Myocardial focal scar detection method, risk prediction method and related apparatus

By applying the automatic segmentation and positioning model of myocardial muscle on the CCTA scan image, the rapid and low-cost detection and risk prediction of myocardial focal scars are solved, and the problems of long detection, high cost and large radiation dose in the prior art are solved. Rapid and non-invasive detection and risk assessment methods for focal myocardial scars.

WO2024230714A1PCT designated stage expired Publication Date: 2024-11-14THE FIFTH AFFILIATED HOSPITAL SUN YAT SEN UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2024/091574
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-08
Filing Date
2024-05-08
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

The prior art is used to detect focal scars of myocardial muscles, which have problems such as difficult operation, long detection time, high cost, large amount of iodine contrast agents and high radiation dose.

Method used

The preset left ventricular wall myocardium automatic segmentation model and myocardial focal scar positioning model were used to voxel binary classification and tricategorize the single-energy or double-energy CCTA scan images to achieve rapid and effective detection of myocardium focal scars. And risk prediction is carried out by extracting imaging features and fusion of multidimensional medical data.

Benefits of technology

It realizes rapid, non-invasive and low-cost detection of focal scars in myocardial heart, reduces the dosage of iodine contrast agents and radiation dose, can reflect the overall change of myocardial muscles, and conducts risk prediction, solving the shortcomings of the existing technology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2024091574_14112024_PF_FP_ABST
    Figure CN2024091574_14112024_PF_FP_ABST
Patent Text Reader

Abstract

A myocardial focal scar detection method, a risk prediction method and a related apparatus, relating to the technical field of medical image analyses. The detection method comprises: on the basis of a preset automatic left ventricular myocardium segmentation model, performing voxel binary classification segmentation on a myocardial image scanned by single-energy CCTA or dual-energy CCTA, to obtain a left ventricular myocardium image; and performing voxel ternary classification segmentation on a low attenuation lesion of the left ventricular myocardium image on the basis of a preset myocardial focal scar localization model, to obtain a myocardial focal scar localization image. The technical problems of difficult operation technology, long detection time, high cost, large consumption of an iodine contrast agent and high radiation dose in the prior art are solved.
Need to check novelty before this filing date? Find Prior Art

Description

A myocardial focal scar detection method, risk prediction method and related device

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on May 8, 2023, with application number 202310505772.X and invention name “A method for detecting focal myocardial scars, a risk prediction method and related devices”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present invention relates to the technical field of medical image analysis, and in particular to a myocardial focal scar detection method, a risk prediction method and related devices. Background Art

[0003] Myocardial focal scar (MFS) is the replacement of fibrous tissue after myocardial trauma. MFS is present in the pathological changes of many cardiomyopathies, such as myocardial infarction (MI), hypertrophic cardiomyopathy, sarcoidosis, myocarditis, and uremic cardiomyopathy.

[0004] The existing technologies currently used to detect focal myocardial scars include subendocardial puncture biopsy, cardiac magnetic resonance imaging (CMR) delayed gadolinium contrast enhancement imaging (LGE), and late iodine enhancement (LIE) based on coronary CT angiography (CCTA). However, subendocardial puncture biopsy is technically difficult to perform, places extremely high demands on the operator, and cannot reflect the overall changes in the myocardium. Cardiac magnetic resonance delayed gadolinium contrast enhancement imaging technology is not only technically difficult to perform, but also time-consuming and expensive. The delayed iodine contrast enhancement technology based on coronary CT angiography has the problems of large iodine contrast agent usage and high radiation dose. Therefore, researching new methods for detecting focal myocardial scars to solve the above technical problems is a technical problem that needs to be solved urgently by those skilled in the art.

[0005] Summary of the Invention

[0006] The present invention provides a myocardial focal scar detection method, a risk prediction method and related devices, which are used to solve the technical problems of the existing technology, such as high difficulty in operation technology, long detection time, high cost, large amount of iodine contrast agent used and high radiation dose.

[0007] In view of this, the first aspect of the present invention provides a method for detecting focal myocardial scars, comprising:

[0008] Based on a preset left ventricular wall myocardial automatic segmentation model, a voxel-based binary classification segmentation is performed on the myocardial image scanned by single-energy CCTA or dual-energy CCTA to obtain a left ventricular wall myocardial image, wherein the voxel-based binary classification segmentation result includes left ventricular wall myocardial tissue and non-left ventricular wall myocardial tissue;

[0009] Based on the preset myocardial focal scar localization model, the low-density lesions in the left ventricular wall myocardial image are segmented by voxel three-category classification to obtain a myocardial focal scar localization image. The voxel three-category segmentation result includes myocardial focal scars, myocardial radiation hardening artifacts and normal myocardium.

