Ischemia reperfusion injury level analysis method and device
Through the pre-trained injury level analysis model, the multi-sequence medical images of myocardial tissue are automatically classified and segmented, which solves the problem that cardiac MRI image analysis relies on doctors' subjective judgment, and improves the accuracy and efficiency of ischemia-reperfusion injury level.
Patent Information
- Application Number
- CN202510367722.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-26
- Publication Date
- 2025-08-12
AI Technical Summary
The prior art central heart MRI image analysis relies on the doctor's subjective judgment, resulting in insufficient accuracy of the level of ischemia-reperfusion injury and is long-term.
The pre-trained injury-level analysis model is used to classify and segment predict the multi-sequence medical images of myocardial tissue through convolutional neural network, visual transformer and codec. Combined with the preset CCS staging standards, the area and type of myocardial injury are automatically identified.
It significantly reduces the workload of doctors in image analysis, shortens analysis time, and improves the accuracy of ischemia-reperfusion injury levels, allowing quick identification of subtle features of myocardial injury.
Smart Images
Figure CN120473112A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a method and device for analyzing the level of ischemia-reperfusion injury. Background Art
[0002] In the diagnosis and treatment of cardiovascular disease, accurately assessing the extent of myocardial damage is crucial for developing treatment plans and predicting patient prognosis. Cardiac magnetic resonance imaging (MRI), as a non-invasive diagnostic tool, can provide detailed information on myocardial tissue properties and signal changes. In patients with acute myocardial infarction (AMI), cardiac MRI can particularly identify myocardial edema, necrosis, microcirculatory obstruction (MVO), and hemorrhage.
[0003] Currently, the analysis of cardiac MRI images relies primarily on the expertise and experience of radiologists. Physicians visually interpret T2 mapping, T2* mapping, and synthesized virtual late gadolinium-enhanced images, manually annotating regions of interest (ROIs) to identify myocardial lesions. The identification and grading of these lesions typically adhere to the Canadian Cardiovascular Society (CCS) staging criteria. However, this process is not only time-consuming but also highly dependent on the physician's subjective judgment, leading to errors in the results and reducing the accuracy of the ischemia-reperfusion injury grade. Summary of the Invention
[0004] The present embodiments provide a method and apparatus for analyzing the level of ischemia-reperfusion injury. 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 components, or delineate the scope of protection of these embodiments. Its sole purpose is to present some concepts in a simplified form, serving as a prelude to the detailed description that follows.
[0005] In a first aspect, an embodiment of the present application provides a method for analyzing the level of ischemia-reperfusion injury, the method comprising:
[0006] Acquire historical multi-series medical images of the myocardial tissue of the object to be identified, generated based on nuclear magnetic resonance;
[0007] Inputting historical multi-series medical images into a pre-trained injury level analysis model to perform classification prediction and segmentation prediction on the historical multi-series medical images, wherein the pre-trained injury level analysis model is generated based on a first training sample for a classification task and a second training sample for a segmentation task;
[0008] Output the myocardial tissue damage area and myocardial damage type corresponding to historical multi-series medical images;
[0009] The ischemia-reperfusion injury level of the myocardial tissue injury area is determined based on the myocardial injury type and the preset CCS staging criteria. The preset CCS staging criteria include a mapping relationship between the myocardial injury type and the ischemia-reperfusion injury level.
[0010] Optionally, the historical multi-series medical images include T1 images, T2 images, T1ρ images, T2* images, and VNE images. The T1 images, T2 images, T1ρ images, and T2* images are obtained through different magnetic resonance scanning sequences. The VNE images are virtual late gadolinium-enhanced images obtained by processing MTT images through a preset deep neural network. The MTT images are obtained by measuring the flow time of a contrast agent in myocardial tissue to evaluate myocardial blood perfusion information.
[0011] Pre-trained damage level analysis models include convolutional neural networks, visual transformers, and encoder-decoders;
[0012] Perform classification and segmentation predictions on historical multi-sequence medical images, including:
[0013] Convolutional neural network performs multi-scale feature extraction on T1 images, T2 images, T1ρ images, T2* images and VNE images to obtain T1 image features, T2 image features, T1ρ image features, T2* image features and VNE image features;
[0014] Use T1 image features and T1ρ image features as auxiliary features;
[0015] The visual transformer classifies and predicts T2 image features, T2* image features, and VNE image features to obtain myocardial injury type, T2 visual features, T2* visual features, and VNE visual features;
[0016] The codec performs segmentation prediction on the auxiliary features, T2 visual features, T2* visual features, and VNE visual features to obtain the myocardial tissue damage area.
[0017] Optionally, the visual transformer includes a sequence embedding layer, a position embedding layer, and a self-attention layer;
[0018] Use the visual transformer to classify and predict T2 image features, T2* image features, and VNE image features to obtain myocardial injury types, T2 visual features, T2* visual features, and VNE visual features, including:
[0019] The sequence embedding layer maps T2 image features, T2* image features, and VNE image features into low-dimensional feature vectors to capture the similarity between pixels, and obtains T2 embedding vectors, T2* embedding vectors, and VNE embedding vectors.
[0020] The position embedding layer extracts the relative position features of different pixel positions from the T2 image features, T2* image features, and VNE image features to obtain T2 position features, T2* position features, and VNE position features;
[0021] The T2 embedding vector, T2* embedding vector, VNE embedding vector, T2 position feature, T2* position feature, and VNE position feature are input into the self-attention layer, and different preset semantic tokens are collected for global classification to predict the type of myocardial injury;
[0022] The T2 embedding vector, T2* embedding vector, VNE embedding vector, T2 position feature, T2* position feature, and VNE position feature are mapped and associated to obtain T2 visual feature, T2* visual feature, and VNE visual feature.
