Method and device for determining ventricular septal thickness, computer device and storage medium

Abstract

The application relates to a method and device for determining the thickness of the interventricular septum, a computer device and a storage medium. The method comprises the following steps: acquiring a heart scan image, performing image segmentation on the heart scan image by using an image segmentation model, and determining a segmentation mask of a target detection region in the heart scan image; converting the segmentation mask of the target detection region by using a moving cube algorithm, and determining target three-dimensional point cloud data corresponding to each segmentation mask; determining key point position information of a heart key point according to the target three-dimensional point cloud data and a key point detection model; determining the position information of a three-dimensional section of a left ventricular interventricular septum region according to the key point position information; the three-dimensional section comprises a mitral valve horizontal short-axis section, a left ventricular long-axis section and a four-chamber heart section; and determining the thickness of the interventricular septum of the left ventricular interventricular septum region according to the position information of the three-dimensional section. The above scheme improves the accuracy of the determined thickness of the interventricular septum.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a method, apparatus, computer device, and storage medium for determining interventricular septum thickness. Background Technology

[0002] Hypertrophic cardiomyopathy is characterized by thickening of the left ventricular septum. This thickening can lead to left ventricular outflow tract obstruction and hemodynamic abnormalities, resulting in symptoms such as chest pain, syncope, and heart failure. For patients with severe outflow tract obstruction, septal myocardial resection is the gold standard in clinical treatment. However, the successful implementation of septal myocardial resection depends on the accurate assessment of key indicators such as septal thickness. Traditionally, two-dimensional echocardiography is the primary method for measuring septal thickness. However, this method provides only limited sectional information and cannot fully reflect the three-dimensional morphology of the septum, easily leading to errors in surgical area assessment and significant measurement errors. Furthermore, physicians must manually sift through numerous two-dimensional tomographic images, which is time-consuming and prone to missing crucial information, further reducing measurement accuracy. Therefore, there is an urgent need for precise, efficient, and real-time septal thickness measurement technology to improve the safety and effectiveness of myocardial resection. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, apparatus, computer equipment, and storage medium for determining interventricular septum thickness that can improve the accuracy and efficiency of interventricular septum thickness measurement, in order to address the aforementioned technical problems.

[0004] In a first aspect, this application provides a method for determining interventricular septal thickness, the method comprising:

[0005] Acquire cardiac scan images and segment the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0006] The moving cube algorithm is used to transform the segmentation mask of the target detection region, and the target 3D point cloud data corresponding to each segmentation mask is determined respectively;

[0007] Based on the target 3D point cloud data and key point detection model, the key point location information of the heart key points is determined; the heart key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point;

[0008] Based on the key point location information, the section position information of the three-dimensional section of the left ventricular interventricular septum region is determined; the three-dimensional section includes the mitral valve level short axis section, the left ventricular long axis section, and the four-chamber section;

[0009] Based on the section position information of the three-dimensional section, the interventricular septum thickness in the left ventricular septum region is determined.

[0010] In one embodiment, the target aortic valve key point is the junction of the left coronary lobes and the non-coronary lobes of the target aortic valve; the target mitral valve key point includes the intersection of two target mitral valve leaflets and the midpoint of the posterior mitral valve leaflet; the target right ventricular key point is the midpoint of the tricuspid valve; and the target left ventricular key point is the apex of the left ventricle.

[0011] In one embodiment, the key point location information of the cardiac key points is determined based on the target 3D point cloud data and the key point detection model, including:

[0012] Based on the target 3D point cloud data and key point detection model, determine the key point location information of the target aortic valve key point, the target mitral valve key point, and the target right ventricle key point;

[0013] Uniform sampling is performed on the target three-dimensional point cloud data corresponding to the target left ventricular region to determine candidate sampling points, and the spatial plane formed by each candidate sampling point, the left ventricular junction point, and the mitral valve center point is determined.

[0014] Determine the area of ​​the cross section after each spatial plane is tangent to the original left ventricular point cloud in the target three-dimensional point cloud data corresponding to the target left ventricular region;

[0015] Based on the cross-sectional area, the sampling point location information of the candidate sampling points, the boundary point location information of the left non-intersection point, and the center point location information of the mitral valve center point, the target left ventricular key point is determined from the candidate sampling points, and the key point location information of the target left ventricular key point is determined.

[0016] In one embodiment, based on the cross-sectional area, the sampling point location information of the candidate sampling points, the boundary point location information of the left ventricular non-boundary point, and the center point location information of the mitral valve center point, a target left ventricular key point is determined from the candidate sampling points, and the key point location information of the target left ventricular key point is determined, including:

[0017] The largest area is determined from the cross-sectional areas, and the candidate sampling point corresponding to the largest area is determined as the sampling point to be screened;

[0018] Based on the sampling point location information of the sampling point to be screened, the boundary point location information of the left no-boundary point and the center point location information of the mitral valve center point, the sampling point to be screened located below the left no-boundary point and the mitral valve center point is determined, and the sampling point to be screened located below the left no-boundary point and the mitral valve center point is taken as the target sampling point.