[0010] Optionally, based on a preset left ventricular wall myocardial automatic segmentation model, voxel binary classification segmentation is performed on the myocardial image scanned based on single-energy CCTA or dual-energy CCTA to obtain a left ventricular wall myocardial image, which also includes:

[0011] Construct a left ventricular wall myocardial annotation dataset based on single-energy CCTA scan images and dual-energy CCTA scan images;

[0012] The left ventricular wall myocardium annotation dataset is used to train, verify and test the left ventricular wall myocardium automatic segmentation model based on voxel binary classification segmentation, and a preset left ventricular wall myocardium automatic segmentation model is obtained.

[0013] Optionally, based on a preset myocardial focal scar localization model, voxel-based three-category segmentation is performed on the low-density lesions in the left ventricular wall myocardial image to obtain a myocardial focal scar localization image, which also includes:

[0014] Based on left ventricular wall myocardial images annotated with focal myocardial scars and myocardial radiation hardening artifacts, a dataset of annotations of focal myocardial scar artifacts and radiation hardening artifacts was constructed.

[0015] The myocardial focal scar artifact and radiation hardening artifact annotation datasets were used to train, validate, and test the focal scar localization model based on voxel-based three-category segmentation, and a preset myocardial focal scar localization model was obtained.

[0016] Optionally, based on a preset myocardial focal scar localization model, voxel-based three-category segmentation is performed on the low-density lesions in the left ventricular wall myocardial image to obtain a myocardial focal scar localization image, which also includes:

[0017] Based on left ventricular wall myocardial images annotated with focal myocardial scars and myocardial radiation hardening artifacts, a dataset of annotations of focal myocardial scar artifacts and radiation hardening artifacts was constructed.

[0018] The myocardial focal scar artifact and radiation hardening artifact annotation datasets were used to train, validate, and test the focal scar localization model based on voxel-based three-category segmentation, and a preset myocardial focal scar localization model was obtained.

[0019] Optionally, performing voxel three-category segmentation on the low-density lesions in the left ventricular wall myocardial image based on a preset myocardial focal scar localization model to obtain a myocardial focal scar localization image includes:

[0020] Based on the preset myocardial focal scar localization model, the low-density lesions in the left ventricular wall myocardial image are segmented by three-dimensional voxel classification to obtain the myocardial focal scar image;

[0021] The volume of myocardial focal scar is calculated from the myocardial focal scar image to obtain quantitative information of myocardial focal scar;

[0022] Based on the quantitative information of myocardial focal scars and the myocardial segmentation information, the myocardial focal scars are located to obtain a myocardial focal scar localization image.

[0023] A second aspect of the present invention provides a risk prediction method, comprising any one of the myocardial focal scar detection methods described in the first aspect, further comprising:

[0024] Extract imaging features of myocardial focal scar localization images;

[0025] Based on the relationship between imaging features of focal myocardial scar localization images and major adverse cardiovascular events, risk factors for major adverse cardiovascular events based on single-energy CCTA images or dual-energy CCTA images were screened;

[0026] A preset risk prediction model was used to perform risk stratification prediction of major adverse cardiovascular events based on major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results.

[0027] Optionally, imaging features of the myocardial focal scar localization image are extracted, including:

[0028] Based on the myocardial focal scar localization image, multiple imaging features are extracted, and normalized, dimensionally reduced, and de-redundanted to obtain the imaging features of the myocardial focal scar localization image.

[0029] Optionally, imaging features of the myocardial focal scar localization image are extracted, including:

[0030] Based on the myocardial focal scar localization image, multiple imaging features are extracted, and visual evaluation, normalization, dimensionality reduction and redundancy removal are performed on the multiple imaging features to obtain the imaging features of the myocardial focal scar localization image. Among them, the multiple imaging features include imaging features visually evaluated by the human eye and imaging omics features based on computer calculation and recognition.