[0023] Optionally, the codec includes a feature fusion module and a segmentation prediction module;
[0024] The codec performs segmentation prediction based on the auxiliary features, T2 visual features, T2* visual features, and VNE visual features to obtain the myocardial tissue damage area, including:
[0025] Input the auxiliary features, T2 visual features, T2* visual features, and VNE visual features into the feature fusion module to perform feature fusion and feature dimensionality reduction on the auxiliary features with the T2 visual features, T2* visual features, and VNE visual features, respectively, to obtain T2 fusion features, T2* fusion features, and VNE fusion features;
[0026] The codec concatenates the T2 image features, T2* image features, and VNE image features with the T2 fusion features, T2* fusion features, and VNE fusion features to obtain T2 concatenation features, T2* concatenation features, and VNE concatenation features;
[0027] The T2 splicing features, T2* splicing features and VNE splicing features are decoded to obtain the myocardial tissue damage area.
[0028] Optionally, the preset CCS staging standard includes a mapping relationship between myocardial injury types and ischemia-reperfusion injury levels; the ischemia-reperfusion injury levels include a first ischemia-reperfusion injury level, a second ischemia-reperfusion injury level, a third ischemia-reperfusion injury level, and a fourth ischemia-reperfusion injury level; the myocardial injury types corresponding to the first ischemia-reperfusion injury level include edema; the myocardial injury types corresponding to the second ischemia-reperfusion injury level include edema and myocardial necrosis; the myocardial injury types corresponding to the third ischemia-reperfusion injury level include edema, myocardial necrosis, and microcirculation obstruction; and the myocardial injury types corresponding to the fourth ischemia-reperfusion injury level include intramyocardial hemorrhage;
[0029] The degree of ischemia-reperfusion injury in the myocardial tissue injury area is determined based on the type of myocardial injury and the preset CCS staging criteria, including:
[0030] According to the type of myocardial injury, the corresponding ischemia-reperfusion injury level is obtained from the mapping relationship.
[0031] Optionally, generate a pre-trained damage level analysis model by following these steps:
[0032] Collecting sample multi-sequence medical images of target subjects with and without ischemia-reperfusion injury;
[0033] Create a hybrid structure model, which includes a convolutional neural network, a visual transformer, and an encoder-decoder. The visual transformer includes a sequence embedding layer, a position embedding layer, and a self-attention layer. The encoder-decoder includes a feature fusion module and a segmentation prediction module.
[0034] Labeling the myocardial injury type labels on the sample multi-sequence medical images to obtain the first training sample for the classification task. The myocardial injury type labels include whether there is edema, whether there is myocardial necrosis, whether there is microcirculation obstruction, and whether there is intramyocardial hemorrhage;
[0035] Segmenting the myocardial tissue damage area in the sample multi-series medical image to obtain a second training sample for the segmentation task;
[0036] A pre-trained damage level analysis model is generated based on the first training sample and the second training sample.
[0037] Optionally, the sample multi-series medical images include historical T2 images, historical T2* images, and historical VNE images;
[0038] The myocardial tissue damage area in the multi-series medical image segmentation sample is obtained to obtain a second training sample for the segmentation task, including:
[0039] In the historical T2 images, the edema area was marked as the region of interest;
[0040] In the historical T2* images, the intramyocardial hemorrhage area was marked as the region of interest;
[0041] In the historical VNE images, the myocardial necrosis area and the microcirculatory obstruction area were marked as the region of interest;
[0042] The historical T2 images, historical T2* images, and historical VNE images with marked regions of interest are used as the second training samples for the segmentation task.
[0043] Optionally, generating a pre-trained damage level analysis model based on the first training sample and the second training sample includes:
[0044] Use the multi-label cross entropy loss function as the first loss function for the classification task;
[0045] Use the Dice loss function as the second loss function for the segmentation task;
[0046] The first loss function and the second loss function are integrated into the hybrid structural model to obtain a damage level analysis model;
[0047] The damage level analysis model is trained according to the first training sample and the second training sample to obtain a pre-trained damage level analysis model.
[0048] Optionally, training the damage level analysis model based on the first training sample and the second training sample to obtain a pre-trained damage level analysis model includes:
[0049] Inputting the first training sample and the second training sample into the damage level analysis model to determine a first model loss value according to a first loss function, and to determine a second model loss value according to a second loss function;
[0050] When the first model loss value and the second model loss value reach a minimum, a pre-trained damage level analysis model is generated.
[0051] In a second aspect, an embodiment of the present application provides an ischemia-reperfusion injury level analysis device, comprising:
[0052] An acquisition module, configured to acquire historical multi-series medical images of the myocardial tissue of the subject to be identified, generated based on nuclear magnetic resonance imaging;
[0053] An input module is used to input historical multi-series medical images into a pre-trained injury level analysis model to perform classification prediction and segmentation prediction on the historical multi-series medical images, where the pre-trained injury level analysis model is generated based on a first training sample for the classification task and a second training sample for the segmentation task;
[0054] An output module is used to output the myocardial tissue damage area and myocardial damage type corresponding to historical multi-series medical images;
[0055] The determination module is used to determine the ischemia-reperfusion injury level of the myocardial tissue injury area according to the myocardial injury type and the preset CCS staging standard. The preset CCS staging standard includes a mapping relationship between the myocardial injury type and the ischemia-reperfusion injury level.
[0056] The technical solutions provided by the embodiments of the present application may have the following beneficial effects:
[0057] In the embodiments of the present application, on the one hand, the workload of doctors in image analysis can be significantly reduced through the automated injury level analysis model. The model is pre-trained to quickly perform classification predictions and segmentation predictions on historical multi-series medical images, thereby shortening the time from image acquisition to analysis result output. On the other hand, the automated injury level analysis model can learn and identify subtle features of myocardial injury through deep learning technology, and can accurately output the myocardial tissue damage area and damage type, thereby improving the accuracy of the ischemia-reperfusion injury level.
