Real-time left ventricular ejection fraction monitoring method and device based on artificial intelligence image analysis

Abstract

The invention discloses a real-time left ventricle ejection fraction monitoring method and device based on artificial intelligence image analysis, and relates to the field of image processing.The method comprises the steps that an X-ray perspective / angiography image sequence of a patient in the whole process of an operation is obtained, image segmentation is carried out, and a left ventricle profile diagram corresponding to each X-ray perspective / angiography image is obtained; constructing a training sample set according to each left ventricular profile diagram and the corresponding left ventricular ejection fraction; and training a machine learning model by using the left ventricular profile diagram in the training sample set and the corresponding left ventricular ejection fraction to obtain a left ventricular ejection fraction prediction model. According to the method, the left ventricular contour is extracted based on the X-ray perspective / angiography image, then the machine learning model is applied to predict the left ventricular ejection fraction, and the left ventricular ejection fraction of a patient can be monitored in real time in a cardiovascular disease operation.

Description

Technical Field

[0001] This application relates to the field of image processing, and in particular to a method and device for real-time monitoring of left ventricular ejection fraction based on artificial intelligence image analysis. Background Technology

[0002] Left ventricular ejection fraction (LVEF) is a core indicator for assessing cardiac pumping function and is crucial for the diagnosis, treatment planning, and prognostic evaluation of heart failure. In interventional cardiac procedures such as percutaneous coronary intervention (PCI) and transcatheter aortic valve replacement (TAVR), real-time and accurate assessment of cardiac function during the procedure is essential. However, current clinical practice primarily relies on preoperative transthoracic echocardiography (TTE), cardiac magnetic resonance imaging (CMR), or intraoperative transesophageal echocardiography (TEE). These methods have significant limitations: TTE cannot be performed intraoperatively; while TEE can be used intraoperatively, it is a semi-invasive procedure requiring a specialist sonographer and making truly continuous monitoring difficult; and CMR is completely unsuitable for use in a catheterization laboratory setting. Summary of the Invention

[0003] The purpose of this application is to provide a method and device for real-time monitoring of left ventricular ejection fraction based on artificial intelligence image analysis, which can monitor the left ventricular ejection fraction of patients in real time during cardiovascular surgery.

[0004] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides a method for real-time monitoring of left ventricular ejection fraction based on artificial intelligence image analysis, including: Obtain a sequence of X-ray fluoroscopy / contrast images of the patient throughout the entire surgical procedure; Image segmentation was performed on the X-ray fluoroscopy / contrast imaging sequence of the patient throughout the entire operation to obtain the left ventricular contour map corresponding to each X-ray fluoroscopy / contrast imaging image; A training sample set is constructed based on each left ventricular contour map and the corresponding left ventricular ejection fraction; the left ventricular ejection fraction corresponding to the left ventricular contour map includes the left ventricular ejection fraction monitored by transesophageal echocardiography during the operation and the left ventricular ejection fraction estimated by the area-length method or Simpson method based on the ventricular area. Using left ventricular contour maps and corresponding left ventricular ejection fractions from the training sample set, a machine learning model is trained to derive a left ventricular ejection fraction prediction model. This model is used to output the corresponding left ventricular ejection fraction when a patient's left ventricular contour map is input, so as to monitor the left ventricular ejection fraction in real time during the operation.

[0005] Secondly, this application provides a real-time left ventricular ejection fraction monitoring device based on artificial intelligence image analysis, comprising: The image acquisition module is used to acquire X-ray fluoroscopy / contrast image sequences of the patient throughout the entire surgical process; The image segmentation module is used to segment the X-ray fluoroscopy / contrast image sequence of the patient throughout the entire operation and obtain the left ventricular contour map corresponding to each X-ray fluoroscopy / contrast image; The training sample construction module is used to construct a training sample set based on each left ventricular contour map and the corresponding left ventricular ejection fraction. The left ventricular ejection fraction corresponding to the left ventricular contour map includes the left ventricular ejection fraction monitored by transesophageal echocardiography during the operation and the left ventricular ejection fraction estimated by the area-length method or Simpson method based on the ventricular area. The model training module is used to train a machine learning model using the left ventricular contour map and the corresponding left ventricular ejection fraction in the training sample set, and to obtain a left ventricular ejection fraction prediction model. The left ventricular ejection fraction prediction model is used to output the corresponding left ventricular ejection fraction when the patient's left ventricular contour map is input, so as to monitor the left ventricular ejection fraction in real time during the patient's operation.