[0031] A third aspect of the present invention provides a device for detecting focal myocardial scars, comprising:

[0032] A left ventricular wall myocardial image acquisition module is used to perform voxel-based binary classification segmentation on the myocardial image scanned based on single-energy CCTA or dual-energy CCTA based on a preset left ventricular wall myocardial automatic segmentation model to obtain a left ventricular wall myocardial image, wherein the voxel-based binary classification segmentation result includes left ventricular wall myocardial tissue and non-left ventricular wall myocardial tissue;

[0033] The focal scar localization module is used to perform voxel three-category segmentation on the low-density lesions in the left ventricular wall myocardial image based on a preset myocardial focal scar localization model to obtain a myocardial focal scar localization image, wherein the voxel three-category segmentation result includes myocardial focal scars, myocardial radiation hardening artifacts and normal myocardium.

[0034] Optionally, it further includes a first model building module;

[0035] The first model building module is used to:

[0036] Construct a left ventricular wall myocardial annotation dataset based on single-energy CCTA scan images and dual-energy CCTA scan images;

[0037] The left ventricular wall myocardium annotation dataset is used to train, verify and test the left ventricular wall myocardium automatic segmentation model based on voxel binary classification segmentation, and a preset left ventricular wall myocardium automatic segmentation model is obtained.

[0038] Optionally, a second model building module is further included;

[0039] The second model building module is used to:

[0040] Based on left ventricular wall myocardial images annotated with focal myocardial scars and myocardial radiation hardening artifacts, a dataset of annotations of focal myocardial scar artifacts and radiation hardening artifacts was constructed.

[0041] The myocardial focal scar artifact and radiation hardening artifact annotation datasets were used to train, validate, and test the focal scar localization model based on voxel-based three-category segmentation, and a preset myocardial focal scar localization model was obtained.

[0042] A fourth aspect of the present invention provides a risk prediction device, comprising the myocardial focal scar detection device according to any one of the second aspects, further comprising:

[0043] A feature extraction module is used to extract imaging features of myocardial focal scar localization images;

[0044] The risk factor screening module is used to screen the risk factors for major adverse cardiovascular events based on single-energy CCTA images or dual-energy CCTA images based on the relationship between the imaging characteristics of myocardial focal scar localization images and major adverse cardiovascular events;

[0045] The risk stratification prediction module is used to use a preset risk prediction model to perform risk stratification prediction of major adverse cardiovascular events based on major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results.

[0046] It can be seen from the above technical solutions that the method for detecting focal myocardial scars provided by the present invention has the following advantages:

[0047] The method for detecting focal myocardial scars provided by the present invention first performs voxel-based binary segmentation on myocardial images scanned using single-energy CCTA or dual-energy CCTA based on a preset left ventricular wall myocardial automatic segmentation model to obtain a left ventricular wall myocardial image. It then performs voxel-based three-class segmentation on low-density lesions in the left ventricular wall myocardial image based on a preset focal myocardial scar localization model to obtain a focal myocardial scar localization image. This method achieves rapid and effective detection of focal myocardial scars.

[0048] Compared with subendocardial puncture biopsy, the myocardial focal scar detection method provided by the present invention has the advantages of being non-invasive, less difficult to operate, and able to reflect the overall changes in the myocardium.

[0049] Compared with the cardiac magnetic resonance gadolinium contrast agent delayed enhancement imaging technology, the myocardial focal scar detection method provided by the present invention has the advantages of less operating difficulty, shorter detection time and lower cost.

[0050] Compared with the iodine contrast agent delayed enhancement technology based on coronary CT angiography, the myocardial focal scar detection method provided by the present invention performs myocardial focal scar detection based on single-energy CCTA or dual-energy CCTA scanning images, does not require delayed scanning, and therefore does not require additional iodine contrast agent usage and radiation dose, thereby having the advantage of halving the iodine contrast agent usage and radiation dose.

[0051] Therefore, the myocardial focal scar detection method provided by the present invention solves the technical problems of the prior art, such as high difficulty in operation technology, long detection time, high cost, large amount of iodine contrast agent used and high radiation dose.