[0058] 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
[0059] 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.
[0060] Figure 1 This is a flow chart of a method for analyzing the level of ischemia-reperfusion injury provided in an embodiment of the present application;
[0061] Figure 2 This is a schematic diagram of a network architecture of a damage level analysis model provided in an embodiment of the present application;
[0062] Figure 3 This is a schematic diagram of different myocardial injury types corresponding to different ischemia-reperfusion injury levels in a preset CCS staging standard provided in an embodiment of the present application;
[0063] Figure 4 This is a flow chart of a model training method provided in an embodiment of the present application;
[0064] Figure 5 This is a schematic structural diagram of an ischemia-reperfusion injury level analysis device provided by the present application;
[0065] Figure 6 This is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0066] The following description and the drawings sufficiently illustrate specific embodiments of the application to enable those skilled in the art to practice them.
[0067] 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.
[0068] 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.
[0069] 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.
[0070] The present application provides an ischemia-reperfusion injury level analysis method and device to solve the problems existing in the related technical issues. In the embodiment of the present application, on the one hand, the workload of doctors in image analysis can be significantly reduced through the automated injury level analysis model. Through pre-training, the model can quickly perform classification prediction and segmentation prediction on historical multi-series medical images, thereby shortening the time from image acquisition to analysis result output. On the other hand, the automated injury level analysis model can learn and identify subtle features of myocardial injury through deep learning technology, and can accurately output the myocardial tissue damage area and damage type, thereby improving the accuracy of the ischemia-reperfusion injury level. The following exemplary embodiments are used for detailed description.
[0071] The following will be combined with the Figure 1 -Attached Figure 4 This article details the ischemia-reperfusion injury level analysis method provided in the embodiments of this application. This method can be implemented using a computer program and run on an ischemia-reperfusion injury level analysis device based on the von Neumann architecture. This computer program can be integrated into an application or run as a standalone tool.
[0072] See Figure 1 , provides a flow chart of a method for analyzing the level of ischemia-reperfusion injury in an embodiment of the present application. Figure 1 As shown, the method of the embodiment of the present application may include the following steps:
[0073] S101, obtaining a historical multi-series medical image of the myocardial tissue of the subject to be identified, generated based on nuclear magnetic resonance imaging;
[0074] The object to be identified refers to a patient or subject who needs to be identified and evaluated for myocardial damage. Myocardial tissue refers to the muscle tissue of the heart, which is responsible for the heart's contraction and pumping function. Magnetic resonance imaging (MRI) is an imaging technology that uses strong magnetic fields and radio waves to generate detailed images of the human body. Multi-sequence medical imaging refers to a series of images acquired using different MRI scanning parameters and pulse sequences. These images can provide detailed information about myocardial tissue characteristics and signal changes.
[0075] In some embodiments, historical cardiac magnetic resonance imaging (MRI) data pre-stored for the object to be identified is collected, including T1, T2, T1ρ, T2* images, and MTT images obtained by the MTT sequence. T1, T2, T1ρ, and T2* images are obtained through different magnetic resonance scanning sequences, and each sequence is optimized for the different characteristics and signal changes of myocardial tissue. The MTT image evaluates myocardial blood perfusion information by measuring the flow time of the contrast agent in the myocardial tissue. The MTT image is processed using a preset deep neural network to generate a virtual late gadolinium enhancement (VNE) image. VNE images help identify myocardial damage such as edema, necrosis, microcirculatory obstruction (MVO), and intramyocardial hemorrhage.
[0076] S102, inputting the historical multi-series medical images into a pre-trained injury level analysis model to perform classification prediction and segmentation prediction on the historical multi-series medical images, where the pre-trained injury level analysis model is generated based on a first training sample for the classification task and a second training sample for the segmentation task;
[0077] Among them, the injury level analysis model is a deep learning model used to analyze cardiac MRI images and identify the degree and type of myocardial injury. The classification prediction task refers to predicting whether the myocardial tissue has a specific type of injury, such as edema, necrosis, MVO, and intramyocardial hemorrhage. The segmentation prediction task refers to accurately identifying and segmenting myocardial injury areas in the image, such as edema areas, necrosis areas, MVO areas, and intramyocardial hemorrhage areas. The first training sample is a data sample used to train the model for classification prediction, which contains annotation information of known injury types. The second training sample is a data sample used to train the model for segmentation prediction, which contains precise boundary annotations of myocardial injury areas.
[0078] Among them, historical multi-sequence medical images include T1 images, T2 images, T1ρ images, T2* images and VNE images. T1 images, T2 images, T1ρ images, and T2* images are obtained through different magnetic resonance scanning sequences. VNE images are virtual late gadolinium-enhanced images obtained by processing MTT images through a preset deep neural network. MTT images are obtained by measuring the flow time of contrast agent in myocardial tissue to evaluate myocardial blood perfusion information.
[0079] Among them, for example Figure 2 As shown in Figure 2, the pre-trained damage level analysis model includes a convolutional neural network, a visual transformer, and an encoder-decoder. The visual transformer includes a sequence embedding layer, a position embedding layer, and a self-attention layer. The encoder-decoder includes a feature fusion module and a segmentation prediction module.
[0080] In the Vision Transformer, the sequence embedding layer converts image features into a serialized embedding vector that the model can process. The position embedding layer adds the position information of each small patch in the original image to the embedding vector, allowing the model to understand the spatial relationship between features. The encoder-decoder is a neural network architecture used for image segmentation tasks. It consists of an encoder to extract features and a decoder to generate segmentation results.