[0006] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis.

[0007] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method and device for real-time left ventricular ejection fraction (LVEF) monitoring based on artificial intelligence image analysis. The method includes: acquiring a sequence of X-ray fluoroscopy / contrast imaging images of the patient throughout the entire surgical procedure, performing image segmentation to obtain a left ventricular contour map corresponding to each X-ray fluoroscopy / contrast imaging image; constructing a training sample set based on each LVEF contour map and its corresponding LVEF; and training a machine learning model using the LVEF contour maps and corresponding LVEFs in the training sample set to obtain a LVEF prediction model. This application, by extracting the left ventricular contour from X-ray fluoroscopy / contrast imaging images and then applying a machine learning model to predict the LVEF, enables real-time monitoring of the patient's LVEF during cardiovascular surgeries (PCI or TAVR procedures, etc.). It features real-time capability (direct intraoperative display of LVEF), continuity (continuous display via multi-channel monitor under DSA imaging), convenience (no other auxiliary equipment required), and safety (providing continuous "cardiac function navigation" for the surgeon, allowing for immediate detection and intervention of abnormalities). Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is an application environment diagram of a real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis in one embodiment of this application; Figure 2 A flowchart illustrating a real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis, provided as an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an image segmentation model provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a residual module provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an edge perception enhancement module provided in an embodiment of this application; Figure 6 A schematic diagram of the structure of a detail-enhancing convolutional layer provided in an embodiment of this application; Figure 7 A schematic diagram of the functional modules of a real-time left ventricular ejection fraction monitoring device based on artificial intelligence image analysis provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis provided in this application embodiment can be applied to, for example... Figure 1The application environment shown is illustrated. The terminal communicates with the server via a network. The data storage system stores the data the server needs to process. The data storage system can be set up independently, integrated into the server, or placed in the cloud or on another server. The terminal can send the entire intraoperative X-ray fluoroscopy / contrast imaging sequence and the patient's left ventricular ejection fraction (LVEF) monitored intermittently via intraoperative transesophageal echocardiography to the server. After receiving the data, the server acquires the entire intraoperative X-ray fluoroscopy / contrast imaging sequence; performs image segmentation on the entire intraoperative X-ray fluoroscopy / contrast imaging sequence to obtain the corresponding LVEF contour map for each X-ray fluoroscopy / contrast imaging image; and constructs a training sample set based on each LVEF contour map and its corresponding LVEF. The LVEF contour map... The corresponding left ventricular ejection fraction (LVEF) includes the LVEF monitored intermittently during surgery via intraoperative transesophageal echocardiography and the LVEF estimated based on ventricular area using the area-length method or the Simpson method. A machine learning model is trained using left ventricular contour maps and corresponding LVEFs from a training sample set to derive a LVEF prediction model. This model outputs the corresponding LVEF when a patient's left ventricular contour map is input, enabling real-time monitoring of the LVEF during surgery. The server can then feed the obtained LVEF prediction model back to the terminal. Furthermore, in some embodiments, the real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis can also be implemented independently by a server or a terminal. For example, the terminal can directly perform real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis on the X-ray fluoroscopy / contrast imaging sequence of the patient throughout the operation and the left ventricular ejection fraction of the patient intermittently monitored by intraoperative transesophageal echocardiography during the operation. Alternatively, the server can obtain the X-ray fluoroscopy / contrast imaging sequence of the patient throughout the operation and the left ventricular ejection fraction of the patient intermittently monitored by intraoperative transesophageal echocardiography from the data storage system and perform real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis.

[0013] The terminal can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. The server can be a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0014] In one exemplary embodiment, such as Figure 2 As shown, a method for real-time monitoring of left ventricular ejection fraction based on artificial intelligence image analysis is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1The following steps, 101 to 104, are used as an example to illustrate the process of using a server in the example.

[0015] Step 101: Obtain the X-ray fluoroscopy / contrast image sequence of the patient throughout the entire surgical process.

[0016] Step 102: Perform image segmentation on the X-ray fluoroscopy / contrast imaging sequence of the patient throughout the entire procedure to obtain the left ventricular contour map corresponding to each X-ray fluoroscopy / contrast imaging image.