[0052] The risk prediction method provided by the present invention, based on the myocardial focal scar localization image obtained based on the myocardial focal scar detection method provided by the invention, further extracts the imaging features of the myocardial focal scar localization image, and based on the relationship between the imaging features of the myocardial focal scar localization image and major adverse cardiovascular events, screens out the major adverse cardiovascular event risk factors based on single-energy CCTA images or dual-energy CCTA images, and finally uses a preset risk prediction model to perform risk stratification prediction of major adverse cardiovascular events based on the major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results. By integrating imaging features with multidimensional medical data through a deep learning method, risk stratification prediction of major adverse cardiovascular events is achieved. On the basis of achieving the technical effect of the technical problem solved by the myocardial focal scar detection method provided by the present invention, it also solves the problem that the delayed enhancement technology of iodine contrast agent (LIE-CT) based on coronary CT angiography cannot be used for myocardial focal scar risk stratification.

[0053] The myocardial focal scar detection device and risk prediction equipment provided by the present invention are used to execute the myocardial focal scar detection method and risk prediction method provided by the present invention. The principles and technical effects achieved are the same as those of the myocardial focal scar detection method and risk prediction method provided by the present invention, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0055] FIG1 is a schematic flow chart of a method for detecting focal myocardial scars provided by the present invention;

[0056] FIG2 is a logic block diagram of a method for detecting focal myocardial scars provided by the present invention;

[0057] FIG3 is a schematic flow chart of a risk prediction method provided by the present invention;

[0058] FIG4 is a schematic structural diagram of a device for detecting focal myocardial scars provided by the present invention;

[0059] FIG5 is a schematic structural diagram of a risk prediction device provided by the present invention. DETAILED DESCRIPTION

[0060] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0061] For ease of understanding, please refer to FIG1 and FIG2 . The present invention provides an embodiment of a method for detecting focal myocardial scars, comprising:

[0062] Step 101: Perform voxel-based binary classification segmentation on the myocardial image scanned by single-energy CCTA or dual-energy CCTA based on a preset left ventricular wall myocardial automatic segmentation model to obtain a left ventricular wall myocardial image, wherein the voxel-based binary classification segmentation result includes left ventricular wall myocardial tissue and non-left ventricular wall myocardial tissue.

[0063] It should be noted that the location, volume, and degree of transmurality of focal myocardial scars can affect left ventricular structural remodeling and functional recovery. Scar volume is directly related to left ventricular function. Anterior wall scars are more likely to lead to heart failure and cardiac death. Even scar geometry and structure can affect the mechanical and electrical function of the heart. Therefore, it is necessary to detect focal myocardial scars. In this embodiment of the present invention, a preset left ventricular wall myocardial automatic segmentation model is used to perform voxel-based binary segmentation on myocardial images scanned using single-energy CCTA or dual-energy CCTA to obtain a left ventricular wall myocardial image. The pre-set left ventricular wall myocardium automatic segmentation model can be constructed by combining manual and machine annotation. First, a left ventricular wall myocardium annotation dataset based on single-energy CCTA scan images and dual-energy CCTA scan images is established. This dataset is then divided into training, validation, and test datasets. The deep learning network-based left ventricular wall myocardium automatic segmentation model is then trained, validated, and tested using voxel-based binary classification using the training, validation, and test datasets, respectively. This yields the pre-set left ventricular wall myocardium automatic segmentation model. Myocardial images obtained using single-energy CCTA or dual-energy CCTA scans are then input into the pre-set left ventricular wall myocardium automatic segmentation model. This model then performs voxel-based binary classification segmentation on the myocardial image, yielding a left ventricular wall myocardium image. The voxel-based binary classification segmentation results are Class 1 and Class 2, corresponding to left ventricular wall myocardial tissue and non-left ventricular wall myocardial tissue, respectively. It should be noted that because dual-energy CCTA can reconstruct multiple virtual single-energy images and base material images from a single scan, if the myocardial image is obtained from a dual-energy CCTA scan, the pre-set left ventricular wall myocardial segmentation model must perform voxel-wise binary segmentation on the multiple virtual single-energy images and base material images. This process must then be performed using advanced normalization tools (ANTs) to align and normalize the voxel-wise binary segmentation results before outputting the final left ventricular wall myocardial image. Accordingly, the same process must be performed when constructing the pre-set left ventricular wall myocardial segmentation model.

[0064] Step 102 : performing voxel-based three-category segmentation on the low-density lesions in the left ventricular wall myocardial image based on a preset myocardial focal scar localization model to obtain a myocardial focal scar localization image, wherein the voxel-based three-category segmentation result includes myocardial focal scars, myocardial radiation hardening artifacts, and normal myocardium.