[0081] In some embodiments of the present application, the specific process of classification prediction and segmentation prediction of historical multi-sequence medical images includes: a convolutional neural network performs multi-scale feature extraction on T1 images, T2 images, T1ρ images, T2* images and VNE images to obtain T1 image features, T2 image features, T1ρ image features, T2* image features and VNE image features; T1 image features and T1ρ image features are used as auxiliary features; a visual transformer performs classification prediction on T2 image features, T2* image features and VNE image features to obtain myocardial injury types, T2 visual features, T2* visual features and VNE visual features; a codec performs segmentation prediction on the auxiliary features and T2 visual features, T2* visual features and VNE visual features to obtain myocardial tissue damage areas.
[0082] Specifically, the visual transformer is used to classify and predict T2 image features, T2* image features and VNE image features to obtain the myocardial injury type, T2 visual features, T2* visual features and VNE visual features. The specific process includes: the sequence embedding layer maps the T2 image features, T2* image features and VNE image features into low-dimensional feature vectors to capture the similarity between pixels, and obtains T2 embedding vectors, T2* embedding vectors and VNE embedding vectors; the position embedding layer extracts the relative positions of different pixel positions from the T2 image features, T2* image features and VNE image features. For the position features, T2 position features, T2* position features and VNE position features are obtained; the T2 embedding vector, T2* embedding vector and VNE embedding vector, T2 position features, T2* position features and VNE position features are input into the self-attention layer, and different preset semantic tokens are collected for global classification to predict the type of myocardial injury; the T2 embedding vector, T2* embedding vector and VNE embedding vector, T2 position features, T2* position features and VNE position features are mapped and associated to obtain T2 visual features, T2* visual features and VNE visual features.
[0083] Specifically, the codec performs segmentation prediction on the auxiliary features and T2 visual features, T2* visual features and VNE visual features to obtain the myocardial tissue damage area. The specific process includes: inputting the auxiliary features and T2 visual features, T2* visual features and VNE visual features into the feature fusion module to perform feature fusion and feature dimension reduction on the auxiliary features with the T2 visual features, T2* visual features and VNE visual features respectively to obtain T2 fusion features, T2* fusion features and VNE fusion features; the codec splices the T2 image features, T2* image features and VNE image features with the T2 fusion features, T2* fusion features and VNE fusion features to obtain T2 splicing features, T2* splicing features and VNE splicing features; and performing feature decoding on the T2 splicing features, T2* splicing features and VNE splicing features to obtain the myocardial tissue damage area.
[0084] For example, Figure 2The model network structure is a convolution-transformer hybrid structure, which consists of three stages. In the first stage, a convolutional neural network (CNN) is used to extract multi-scale features specific to the sequence type for multi-sequence inputs. In the second stage, a vision transformer is used to perform multi-sequence deep feature fusion, which explores the correlation and difference of image features between different sequences through the self-attention mechanism. At the same time, by adding different semantic tokens, global classification based on multi-sequence inputs is achieved, and tissue feature damage is predicted to achieve automatic severity grading. In the third stage, an encoder-decoder structure is constructed. The main input sequence features obtained from the vision transformer and the auxiliary input sequence features obtained using a lightweight CNN are fused and reduced in the feature fusion module. After that, the multi-scale features obtained in the first stage are used through upsampling calculations and CNN structures combined with skip connections to perform sequence-specific automatic recognition of myocardial tissue damage areas. According to the recognition results, relevant quantitative analysis can be performed.
[0085] S103, outputting the myocardial tissue damage area and myocardial damage type corresponding to the historical multi-series medical images;
[0086] Myocardial tissue damage refers to areas that show abnormalities in cardiac MRI images and may be affected by pathological processes. Myocardial injury types refer to specific classifications of myocardial damage, including but not limited to edema, necrosis, microcirculatory obstruction (MVO), and hemorrhage.
[0087] In some embodiments, the model performs classification prediction on multiple sequences of input medical images to determine whether a specific type of myocardial injury exists. Simultaneously, the model performs segmentation prediction, identifying and segmenting specific areas of myocardial injury. The model then outputs the myocardial tissue injury area and the type of myocardial injury.
[0088] S104 , determining the ischemia-reperfusion injury level of the myocardial tissue injury area according to the myocardial injury type and a preset CCS staging standard, wherein the preset CCS staging standard includes a mapping relationship between the myocardial injury type and the ischemia-reperfusion injury level.
[0089] Among them, the preset CCS staging standards include the mapping relationship between myocardial injury types and ischemia-reperfusion injury levels; the ischemia-reperfusion injury levels include the first ischemia-reperfusion injury level, the second ischemia-reperfusion injury level, the third ischemia-reperfusion injury level, and the fourth ischemia-reperfusion injury level. The myocardial injury types corresponding to the first ischemia-reperfusion injury level include edema, the myocardial injury types corresponding to the second ischemia-reperfusion injury level include edema and myocardial necrosis, the myocardial injury types corresponding to the third ischemia-reperfusion injury level include edema, myocardial necrosis and microcirculation obstruction, and the myocardial injury types corresponding to the fourth ischemia-reperfusion injury level include intramyocardial hemorrhage.
[0090] In some embodiments of the present application, the specific process of determining the ischemia-reperfusion injury level of the myocardial tissue damage area according to the myocardial injury type and the preset CCS staging standard includes: obtaining the corresponding ischemia-reperfusion injury level from the mapping relationship according to the myocardial injury type.
[0091] For example Figure 3 As shown in the figure, different myocardial injury types in the preset CCS staging standard correspond to different levels of ischemia-reperfusion injury.
[0092] In the embodiments of the present application, on the one hand, the workload of doctors in image analysis can be significantly reduced through the automated injury level analysis model. The model is pre-trained to quickly perform classification predictions and segmentation predictions on historical multi-series medical images, thereby shortening the time from image acquisition to analysis result output. On the other hand, the automated injury level analysis model can learn and identify subtle features of myocardial injury through deep learning technology, and can accurately output the myocardial tissue damage area and damage type, thereby improving the accuracy of the ischemia-reperfusion injury level.