[0017] Step 103: Construct a training sample set based on each left ventricular contour map and the corresponding left ventricular ejection fraction; the left ventricular ejection fraction corresponding to the left ventricular contour map includes the left ventricular ejection fraction monitored intermittently by intraoperative transesophageal echocardiography and the left ventricular ejection fraction estimated based on the ventricular area using the area-length method or the Simpson method.

[0018] Step 104: Using the left ventricular contour map and the corresponding left ventricular ejection fraction in the training sample set, train the machine learning model to obtain the left ventricular ejection fraction prediction model; the left ventricular ejection fraction prediction model is used to output the corresponding left ventricular ejection fraction when the patient's left ventricular contour map is input, so as to monitor the left ventricular ejection fraction in real time during the patient's operation.

[0019] Performing steps 101 to 104 above, cardiac interventional surgery is a dynamic process. Procedures such as balloon dilation, stent deployment, and valve implantation instantly alter the hemodynamic state of the heart, potentially leading to acute changes in cardiac function. Traditional intermittent assessment methods cannot capture these transient changes, making it difficult for surgeons to detect acute cardiac function deterioration caused by acute ischemia, coronary complications, conduction block, etc., thus missing the optimal intervention opportunity. Therefore, it is necessary to develop a real-time LVEF monitoring technology that can be seamlessly integrated into the interventional procedure. This technology extracts the left ventricular contour based on X-ray fluoroscopy / angiography images and then trains a machine learning model using the corresponding left ventricular ejection fraction to derive a left ventricular ejection fraction prediction model. This model can then be directly applied to predict the left ventricular ejection fraction, enabling dynamic real-time monitoring of LVEF under DSA imaging (angiography). It allows for continuous real-time monitoring of LVEF during interventional procedures using existing DSA images, without requiring additional invasive / non-invasive procedures. Real-time continuous monitoring of LVEF during interventional procedures can provide surgeons with continuous "cardiac function navigation," enabling them to detect acute cardiac function deterioration caused by acute ischemia, coronary complications, conduction block, or valvular regurgitation at the earliest possible time. This allows for intervention at the optimal time, achieving a leap from "anatomical success" to "functional success," and has revolutionary potential value in reducing surgical risks, optimizing surgical strategies, and improving patients' short- and long-term prognosis.

[0020] In another exemplary embodiment of this application, a training sample set is constructed based on each left ventricular contour map and the corresponding left ventricular ejection fraction, specifically including: (1) Obtain the left ventricular ejection fraction intermittently via intraoperative transesophageal echocardiography during cardiovascular surgery. Cardiovascular surgery includes TAVR, TEER, atrial fibrillation, etc.

[0021] (2) Select the first left ventricular contour map and construct a training sample set based on the first left ventricular contour map and the corresponding left ventricular ejection fraction; the first left ventricular contour map refers to the left ventricular contour map with left ventricular ejection fraction monitored by transesophageal echocardiography.

[0022] (3) For the second left ventricular contour map, the stroke volume and left ventricular ejection fraction are estimated by the area length method or Simpson method based on the ventricular area; and the second left ventricular contour map and the corresponding left ventricular ejection fraction are added to the training sample set; the second left ventricular contour map refers to the left ventricular contour map without left ventricular ejection fraction monitored by transesophageal echocardiography.

[0023] In another exemplary embodiment of this application, the input to the left ventricular ejection fraction prediction model based on the machine learning model further includes physiological signals; the physiological signals include ECG signals and pressure waveforms. To ensure the accuracy of left ventricular ejection fraction prediction, prediction can be based not only on the left ventricular profile but also on multimodal data, incorporating physiological signals such as ECG signals and pressure waveforms.

[0024] In another exemplary embodiment of this application, step 104 involves training a machine learning model using left ventricular contour maps and corresponding left ventricular ejection fractions from the training sample set to derive a left ventricular ejection fraction prediction model. This specifically includes: (1) Use the left ventricular contour map and the corresponding left ventricular ejection fraction in the training sample set to train a variety of machine learning models.

[0025] Machine learning models that can be selected include: K-nearest neighbor model, support vector machine, multilayer perceptron, XGBoost, random forest, and LightGBM, etc.

[0026] (2) Evaluate the performance metrics of each trained machine learning model.

[0027] Performance metrics can include: AUC, PPV, NPV of the ROC curve, sensitivity, specificity, accuracy, and F1 score.

[0028] (3) Select the machine learning model with the best performance index after training as the left ventricular ejection fraction prediction model.