[0065] It should be noted that myocardial radiation hardening artifacts are another common artifact during cardiac CT examinations and a significant issue affecting the accuracy of CT myocardial perfusion assessments. The mechanism of their formation is as follows: During industrial X-ray CT imaging, the X-ray energy spectrum is polychromatic, consisting of rays of varying energies. Therefore, when X-rays pass through tissue, their attenuation does not scale linearly with tissue thickness. When X-rays pass through dense tissue, lower-energy rays / photons are preferentially absorbed, while higher-energy photons pass through the tissue. This photoelectric effect narrows the radiation energy spectrum, gradually filtering the radiation to higher-energy photons. This process is known as radiation hardening. This radiation hardening phenomenon results in different radiation energies being received by different tissues, leading to distortion of tissue CT values ​​and the generation of radiation hardening artifacts. Radiation hardening artifacts primarily manifest as banded or cup-shaped low-density shadows between high-density tissues. Therefore, when detecting focal myocardial scars, it is important to distinguish between focal scars and myocardial radiation hardening artifacts. Therefore, in an embodiment of the present invention, low-density lesions in the left ventricular wall myocardial image are segmented into three voxel categories based on a preset myocardial focal scar localization model, where the three categories correspond to focal scars, myocardial radiation sclerosis artifacts, and normal myocardium, respectively.

[0066] To construct a pre-defined focal myocardial scar localization model, the authors first annotate myocardial focal scars (MFS) on left ventricular wall images using cardiac magnetic resonance imaging with delayed gadolinium contrast enhancement (CMR-LGE). They also annotate other focal areas of increased or decreased myocardial density as myocardial radiation hardening artifacts, thereby constructing a dataset of labeled myocardial focal scar artifacts and radiation hardening artifacts. The dataset is then divided into training, validation, and testing data sets. The training, validation, and testing data sets are then used to train, validate, and test the focal myocardial scar localization model using voxel-based three-category segmentation, respectively, to obtain the pre-defined focal myocardial scar localization model.

[0067] In one embodiment, a preset myocardial focal scar localization model is used to perform voxel-based three-category segmentation of low-density lesions in the left ventricular wall myocardial image. The resulting segmented myocardial focal scar image is first obtained. The myocardial focal scar volume is calculated based on the myocardial focal scar image to obtain quantitative information about the myocardial focal scar. Finally, the myocardial focal scar is located using myocardial segmentation information to obtain a precisely localized myocardial focal scar localization image. Methods for left ventricular myocardial segmentation are currently available and will not be further elaborated here.

[0068] Compared with subendocardial puncture biopsy, the myocardial focal scar detection method provided by the present invention has the advantages of being non-invasive, less difficult to operate, and able to reflect the overall changes in the myocardium.

[0069] Compared with the cardiac magnetic resonance gadolinium contrast agent delayed enhancement imaging technology, the myocardial focal scar detection method provided by the present invention has the advantages of less operating difficulty, shorter detection time and lower cost.

[0070] Compared with the iodine contrast agent delayed enhancement technology based on coronary CT angiography, the myocardial focal scar detection method provided by the present invention performs myocardial focal scar detection based on single-energy CCTA or dual-energy CCTA scanning images, does not require delayed scanning, and therefore does not require additional iodine contrast agent usage and radiation dose, thereby having the advantage of halving the iodine contrast agent usage and radiation dose.

[0071] Therefore, the myocardial focal scar detection method provided by the present invention solves the technical problems of the prior art, such as high difficulty in operation technology, long detection time, high cost, large amount of iodine contrast agent used and high radiation dose.

[0072] For ease of understanding, please refer to FIG3 . The present invention provides an embodiment of a risk prediction method. The risk prediction method includes the myocardial focal scar detection method provided by the present invention, and further includes:

[0073] Step 201: Extract imaging features of the myocardial focal scar localization image.

[0074] It should be noted that after obtaining a focal myocardial scar localization image using the myocardial scar detection method provided by the present invention, a feature extraction method is used to extract imaging features from the focal myocardial scar localization image. Specifically, multiple imaging features are extracted from the focal myocardial scar localization image, including imaging features assessed by human visual evaluation and radiomics features identified through computer computation. These multiple imaging features are then visually evaluated, normalized, dimensionally reduced, and de-redundanted to obtain imaging features of the focal myocardial scar localization image with high correlation, high discrimination, and high reliability.