[0093] See Figure 4 , provides a flowchart of a damage level analysis model training method according to an embodiment of the present application. Figure 4 As shown, the method of the embodiment of the present application may include the following steps:
[0094] S201, collecting sample multi-sequence medical images of target subjects with and without ischemia-reperfusion injury;
[0095] Ischemia-reperfusion injury refers to the further damage to myocardial cells and microvasculature that may occur when blood flow is restored (reperfusion) after a period of insufficient blood flow (ischemia) to myocardial tissue. This damage may lead to myocardial cell death, microcirculatory obstruction (MVO), and intramyocardial hemorrhage. The target subject refers to the subject of research or diagnosis, namely the patient or subject. Sample multi-sequence medical images refer to a series of cardiac magnetic resonance imaging (MRI) data collected from the target subject. These data include images from different sequences, such as T1, T2, T1ρ, T2*, and VNE. These images are acquired using different MRI scan sequences and provide detailed information on myocardial tissue properties and signal changes. Samples with I / R injury refer to medical imaging data from patients or subjects whose cardiac MRI images show signs of I / R injury. These signs may include myocardial edema, necrosis, MVO, and hemorrhage. Samples without I / R injury refer to medical imaging data from patients or subjects whose cardiac MRI images do not show signs of I / R injury. These samples serve as a control group to help distinguish normal from damaged myocardial tissue.
[0096] S202, creating a hybrid structure model, the hybrid structure model including a convolutional neural network, a visual transformer, and an encoder-decoder, the visual transformer including a sequence embedding layer, a position embedding layer, and a self-attention layer, and the encoder-decoder including a feature fusion module and a segmentation prediction module;
[0097] S203, labeling the sample multi-series medical images with myocardial injury type labels to obtain a first training sample for the classification task, where the myocardial injury type labels include whether edema, myocardial necrosis, microcirculatory obstruction, and intramyocardial hemorrhage are present;
[0098] Among them, myocardial injury type labels are used to describe and classify the types of myocardial injuries. They serve as data annotations for model training to help the model learn to identify different myocardial injuries.
[0099] In some embodiments, quality control is performed on the collected MRI images, including denoising, correction of motion artifacts and magnetic field inhomogeneities, and the MRI images are analyzed to identify the type of myocardial injury. The images are annotated with corresponding injury type labels, such as edema, myocardial necrosis, MVO, and intramyocardial hemorrhage. The annotated image dataset is used as the first training sample for classification task training of the model.
[0100] S204, segmenting the myocardial tissue damage area in the sample multi-series medical image to obtain a second training sample for the segmentation task;
[0101] The sample multi-sequence medical images include historical T2 images, historical T2* images, and historical VNE images.
[0102] In some embodiments, the specific process of segmenting the myocardial tissue damage area present in the multi-series medical images of the sample to obtain the second training sample for the segmentation task includes: marking the edema area in the historical T2 image as the region of interest; marking the intramyocardial hemorrhage area in the historical T2* image as the region of interest; marking the myocardial necrosis area and the microcirculation obstruction area in the historical VNE image as the region of interest; and using the historical T2 images, historical T2* images and historical VNE images with the marked regions of interest as the second training sample for the segmentation task.
[0103] S205: Generate a pre-trained damage level analysis model based on the first training sample and the second training sample.
[0104] In some embodiments of the present application, the specific process of generating a pre-trained damage level analysis model based on the first training sample and the second training sample includes: using the multi-label cross entropy loss function as the first loss function of the classification task; using the Dice loss function as the second loss function of the segmentation task; integrating the first loss function and the second loss function into the hybrid structure model to obtain the damage level analysis model; training the damage level analysis model based on the first training sample and the second training sample to obtain a pre-trained damage level analysis model.
[0105] Specifically, the damage level analysis model is trained according to the first training sample and the second training sample. The specific process of obtaining the pre-trained damage level analysis model includes: inputting the first training sample and the second training sample into the damage level analysis model to determine the first model loss value according to the first loss function, and determining the second model loss value according to the second loss function; when the first model loss value and the second model loss value reach the minimum, a pre-trained damage level analysis model is generated.
[0106] In the embodiments of the present application, on the one hand, the workload of doctors in image analysis can be significantly reduced through the automated injury level analysis model. The model is pre-trained to quickly perform classification predictions and segmentation predictions on historical multi-series medical images, thereby shortening the time from image acquisition to analysis result output. On the other hand, the automated injury level analysis model can learn and identify subtle features of myocardial injury through deep learning technology, and can accurately output the myocardial tissue damage area and damage type, thereby improving the accuracy of the ischemia-reperfusion injury level.
[0107] 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.
[0108] See Figure 5 , which shows a schematic diagram of the structure of an ischemia-reperfusion injury level analysis device provided by an exemplary embodiment of the present application. The ischemia-reperfusion injury level 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 acquisition module 10, an input module 20, an output module 30, and a determination module 40.
[0109] An acquisition module 10 is used to acquire historical multi-series medical images of the myocardial tissue of the subject to be identified, generated based on nuclear magnetic resonance imaging;
[0110] An input module 20 is configured to input historical multi-series medical images into a pre-trained injury level analysis model to perform classification prediction and segmentation prediction on the historical multi-series medical images, wherein the pre-trained injury level analysis model is generated based on a first training sample for the classification task and a second training sample for the segmentation task;
[0111] An output module 30 is used to output the myocardial tissue damage area and myocardial damage type corresponding to the historical multi-series medical images;
[0112] The determination module 40 is used to determine the ischemia-reperfusion injury level of the myocardial tissue injury area according to the myocardial injury type and the preset CCS staging standard. The preset CCS staging standard includes a mapping relationship between the myocardial injury type and the ischemia-reperfusion injury level.