[0029] In another exemplary embodiment of this application, image segmentation is performed on the X-ray fluoroscopy / contrast imaging sequence of the patient throughout the entire surgical procedure to obtain the left ventricular contour map corresponding to each X-ray fluoroscopy / contrast imaging image, specifically including: The X-ray fluoroscopy / contrast imaging sequence of the patient throughout the entire operation is input into the image segmentation model to obtain the left ventricular contour map corresponding to each X-ray fluoroscopy / contrast imaging image.

[0030] The image segmentation model employs an improved U-Net network. For example... Figure 3 As shown, the improved U-Net network includes an encoder and a decoder.

[0031] The encoder includes a first coding layer and a second coding layer; the first coding layer and the second coding layer are arranged in a cross manner; the first coding layer includes a residual module and a CBAM module connected in sequence; the second coding layer includes a first convolution module and an EPEM module (edge ​​awareness enhancement module) connected in sequence.

[0032] The decoder consists of multiple decoding layers; each decoding layer consists of a second convolutional module and a cross-attention mechanism module (CA module) connected in sequence.

[0033] In this application, a residual module is used in the coding layer to replace the convolutional layers of U-Net, such as... Figure 4 As shown, this residual module mainly consists of two 3×3 convolutional layers, two 1×1 convolutional layers, a batch normalization (BN) layer, and a ReLU activation layer. It also performs one feature fusion operation and one feature addition operation, which addresses the degradation problem of each convolutional layer, extracts more image features, improves feature utilization, and mitigates channel dependencies. Furthermore, it incorporates a CBAM (Convolutional Block Attention) module, which combines a spatial attention mechanism (SAM) and a channel attention mechanism (CAM). These two attention mechanisms focus on feature information in the channel dimension and spatial dimension, respectively, to enhance the network's representational capabilities.

[0034] In another exemplary embodiment of this application, such as Figure 5 As shown, the EPEM module (edge ​​awareness enhancement module) includes: a first convolutional layer, a first normalization and activation layer, a second convolutional layer, a first activation layer, a detail enhancement convolutional layer, a second activation layer, a first additive layer, a third convolutional layer, a second normalization and activation layer, and a second additive layer, connected in sequence.

[0035] The input of the first additive layer is also connected to the output of the first activation layer; the input of the second additive layer is also connected to the input of the first convolutional layer.

[0036] The edge awareness enhancement module is based on convolutional neural networks, which can effectively capture local features such as edges and textures. It also incorporates detail enhancement convolution, which can improve the representation and generalization ability of ordinary convolution and further enhance the model's ability to extract defect textures and edge features.

[0037] Among them, such as Figure 6 As shown, the detail enhancement convolutional layer includes a third additive layer and a fourth convolutional layer connected in parallel, a central difference convolutional layer, an angular difference convolutional layer, a horizontal difference convolutional layer, and a vertical difference convolutional layer.

[0038] The inputs of the fourth convolutional layer, the central difference convolutional layer, the angular difference convolutional layer, the horizontal difference convolutional layer, and the vertical difference convolutional layer are all connected to the output of the first activation layer.

[0039] The outputs of the fourth convolutional layer, the central difference convolutional layer, the angular difference convolutional layer, the horizontal difference convolutional layer, and the vertical difference convolutional layer are all connected to the input of the third additive layer.

[0040] The output of the third addition layer is connected to the input of the second activation layer.

[0041] The detail-enhancing convolution borrows the idea of ​​differential convolution, deploying four differential convolutions (central differential convolution, corner differential convolution, horizontal differential convolution, and vertical differential convolution) and a regular convolution in parallel. The four different differential convolutions are used to enhance gradient-level information, while the regular convolution is used to obtain intensity-level information. By using reparameterization techniques, the detail-enhancing convolution is equivalently converted into a normal convolution without additional parameters and computational cost.

[0042] The number of encoding layers is the same as the number of decoding layers, and each encoding layer is connected to the corresponding decoding layer through skip connections.

[0043] The image segmentation model described above can solve the problems of noise and artifact processing in dynamic images, and address the challenge of machine learning models to perform continuous, stable, and automatic segmentation of cardiac chamber boundaries in complex environments such as low-dose X-rays, respiratory movements, cardiac pulsation, and uneven contrast agent filling.