[0075] Step 202: Based on the relationship between the imaging features of the myocardial focal scar localization image and major adverse cardiovascular events, the risk factors for major adverse cardiovascular events based on the single-energy CCTA image or the dual-energy CCTA image are screened.

[0076] It should be noted that major adverse cardiovascular events (MACE) include recurrent angina, acute myocardial infarction, severe arrhythmias, heart failure, and coronary heart disease death. Risk factors for major adverse cardiovascular events include hypertension, atrial fibrillation, hyperlipidemia, and smoking. The imaging features of focal myocardial scar localization images correspond to major adverse cardiovascular events, and this relationship can be established in advance through massive data analysis. Therefore, after extracting the imaging features of focal myocardial scar localization images, the major adverse cardiovascular events corresponding to the focal myocardial scar localization images can be determined based on the relationship between the imaging features and major adverse cardiovascular events, thereby screening for risk factors for major adverse cardiovascular events based on single-energy CCTA images or dual-energy CCTA images.

[0077] Step 203: Using a preset risk prediction model, perform risk stratification prediction on major adverse cardiovascular events based on major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results.

[0078] It should be noted that multidimensional medical data includes characteristic data from multiple dimensions, such as clinical manifestations, laboratory tests, transthoracic echocardiography, and electrocardiograms. By combining major adverse cardiovascular event risk factors with multidimensional medical data and corresponding risk stratification labels, a preset risk prediction model based on a deep learning network can be constructed. The major adverse cardiovascular event risk factors obtained in step 202 and the corresponding multidimensional medical data are input into the preset risk prediction model to perform risk stratification prediction of major adverse cardiovascular events, thereby obtaining corresponding risk stratification prediction results.

[0079] The risk prediction method provided by the present invention, based on the myocardial focal scar localization image obtained based on the myocardial focal scar detection method provided by the invention, further extracts the imaging features of the myocardial focal scar localization image, and based on the relationship between the imaging features of the myocardial focal scar localization image and major adverse cardiovascular events, screens out the major adverse cardiovascular event risk factors based on single-energy CCTA images or dual-energy CCTA images, and finally uses a preset risk prediction model to perform risk stratification prediction of major adverse cardiovascular events based on the major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results. By integrating imaging features with multidimensional medical data through a deep learning method, risk stratification prediction of major adverse cardiovascular events is achieved. On the basis of achieving the technical effect of the technical problem solved by the myocardial focal scar detection method provided by the present invention, it also solves the problem that the delayed enhancement technology of iodine contrast agent (LIE-CT) based on coronary CT angiography cannot be used for myocardial focal scar risk stratification.

[0080] For ease of understanding, please refer to FIG4 . The present invention provides an embodiment of a device for detecting focal myocardial scars, comprising:

[0081] A left ventricular wall myocardial image acquisition module is used to perform voxel-based binary classification segmentation on the myocardial image scanned based on single-energy CCTA or dual-energy CCTA based on a preset left ventricular wall myocardial automatic segmentation model to obtain a left ventricular wall myocardial image, wherein the voxel-based binary classification segmentation result includes left ventricular wall myocardial tissue and non-left ventricular wall myocardial tissue;

[0082] The focal scar localization module is used to perform voxel three-category segmentation on the low-density lesions in the left ventricular wall myocardial image based on a preset myocardial focal scar localization model to obtain a myocardial focal scar localization image, wherein the voxel three-category segmentation result includes myocardial focal scars, myocardial radiation hardening artifacts and normal myocardium.

[0083] Also included is a first model building module;

[0084] The first model building module is used to:

[0085] Construct a left ventricular wall myocardial annotation dataset based on single-energy CCTA scan images and dual-energy CCTA scan images;

[0086] The left ventricular wall myocardium annotation dataset is used to train, verify and test the left ventricular wall myocardium automatic segmentation model based on voxel binary classification segmentation, and a preset left ventricular wall myocardium automatic segmentation model is obtained.

[0087] Also included is a second model building module;

[0088] The second model building module is used to:

[0089] Based on left ventricular wall myocardial images annotated with focal myocardial scars and myocardial radiation hardening artifacts, a dataset of annotations of focal myocardial scar artifacts and radiation hardening artifacts was constructed.

[0090] The myocardial focal scar artifact and radiation hardening artifact annotation datasets were used to train, validate, and test the focal scar localization model based on voxel-based three-category segmentation, and a preset myocardial focal scar localization model was obtained.