[0113] It should be noted that the ischemia-reperfusion injury level analysis device provided in the above embodiment, when executing the ischemia-reperfusion injury level analysis method, only uses the division of the above-mentioned functional modules as an example. In actual applications, 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 ischemia-reperfusion injury level analysis device provided in the above embodiment and the ischemia-reperfusion injury level analysis method embodiment are based on the same concept. The implementation process is detailed in the method embodiment and will not be repeated here.
[0114] 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.
[0115] In the embodiments of the present application, on the one hand, the workload of doctors in image analysis can be significantly reduced through the automated injury level analysis model. The model is pre-trained to quickly perform classification predictions and segmentation predictions on historical multi-series medical images, thereby shortening the time from image acquisition to analysis result output. On the other hand, the automated injury level analysis model can learn and identify subtle features of myocardial injury through deep learning technology, and can accurately output the myocardial tissue damage area and damage type, thereby improving the accuracy of the ischemia-reperfusion injury level.
[0116] The present application also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implements the ischemia-reperfusion injury level analysis method provided by each of the above method embodiments.
[0117] The present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the ischemia-reperfusion injury level analysis method of each of the above method embodiments.
[0118] See Figure 6, is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 6 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 .
[0119] The communication bus 1002 is used to implement the connection and communication between these components.
[0120] 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.
[0121] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).
[0122] 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.
[0123] 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 6 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an ischemia-reperfusion injury level analysis application.
[0124] exist Figure 6 In the electronic device 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain data input by the user; and the processor 1001 can be used to call the ischemia reperfusion injury level analysis application stored in the memory 1005 and specifically perform the following operations:
[0125] Acquire historical multi-series medical images of the myocardial tissue of the object to be identified, generated based on nuclear magnetic resonance;
[0126] Inputting historical multi-series medical images into a pre-trained injury level analysis model to perform classification prediction and segmentation prediction on the historical multi-series medical images, wherein the pre-trained injury level analysis model is generated based on a first training sample for a classification task and a second training sample for a segmentation task;
[0127] Output the myocardial tissue damage area and myocardial damage type corresponding to historical multi-series medical images;
[0128] The ischemia-reperfusion injury level of the myocardial tissue injury area is determined based on the myocardial injury type and the preset CCS staging criteria. The preset CCS staging criteria include a mapping relationship between the myocardial injury type and the ischemia-reperfusion injury level.
[0129] In one embodiment, when performing classification prediction and segmentation prediction on historical multi-series medical images, the processor 1001 specifically performs the following operations:
[0130] Convolutional neural network performs multi-scale feature extraction on T1 images, T2 images, T1ρ images, T2* images and VNE images to obtain T1 image features, T2 image features, T1ρ image features, T2* image features and VNE image features;
[0131] Use T1 image features and T1ρ image features as auxiliary features;
[0132] The visual transformer classifies and predicts T2 image features, T2* image features, and VNE image features to obtain myocardial injury type, T2 visual features, T2* visual features, and VNE visual features;
[0133] The codec performs segmentation prediction on the auxiliary features, T2 visual features, T2* visual features, and VNE visual features to obtain the myocardial tissue damage area.
[0134] In one embodiment, when the processor 1001 uses the visual transformer to perform classification prediction on the T2 image features, the T2* image features, and the VNE image features to obtain the myocardial injury type, the T2 visual features, the T2* visual features, and the VNE visual features, the processor 1001 specifically performs the following operations:
[0135] The sequence embedding layer maps T2 image features, T2* image features, and VNE image features into low-dimensional feature vectors to capture the similarity between pixels, and obtains T2 embedding vectors, T2* embedding vectors, and VNE embedding vectors.
[0136] The position embedding layer extracts the relative position features of different pixel positions from the T2 image features, T2* image features, and VNE image features to obtain T2 position features, T2* position features, and VNE position features;
[0137] The T2 embedding vector, T2* embedding vector, VNE embedding vector, T2 position feature, T2* position feature, and VNE position feature are input into the self-attention layer, and different preset semantic tokens are collected for global classification to predict the type of myocardial injury;
[0138] The T2 embedding vector, T2* embedding vector, VNE embedding vector, T2 position feature, T2* position feature, and VNE position feature are mapped and associated to obtain T2 visual feature, T2* visual feature, and VNE visual feature.
[0139] In one embodiment, when the processor 1001 executes the codec to perform segmentation prediction on the auxiliary features and the T2 visual features, the T2* visual features, and the VNE visual features to obtain the myocardial tissue damage area, the following operations are specifically performed:
[0140] Input the auxiliary features, T2 visual features, T2* visual features, and VNE visual features into the feature fusion module to perform feature fusion and feature dimensionality reduction on the auxiliary features with the T2 visual features, T2* visual features, and VNE visual features, respectively, to obtain T2 fusion features, T2* fusion features, and VNE fusion features;
[0141] The codec concatenates the T2 image features, T2* image features, and VNE image features with the T2 fusion features, T2* fusion features, and VNE fusion features to obtain T2 concatenation features, T2* concatenation features, and VNE concatenation features;
[0142] The T2 splicing features, T2* splicing features and VNE splicing features are decoded to obtain the myocardial tissue damage area.
[0143] In one embodiment, when the processor 1001 determines the ischemia-reperfusion injury level of the myocardial tissue injury area according to the myocardial injury type and the preset CCS staging standard, the processor 1001 specifically performs the following operations:
[0144] According to the type of myocardial injury, the corresponding ischemia-reperfusion injury level is obtained from the mapping relationship.