[0019] Based on the sampling point location information of the target sampling point and the boundary point location information of the left boundary point, a first distance between the target sampling point and the left boundary point is determined, and based on the sampling point location information of the target sampling point and the center point location information of the mitral valve center point, a second distance between the target sampling point and the mitral valve center point is determined.

[0020] The sum of the first distance and the second distance is taken as the spatial distance between the target sampling point and the left boundary point and the mitral valve center point;

[0021] The maximum distance is determined from the spatial distance, and the sampling point location information of the target sampling point corresponding to the maximum distance is determined as the key point location information of the target left ventricular key point.

[0022] In one embodiment, the section position information of the three-dimensional section of the left ventricular septum region is determined based on the key point location information, including:

[0023] The horizontal short axis section of the mitral valve is determined based on the key points of the target mitral valve using a random sampling consensus algorithm.

[0024] Determine the long axis section of the left ventricle based on the target aortic valve key point, the target mitral valve center point, and the target left ventricle key point;

[0025] The four-chamber view is determined based on the center points of the target mitral valve, the target left ventricular key point, and the target right ventricular key point.

[0026] In one embodiment, the above method for determining interventricular septum thickness further includes:

[0027] Obtain sample cardiac images and sample annotation information of the sample cardiac images; the sample annotation information includes annotation information of the sample aortic region, sample aortic valve region, sample mitral valve region, sample left ventricular region, and sample right ventricular region of the sample cardiac images;

[0028] The deep learning network is trained using the heart-related information of the samples and the sample annotation information to determine the image segmentation model.

[0029] In one embodiment, the above method for determining interventricular septum thickness further includes:

[0030] Acquire the sample three-dimensional point cloud data corresponding to the sample cardiac image and the annotation information of the sample three-dimensional point cloud data; the annotation information of the sample three-dimensional point cloud data includes the cardiac structure label corresponding to the sample three-dimensional point cloud data, and the key point coordinates of the sample key points in the sample three-dimensional point cloud data; the sample key points include the sample aortic valve key points, the sample mitral valve key points and the sample right ventricle key points.

[0031] Based on the sample 3D point cloud data and the annotation information of the sample 3D point cloud data, the point cloud network to be trained is trained to determine the key point detection model; the key point detection model includes an aortic valve detection model, a mitral valve detection model and a right ventricle detection model.

[0032] Secondly, this application also provides a device for determining interventricular septal thickness, the device comprising:

[0033] The image segmentation module is used to acquire cardiac scan images and perform image segmentation on the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0034] The point cloud data acquisition module is used to convert the segmentation mask of the target detection area using the moving cube algorithm, and to determine the target 3D point cloud data corresponding to each segmentation mask;

[0035] The key point determination module is used to determine the key point location information of cardiac key points based on the target 3D point cloud data and the key point detection model; the cardiac key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point;

[0036] The three-dimensional section determination module is used to determine the section position information of the three-dimensional section of the left ventricular interventricular septum region based on the key point position information; the three-dimensional section includes the mitral valve level short axis section, the left ventricular long axis section, and the four-chamber section;

[0037] The interventricular septum thickness determination module is used to determine the interventricular septum thickness in the left ventricular septum region based on the section position information of the three-dimensional section.

[0038] Thirdly, this application also provides a computer device, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0039] Acquire cardiac scan images and segment the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0040] The moving cube algorithm is used to transform the segmentation mask of the target detection region, and the target 3D point cloud data corresponding to each segmentation mask is determined respectively;

[0041] Based on the target 3D point cloud data and key point detection model, the key point location information of the heart key points is determined; the heart key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point;

[0042] Based on the key point location information, the section position information of the three-dimensional section of the left ventricular interventricular septum region is determined; the three-dimensional section includes the mitral valve level short axis section, the left ventricular long axis section, and the four-chamber section;

[0043] Based on the section position information of the three-dimensional section, the interventricular septum thickness in the left ventricular septum region is determined.

[0044] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0045] Acquire cardiac scan images and segment the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0046] The moving cube algorithm is used to transform the segmentation mask of the target detection region, and the target 3D point cloud data corresponding to each segmentation mask is determined respectively;

[0047] Based on the target 3D point cloud data and key point detection model, the key point location information of the heart key points is determined; the heart key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point;

[0048] Based on the key point location information, the section position information of the three-dimensional section of the left ventricular interventricular septum region is determined; the three-dimensional section includes the mitral valve level short axis section, the left ventricular long axis section, and the four-chamber section;

[0049] Based on the section position information of the three-dimensional section, the interventricular septum thickness in the left ventricular septum region is determined.