[0044] This application also provides an application scenario in which the above-mentioned real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis is applied. Specifically, the real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis provided in this embodiment can be applied to the scenario of monitoring the left ventricular ejection fraction of patients during surgery. This scenario includes a data acquisition stage and a left ventricular ejection fraction monitoring stage; the data acquisition stage is used to acquire X-ray fluoroscopy / contrast imaging sequences of the patient throughout the entire surgical process and the left ventricular ejection fraction monitored intermittently by intraoperative transesophageal echocardiography; the left ventricular ejection fraction monitoring stage is used to train a left ventricular ejection fraction prediction model and thereby monitor the left ventricular ejection fraction in real time. The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis provided in this embodiment belongs to the left ventricular ejection fraction monitoring stage.

[0045] Based on the same inventive concept, this application also provides a device for real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis to implement the aforementioned method for real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the device for real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis provided below can be found in the limitations of the method for real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis described above, and will not be repeated here.

[0046] In one exemplary embodiment, such as Figure 7 As shown, a real-time left ventricular ejection fraction monitoring device based on artificial intelligence image analysis is provided, comprising: The image acquisition module M1 is used to acquire X-ray fluoroscopy / contrast image sequences of the patient throughout the entire surgical process.

[0047] The image segmentation module M2 is used to segment the X-ray fluoroscopy / contrast image sequence of the patient throughout the entire operation, and to obtain the left ventricular contour map corresponding to each X-ray fluoroscopy / contrast image.

[0048] The training sample construction module M3 is used to construct a training sample set based on each left ventricular contour map and the corresponding left ventricular ejection fraction. The left ventricular ejection fraction corresponding to the left ventricular contour map includes the left ventricular ejection fraction monitored intermittently by intraoperative transesophageal echocardiography and the left ventricular ejection fraction estimated based on the ventricular area using the area-length method or the Simpson method.

[0049] The model training module M4 is used to train a machine learning model using the left ventricular contour map and the corresponding left ventricular ejection fraction in the training sample set, and to obtain a left ventricular ejection fraction prediction model. The left ventricular ejection fraction prediction model is used to output the corresponding left ventricular ejection fraction when the patient's left ventricular contour map is input, so as to monitor the left ventricular ejection fraction in real time during the patient's operation.

[0050] This device can achieve data interface with DSA equipment, real-time processing, and result display.

[0051] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 8 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores real-time left ventricular ejection fraction monitoring data based on artificial intelligence image analysis. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for real-time left ventricular ejection fraction monitoring based on artificial intelligence image analysis.

[0052] Those skilled in the art will understand that Figure 8 The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0053] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0054] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0055] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0056] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0057] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0058] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis, characterized in that, The method comprises the following steps: obtaining a sequence of X-ray fluoroscopy / angiography images of a patient during an operation; performing image segmentation on the sequence of X-ray fluoroscopy / angiography images of the patient during the operation to obtain a left ventricular contour corresponding to each X-ray fluoroscopy / angiography image; constructing a training sample set according to each left ventricular contour and a corresponding left ventricular ejection fraction; the left ventricular ejection fraction corresponding to each left ventricular contour comprises an intermittent left ventricular ejection fraction monitored by transesophageal echocardiography during the operation of the patient and a left ventricular ejection fraction estimated by applying an area-length method or a Simpson method to a ventricular area; training a machine learning model by using the left ventricular contours and the corresponding left ventricular ejection fractions in the training sample set to obtain a left ventricular ejection fraction prediction model; the left ventricular ejection fraction prediction model is used to output a corresponding left ventricular ejection fraction when a left ventricular contour of the patient is input, so as to monitor the left ventricular ejection fraction of the patient in real time during the operation.