[0145] In one embodiment, when generating a pre-trained damage level analysis model, the processor 1001 specifically performs the following operations:
[0146] Collecting sample multi-sequence medical images of target subjects with and without ischemia-reperfusion injury;
[0147] Create a hybrid structure model, which includes a convolutional neural network, a visual transformer, and an encoder-decoder. The visual transformer includes a sequence embedding layer, a position embedding layer, and a self-attention layer. The encoder-decoder includes a feature fusion module and a segmentation prediction module.
[0148] Labeling the myocardial injury type labels on the sample multi-sequence medical images to obtain the first training sample for the classification task. The myocardial injury type labels include whether there is edema, whether there is myocardial necrosis, whether there is microcirculation obstruction, and whether there is intramyocardial hemorrhage;
[0149] Segmenting the myocardial tissue damage area in the sample multi-series medical image to obtain a second training sample for the segmentation task;
[0150] A pre-trained damage level analysis model is generated based on the first training sample and the second training sample.
[0151] In one embodiment, when the processor 1001 performs segmentation of the myocardial tissue damage area in the sample multi-series medical image to obtain the second training sample for the segmentation task, the processor 1001 specifically performs the following operations:
[0152] In the historical T2 images, the edema area was marked as the region of interest;
[0153] In the historical T2* images, the intramyocardial hemorrhage area was marked as the region of interest;
[0154] In the historical VNE images, the myocardial necrosis area and the microcirculatory obstruction area were marked as the region of interest;
[0155] The historical T2 images, historical T2* images, and historical VNE images with marked regions of interest are used as the second training samples for the segmentation task.
[0156] In one embodiment, when the processor 1001 generates a pre-trained damage level analysis model based on the first training sample and the second training sample, the processor 1001 specifically performs the following operations:
[0157] Use the multi-label cross entropy loss function as the first loss function for the classification task;
[0158] Use the Dice loss function as the second loss function for the segmentation task;
[0159] The first loss function and the second loss function are integrated into the hybrid structural model to obtain a damage level analysis model;
[0160] The damage level analysis model is trained according to the first training sample and the second training sample to obtain a pre-trained damage level analysis model.
[0161] In one embodiment, when the processor 1001 trains the damage level analysis model based on the first training sample and the second training sample to obtain the pre-trained damage level analysis model, the processor 1001 specifically performs the following operations:
[0162] Inputting the first training sample and the second training sample into the damage level analysis model to determine a first model loss value according to a first loss function, and to determine a second model loss value according to a second loss function;
[0163] When the first model loss value and the second model loss value reach a minimum, a pre-trained damage level analysis model is generated.
[0164] In the embodiments of the present application, on the one hand, the workload of doctors in image analysis can be significantly reduced through the automated injury level analysis model. The model is pre-trained to quickly perform classification predictions and segmentation predictions on historical multi-series medical images, thereby shortening the time from image acquisition to analysis result output. On the other hand, the automated injury level analysis model can learn and identify subtle features of myocardial injury through deep learning technology, and can accurately output the myocardial tissue damage area and damage type, thereby improving the accuracy of the ischemia-reperfusion injury level.
[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 related hardware through a computer program. The program for analyzing the level of ischemia-reperfusion injury 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 the level of ischemia-reperfusion injury 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 the level of ischemia-reperfusion injury, characterized in that: The method comprises: Acquire historical multi-series medical images of the myocardial tissue of the object to be identified, generated based on nuclear magnetic resonance; Inputting the historical multi-series medical images into a pre-trained injury level analysis model to perform classification prediction and segmentation prediction on the historical multi-series medical images, wherein the pre-trained injury level analysis model is generated based on a first training sample for a classification task and a second training sample for a segmentation task; Outputting the myocardial tissue damage area and myocardial damage type corresponding to the historical multi-series medical images; The ischemia-reperfusion injury level of the myocardial tissue injury area is determined according to the myocardial injury type and a preset CCS staging standard, wherein the preset CCS staging standard includes a mapping relationship between the myocardial injury type and the ischemia-reperfusion injury level.
2. The method according to claim 1, characterized in that The historical multi-series medical images include T1 images, T2 images, T1ρ images, T2* images, and VNE images. The T1 images, T2 images, T1ρ images, and T2* images are obtained through different magnetic resonance scanning sequences. The VNE images are virtual late gadolinium-enhanced images obtained by processing MTT images through a preset deep neural network. The MTT images are obtained by measuring the flow time of a contrast agent in myocardial tissue to evaluate myocardial blood perfusion information. The pre-trained damage level analysis model includes a convolutional neural network, a visual transformer, and an encoder-decoder; The performing classification prediction and segmentation prediction on the historical multi-sequence medical images includes: The convolutional neural network performs multi-scale feature extraction on the T1 image, T2 image, T1ρ image, T2* image and VNE image to obtain T1 image features, T2 image features, T1ρ image features, T2* image features and VNE image features; Using the T1 image features and T1ρ image features as auxiliary features; The visual transformer classifies and predicts the T2 image features, T2* image features, and VNE image features to obtain myocardial injury types, T2 visual features, T2* visual features, and VNE visual features; The codec performs segmentation prediction on the auxiliary features and the T2 visual features, the T2* visual features, and the VNE visual features to obtain a myocardial tissue damage area.
