.

6. The method according to claim 1, characterized in that The following steps are used to generate a pre-trained myocardial injury information analysis model, including: collecting historical cardiac magnetic resonance data of a hemorrhagic myocardial infarction model subject and a non-hemorrhagic myocardial infarction model subject, wherein the historical cardiac magnetic resonance data is acquired at a plurality of preset time points; Creating a network model that combines convolution and transformer structures, the network model includes a convolutional neural network, a visual transformer, an encoder-decoder structure and feature fusion module, and a region division and quantitative parameter calculation module; Annotating regions of interest on the historical cardiac magnetic resonance data to obtain model training samples for a segmentation task; A pre-trained myocardial injury information analysis model is generated based on the model training samples and the network model.

7. The method according to claim 6, characterized in that The historical cardiac magnetic resonance data includes historical VNE images, historical T2* images, historical CEST images, and historical OS-BOLD images; The step of labeling the historical cardiac magnetic resonance data with regions of interest to obtain model training samples for the segmentation task includes: Marking regions of interest (ROIs) of the left ventricle, infarct, and hemorrhage areas on the historical VNE images; Marking a region of interest (ROI) of a hemorrhage area on the historical T2* image; Marking the regions of interest of the left ventricle and the infarct area on the historical CEST images; Marking the regions of interest of the left ventricle and infarct area on the historical OS-BOLD images; The above labeled results are used as model training samples for the segmentation task.

8. The method according to claim 6, characterized in that Generating a pre-trained myocardial injury information analysis model based on the model training sample and the network model includes: Using the Dice loss function as the loss function for the segmentation task for each preset labeled category; Integrating the loss function into the network model to obtain a myocardial injury information analysis model; Inputting the model training sample into the myocardial injury information analysis model and outputting the loss value of the model; When the loss value reaches the minimum, a pre-trained myocardial injury information analysis model is generated.

9. The method according to claim 8, characterized in that The method further comprises: In the case that the loss value has not reached the minimum, the loss value is back-propagated to update the model parameters of the network model, and the step of inputting the historical multi-sequence magnetic resonance imaging data into the pre-trained myocardial injury information analysis model is continued until the loss value reaches the minimum.

10. A myocardial injury information analysis device based on multi-sequence magnetic resonance imaging, characterized in that: The device comprises: An image acquisition module is used to acquire historical multi-sequence magnetic resonance imaging data of the myocardial tissue of the object to be identified, which is generated based on nuclear magnetic resonance. The historical multi-sequence magnetic resonance imaging data includes MTT sequence, VNE image, CEST sequence and OS-BOLD sequence; a data input module, configured to input the historical multi-sequence magnetic resonance imaging data into a pre-trained myocardial injury information analysis model to analyze the historical multi-sequence magnetic resonance imaging data; The result output module is used to output multiple types of myocardial tissue areas corresponding to the historical multi-sequence magnetic resonance imaging data, and to calculate the quantitative parameters of the MTT sequence, VNE image, CEST sequence and OS-BOLD sequence in the multiple types of myocardial tissue areas to obtain the myocardial damage information of the object to be identified.

Images