[0091] The Focal Scar Localization Module is specifically used for:

[0092] Based on the preset myocardial focal scar localization model, the low-density lesions in the left ventricular wall myocardial image are segmented by three-dimensional voxel classification to obtain the myocardial focal scar image;

[0093] The volume of myocardial focal scar is calculated from the myocardial focal scar image to obtain quantitative information of myocardial focal scar;

[0094] Based on the quantitative information of myocardial focal scars and the myocardial segmentation information, the myocardial focal scars are located to obtain a myocardial focal scar localization image.

[0095] For ease of understanding, please refer to FIG5 . The present invention provides a risk prediction device, including the myocardial focal scar detection device provided in the present invention, and further comprising:

[0096] A feature extraction module is used to extract imaging features of myocardial focal scar localization images;

[0097] The risk factor screening module is used to screen the risk factors for major adverse cardiovascular events based on single-energy CCTA images or dual-energy CCTA images based on the relationship between the imaging characteristics of myocardial focal scar localization images and major adverse cardiovascular events;

[0098] The risk stratification prediction module is used to use a preset risk prediction model to perform risk stratification prediction of major adverse cardiovascular events based on major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results.

[0099] The feature extraction module is specifically used for:

[0100] Based on the myocardial focal scar localization image, multiple imaging features are extracted, and visual evaluation, normalization, dimensionality reduction and redundancy removal are performed on the multiple imaging features to obtain the imaging features of the myocardial focal scar localization image. Among them, the multiple imaging features include imaging features visually evaluated by the human eye and imaging omics features based on computer calculation and recognition.

[0101] The myocardial focal scar detection device and risk prediction equipment provided by the present invention are used to execute the myocardial focal scar detection method and risk prediction method provided by the present invention. The principles and technical effects achieved are the same as those of the myocardial focal scar detection method and risk prediction method provided by the present invention, and will not be repeated here.

[0102] The terms "first," "second," "third," "fourth," and the like (if any) in the specification and drawings of the present invention are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of the present application described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions, e.g., a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0103] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting focal myocardial scars, characterized in that: include: Based on a preset left ventricular wall myocardial automatic segmentation model, a voxel binary classification segmentation is performed on the myocardial image scanned based on single-energy CCTA or dual-energy CCTA to obtain a left ventricular wall myocardial image, wherein the voxel binary classification segmentation result includes left ventricular wall myocardial tissue and non-left ventricular wall myocardial tissue; Based on the preset myocardial focal scar localization model, the low-density lesions in the left ventricular wall myocardial image are segmented by voxel three-category classification to obtain a myocardial focal scar localization image, wherein the voxel three-category segmentation result includes myocardial focal scars, myocardial radiation hardening artifacts and normal myocardium.

2. The method for detecting focal myocardial scars according to claim 1, characterized in that: Based on the preset left ventricular wall myocardial automatic segmentation model, the myocardial image scanned by single-energy CCTA or dual-energy CCTA is segmented by voxel binary classification to obtain the left ventricular wall myocardial image. Previously, it also included: Construct a left ventricular wall myocardial annotation dataset based on single-energy CCTA scan images and dual-energy CCTA scan images; The left ventricular wall myocardium annotation dataset is used to train, verify and test the left ventricular wall myocardium automatic segmentation model based on voxel binary classification segmentation, and a preset left ventricular wall myocardium automatic segmentation model is obtained.

3. The method for detecting focal myocardial scars according to claim 1, characterized in that: Based on the preset myocardial focal scar localization model, the low-density lesions in the left ventricular wall myocardial image are segmented by three-classification voxels to obtain the myocardial focal scar localization image, which also includes: Based on the left ventricular wall myocardial images annotated with myocardial focal scars and myocardial radiation hardening artifacts, a dataset of annotations of myocardial focal scar artifacts and radiation hardening artifacts was constructed. The focal scar localization model was trained, verified and tested based on voxel-based three-category segmentation using the myocardial focal scar artifact and radiation hardening artifact annotation datasets to obtain a preset myocardial focal scar localization model.