2. The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to claim 1, characterized in that, The method of constructing the training sample set according to each left ventricular contour and the corresponding left ventricular ejection fraction comprises the following steps: intermittently monitoring a left ventricular ejection fraction of the patient by transesophageal echocardiography during a cardiovascular disease operation; selecting a first left ventricular contour and constructing a training sample set according to the first left ventricular contour and the corresponding left ventricular ejection fraction; the first left ventricular contour refers to a left ventricular contour with a left ventricular ejection fraction monitored by transesophageal echocardiography; estimating a stroke volume and a left ventricular ejection fraction according to a ventricular area by applying an area-length method or a Simpson method to a second left ventricular contour, and adding the second left ventricular contour and the corresponding left ventricular ejection fraction to the training sample set; the second left ventricular contour refers to a left ventricular contour without a left ventricular ejection fraction monitored by transesophageal echocardiography. 3.The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to claim 1, wherein, The input of the left ventricular ejection fraction prediction model based on the machine learning model further comprises a physiological signal; the physiological signal comprises an ECG signal and a pressure waveform. 4.The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to claim 1, wherein, The method of training a machine learning model by using the left ventricular contours and the corresponding left ventricular ejection fractions in the training sample set to obtain a left ventricular ejection fraction prediction model comprises the following steps: training multiple machine learning models by using the left ventricular contours and the corresponding left ventricular ejection fractions in the training sample set; evaluating a performance index of each trained machine learning model; selecting a trained machine learning model with an optimal performance index as the left ventricular ejection fraction prediction model. 5.The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to claim 4, wherein, The performance index comprises at least one of an AUC index of an ROC curve, a PPV index, a NPV index, a sensitivity, a specificity, an accuracy rate and an F1 value; The machine learning model comprises a K-nearest neighbor model, a support vector machine, a multilayer perceptron, an XGBoost, a random forest and a LightGBM. 6.The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to claim 1, wherein, The method of performing image segmentation on the sequence of X-ray fluoroscopy / angiography images of the patient during the operation to obtain a left ventricular contour corresponding to each X-ray fluoroscopy / angiography image comprises the following steps: inputting the sequence of X-ray fluoroscopy / angiography images of the patient during the operation into an image segmentation model to obtain a left ventricular contour corresponding to each X-ray fluoroscopy / angiography image; wherein the image segmentation model adopts an improved U-Net network. The improved U-Net network comprises an encoder and a decoder; The encoder comprises a first encoding layer and a second encoding layer; the first encoding layer and the second encoding layer are arranged in a cross manner; the first encoding layer comprises a residual module and a CBAM module connected in sequence; and the second encoding layer comprises a first convolution module and an EPEM module connected in sequence. The decoder comprises a plurality of decoding layers; each decoding layer comprises a second convolution module and a cross attention mechanism module connected in sequence.

7. The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to claim 6, characterized in that, The EPEM module comprises a first convolution layer, a first normalization and activation layer, a second convolution layer, a first activation layer, a detail enhancement convolution layer, a second activation layer, a first addition layer, a third convolution layer, a second normalization and activation layer and a second addition layer connected in sequence. The input end of the first addition layer is also connected with the output end of the first activation layer; and the input end of the second addition layer is also connected with the input end of the first convolution layer.

8. The real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to claim 7, characterized in that, The detail enhancement convolution layer comprises a third addition layer and a fourth convolution layer, a center difference convolution layer, an angle difference convolution layer, a horizontal difference convolution layer and a vertical difference convolution layer connected in parallel. The input ends of the fourth convolution layer, the center difference convolution layer, the angle difference convolution layer, the horizontal difference convolution layer and the vertical difference convolution layer are all connected with the output end of the first activation layer. The output ends of the fourth convolution layer, the center difference convolution layer, the angle difference convolution layer, the horizontal difference convolution layer and the vertical difference convolution layer are all connected with the input end of the third addition layer. The output end of the third addition layer is connected with the input end of the second activation layer.

9. A real-time left ventricular ejection fraction monitoring device based on artificial intelligence image analysis, characterized by, The method comprises the steps of: an image acquisition module, configured to acquire an X-ray fluoroscopy / angiography image sequence of a patient during an entire operation process; an image segmentation module, configured to perform image segmentation on the X-ray fluoroscopy / angiography image sequence of the patient during the entire operation process, and obtain a left ventricular contour corresponding to each X-ray fluoroscopy / angiography image; a training sample construction module, configured to construct a training sample set according to each left ventricular contour and a corresponding left ventricular ejection fraction; the left ventricular ejection fraction corresponding to the left ventricular contour comprises a left ventricular ejection fraction intermittently monitored by an intraoperative transesophageal echocardiography during the operation of the patient and a left ventricular ejection fraction estimated by applying an area-length method or a Simpson method to a ventricular area; a model training module, configured to train a machine learning model by using the left ventricular contour and the corresponding left ventricular ejection fraction in the training sample set, and obtain a left ventricular ejection fraction prediction model; the left ventricular ejection fraction prediction model is configured to output the corresponding left ventricular ejection fraction when the left ventricular contour of the patient is input, so as to monitor the left ventricular ejection fraction of the patient in real time during the operation.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by a processor to implement the real-time left ventricular ejection fraction monitoring method based on artificial intelligence image analysis according to any one of claims 1-8.

Images