3. The method according to claim 2, characterized in that The visual transformer includes a sequence embedding layer, a position embedding layer, and a self-attention layer; The method of using the visual transformer to classify and predict the T2 image features, T2* image features, and VNE image features to obtain myocardial injury types, T2 visual features, T2* visual features, and VNE visual features includes: The sequence embedding layer maps the T2 image features, T2* image features, and VNE image features into low-dimensional feature vectors to capture the similarity between pixels, thereby obtaining a T2 embedding vector, a T2* embedding vector, and a VNE embedding vector; The position embedding layer extracts relative position features of different pixel positions from the T2 image features, T2* image features and VNE image features to obtain T2 position features, T2* position features and VNE position features; Inputting the T2 embedding vector, T2* embedding vector, VNE embedding vector, T2 position feature, T2* position feature, and VNE position feature into the self-attention layer, and performing global classification by combining preset different semantic tokens to predict the type of myocardial injury; The T2 embedding vector, the T2* embedding vector, the VNE embedding vector, the T2 position feature, the T2* position feature, and the VNE position feature are mapped and associated to obtain the T2 visual feature, the T2* visual feature, and the VNE visual feature.
4. The method according to claim 2, characterized in that The codec includes a feature fusion module and a segmentation prediction module; The codec performs segmentation prediction on the auxiliary features and the T2 visual features, the T2* visual features, and the VNE visual features to obtain a myocardial tissue damage area, including: Inputting the auxiliary features and the T2 visual features, T2* visual features, and VNE visual features into the feature fusion module to perform feature fusion and feature dimensionality reduction on the auxiliary features with the T2 visual features, T2* visual features, and VNE visual features, respectively, to obtain T2 fusion features, T2* fusion features, and VNE fusion features; The codec splices the T2 image feature, T2* image feature and VNE image feature with the T2 fusion feature, T2* fusion feature and VNE fusion feature to obtain a T2 splicing feature, a T2* splicing feature and a VNE splicing feature; Feature decoding is performed on the T2 splicing features, T2* splicing features, and VNE splicing features to obtain the myocardial tissue damage area.
5. The method according to claim 1, wherein The preset CCS staging standard includes a mapping relationship between myocardial injury types and ischemia-reperfusion injury levels; the ischemia-reperfusion injury levels include a first ischemia-reperfusion injury level, a second ischemia-reperfusion injury level, a third ischemia-reperfusion injury level, and a fourth ischemia-reperfusion injury level. The myocardial injury types corresponding to the first ischemia-reperfusion injury level include edema, the myocardial injury types corresponding to the second ischemia-reperfusion injury level include edema and myocardial necrosis, the myocardial injury types corresponding to the third ischemia-reperfusion injury level include edema, myocardial necrosis, and microcirculation obstruction, and the myocardial injury types corresponding to the fourth ischemia-reperfusion injury level include intramyocardial hemorrhage. Determining the ischemia-reperfusion injury level of the myocardial tissue injury area according to the myocardial injury type and the preset CCS staging standard includes: According to the myocardial injury type, the corresponding ischemia-reperfusion injury level is obtained from the mapping relationship.
6. The method according to claim 1, wherein The following steps are used to generate a pre-trained damage level analysis model, including: Collecting sample multi-sequence medical images of target subjects with and without ischemia-reperfusion injury; Creating a hybrid structure model, the hybrid structure model includes a convolutional neural network, a visual transformer, and an encoder-decoder, the visual transformer includes a sequence embedding layer, a position embedding layer, and a self-attention layer, and the encoder-decoder includes a feature fusion module and a segmentation prediction module; labeling the sample multi-series medical images with myocardial injury type labels to obtain a first training sample for a classification task, wherein the myocardial injury type labels include whether edema, whether myocardial necrosis, whether microcirculation obstruction, and whether intramyocardial hemorrhage exists; Segmenting the myocardial tissue damage area in the sample multi-series medical image to obtain a second training sample for the segmentation task; A pre-trained damage level analysis model is generated based on the first training sample and the second training sample.
7. The method according to claim 6, characterized in that The sample multi-sequence medical images include historical T2 images, historical T2* images and historical VNE images; The step of segmenting the myocardial tissue damage area in the sample multi-series medical image to obtain a second training sample for the segmentation task includes: In the historical T2 image, an edema area is marked as a region of interest; In the historical T2* image, marking the intramyocardial hemorrhage area as a region of interest; In the historical VNE image, the myocardial necrosis area and the microcirculation obstruction area are marked as regions of interest; The historical T2 images, historical T2* images and historical VNE images with the region of interest marked are used as second training samples for the segmentation task.
8. The method according to claim 6, characterized in that Generating a pre-trained damage level analysis model according to the first training sample and the second training sample includes: Using the multi-label cross entropy loss function as the first loss function of the classification task; Using the Dice loss function as the second loss function for the segmentation task; Integrating the first loss function and the second loss function into the hybrid structure model to obtain a damage level analysis model; The damage level analysis model is trained based on the first training sample and the second training sample to obtain a pre-trained damage level analysis model.
9. The method according to claim 8, characterized in that The step of training the damage level analysis model based on the first training sample and the second training sample to obtain a pre-trained damage level analysis model includes: Inputting the first training sample and the second training sample into the damage level analysis model to determine a first model loss value according to the first loss function and to determine a second model loss value according to the second loss function; When the first model loss value and the second model loss value reach a minimum, a pre-trained damage level analysis model is generated.
10. An ischemia-reperfusion injury level analysis device, characterized in that: The device comprises: An acquisition module, configured to acquire historical multi-series medical images of the myocardial tissue of the subject to be identified, generated based on nuclear magnetic resonance imaging; an input module, configured to input the historical multi-series medical images into a pre-trained injury level analysis model to perform classification prediction and segmentation prediction on the historical multi-series medical images, wherein the pre-trained injury level analysis model is generated based on a first training sample for a classification task and a second training sample for a segmentation task; An output module, configured to output the myocardial tissue damage area and myocardial damage type corresponding to the historical multi-series medical images; The determination module is used to determine the ischemia-reperfusion injury level of the myocardial tissue injury area according to the myocardial injury type and a preset CCS staging standard, wherein the preset CCS staging standard includes a mapping relationship between the myocardial injury type and the ischemia-reperfusion injury level.