4. The method for detecting focal myocardial scars according to claim 1, characterized in that: Based on the preset myocardial focal scar localization model, the low-density lesions in the left ventricular wall myocardial image are segmented by three-dimensional voxel classification to obtain a myocardial focal scar localization image, including: Based on the preset myocardial focal scar localization model, low-density lesions in left ventricular wall myocardial images were identified. Perform voxel three-classification segmentation to obtain the image of myocardial focal scar; Calculate the volume of myocardial focal scar from the image of myocardial focal scar to obtain quantitative information of myocardial focal scar; Based on the quantitative information of the myocardial focal scar and the myocardial segmentation information, the myocardial focal scar is located to obtain a myocardial focal scar localization image.

5. A risk prediction method, characterized in that: The method for detecting focal myocardial scar according to any one of claims 1 to 4, further comprising: Extract imaging features of focal myocardial scar localization images; Based on the relationship between the imaging features of focal myocardial scar localization images and major adverse cardiovascular events, the risk factors for major adverse cardiovascular events based on single-energy CCTA images or dual-energy CCTA images were screened out; A preset risk prediction model was used to perform risk stratification prediction of major adverse cardiovascular events based on major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results.

6. The risk prediction method according to claim 5, characterized in that: Extract imaging features of myocardial focal scar localization images, including: Based on the myocardial focal scar localization image, a variety of imaging features are extracted, and visual evaluation, normalization, dimensionality reduction and redundancy removal are performed on the multiple imaging features to obtain the imaging features of the myocardial focal scar localization image. Among them, the multiple imaging features include imaging features visually evaluated by the human eye and imaging omics features based on computer calculation and recognition.

7. A myocardial focal scar detection device, characterized in that: include: A left ventricular wall myocardial image acquisition module is used to perform voxel binary classification segmentation on the myocardial image scanned based on single-energy CCTA or dual-energy CCTA based on a preset left ventricular wall myocardial automatic segmentation model to obtain a left ventricular wall myocardial image, wherein the voxel binary classification segmentation result includes left ventricular wall myocardial tissue and non-left ventricular wall myocardial tissue; The focal scar localization module is used to perform voxel three-category segmentation on the low-density lesions in the left ventricular wall myocardial image based on a preset myocardial focal scar localization model to obtain a myocardial focal scar localization image, wherein the voxel three-category segmentation result includes myocardial focal scars, myocardial radiation hardening artifacts and normal myocardium.

8. The myocardial focal scar detection device according to claim 7, characterized in that: Also included is a first model building module; The first model building module is used to: Construct a left ventricular wall myocardial annotation dataset based on single-energy CCTA scan images and dual-energy CCTA scan images; The left ventricular wall myocardium annotation dataset is used to train, verify and test the left ventricular wall myocardium automatic segmentation model based on voxel binary classification segmentation, and a preset left ventricular wall myocardium automatic segmentation model is obtained.

9. The myocardial focal scar detection device according to claim 7, characterized in that: Also included is a second model building module; The second model building module is used to: Based on the left ventricular wall myocardial images annotated with myocardial focal scars and myocardial radiation hardening artifacts, a dataset of annotations of myocardial focal scar artifacts and radiation hardening artifacts was constructed. The focal scar localization model was trained, verified and tested based on voxel-based three-category segmentation using the myocardial focal scar artifact and radiation hardening artifact annotation datasets to obtain a preset myocardial focal scar localization model.

10. A risk prediction device, characterized in that: The device for detecting focal myocardial scar according to any one of claims 7 to 9, further comprising: A feature extraction module is used to extract imaging features of myocardial focal scar localization images; The risk factor screening module is used to screen out the risk factors of major adverse cardiovascular events based on single-energy CCTA images or dual-energy CCTA images based on the relationship between the imaging characteristics of myocardial focal scar localization images and major adverse cardiovascular events; The risk stratification prediction module is used to use a preset risk prediction model to perform risk stratification prediction of major adverse cardiovascular events based on major adverse cardiovascular event risk factors and multidimensional medical data to obtain risk stratification prediction results.

Citation Information

Patent Citations

  • Method for quantitatively evaluating myocardial infarction on basis of nuclide image and CT (computed tomography) coronary angiography fusion

    CN109498046A

  • Image processing method and device, computer equipment, storage medium and mapping system

    CN113870339A

  • Segmentation method and device for inner myocardial membrane and / or outer myocardial membrane of left ventricle of heart

    CN114419032A

  • Prediction device for main heart adverse events based on magnetic resonance imaging

    CN115965621A

  • Myocardial focal scar detection method, risk prediction method and related device

    CN116245878A