[0050] The aforementioned method, apparatus, computer equipment, and storage medium for determining interventricular septal thickness acquire cardiac scan images and segment these images using an image segmentation model to determine segmentation masks for target detection regions within the cardiac scan images. The target detection regions include the target aortic valve region, target mitral valve region, target left ventricle region, and target right ventricle region. A moving cube algorithm is used to transform the segmentation masks of the target detection regions, determining the target 3D point cloud data corresponding to each segmentation mask. Based on the target 3D point cloud data and a key point detection model, the key point location information of cardiac key points is determined. These cardiac key points include target aortic valve key points, target mitral valve key points, target right ventricle key points, and target left ventricle key points. Based on the key point location information, the cross-sectional location information of the 3D cross-section of the left ventricular interventricular septal region is determined. The 3D cross-section includes a short-axis cross-section at the mitral valve level, a long-axis cross-section of the left ventricle, and a four-chamber cross-section. Based on the cross-sectional location information of the 3D cross-section, the interventricular septal thickness of the left ventricular interventricular septal region is determined. This solution addresses the shortcomings of traditional two-dimensional echocardiography-based methods for measuring interventricular septum thickness. These methods fail to fully reflect the three-dimensional morphology of the septum and require physicians to manually sift through numerous two-dimensional tomographic images, leading to inaccurate measurements. The proposed method first acquires cardiac scan images. An image segmentation model is used to obtain segmentation masks for the aortic valve, mitral valve, left ventricular, and right ventricular regions. A moving cubes algorithm then converts these masks into corresponding three-dimensional point cloud data. A key point detection model determines the locations of key cardiac points. Based on these key point locations, three types of three-dimensional sections are determined: a short-axis section at the mitral valve level, a long-axis section at the left ventricular level, and a four-chamber section. The thickness of the left ventricular septum is measured based on the position of these three-dimensional sections. This approach ensures that the determined three-dimensional sections conform to clinical anatomical standards and avoids the errors caused by manual positioning of the sections. Determining the septal thickness through the position of the three-dimensional sections allows for efficient and accurate acquisition of multi-directional septal thickness data, providing precise data for preoperative planning of procedures such as myocardial resection and improving the accuracy of the determined septal thickness. Attached Figure Description

[0051] Figure 1 This is an application environment diagram of the interventricular septum thickness determination method in one embodiment;

[0052] Figure 2 This is a flowchart illustrating a method for determining interventricular septum thickness in one embodiment;

[0053] Figure 3 This is a flowchart illustrating a method for determining key point location information in one embodiment;

[0054] Figure 4 This is a structural block diagram of the interventricular septum thickness determination device in one embodiment;

[0055] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0057] The method for determining interventricular septal thickness provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on another network server. Server 104 acquires cardiac scan images and performs image segmentation on the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target regions are the aortic valve region, the mitral valve region, the left ventricle region, and the right ventricle region; the segmentation mask of the target detection region is converted using a moving cube algorithm to determine the target three-dimensional point cloud data corresponding to each segmentation mask; based on the target three-dimensional point cloud data and the key point detection model, the key point location information of cardiac key points is determined; the cardiac key points include the target aortic valve key points, the target mitral valve key points, the target right ventricle key points, and the target left ventricle key points; based on the key point location information, the section position information of the three-dimensional section of the left ventricular interventricular septum region is determined; the three-dimensional section includes the mitral valve horizontal short axis section, the left ventricular long axis section, and the four-chamber section; based on the section position information of the three-dimensional section, the interventricular septum thickness of the left ventricular interventricular septum region is determined. The interventricular septum thickness of the left ventricular interventricular septum region is sent to terminal 102 through a communication network. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle systems. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0058] In one embodiment, such as Figure 2 As shown, a method for determining interventricular septal thickness is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] S210. Acquire cardiac scan images and perform image segmentation on the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images.

[0060] Target aortic valve region, target mitral valve region, target left ventricle region, and target right ventricle region.

[0061] The cardiac scan images are of the left ventricular interventricular septum region. This region mainly includes the interventricular septum myocardium, the left ventricular surface of the interventricular septum, and the right ventricular surface. The image segmentation model is a pre-trained nnUnet deep learning network. The nnUnet deep learning network is an adaptive medical image segmentation deep learning framework based on the U-Net architecture. It can automatically optimize the network structure, training parameters, and processing flow according to the features of the input medical image data, achieving high-precision and high-generalization automatic segmentation of anatomical structures or lesions in multimodal images.

[0062] Specifically, CTA (coronary CT angiography) technology is used to acquire high-resolution three-dimensional image data of the left ventricular septum region, i.e., cardiac scan images, or cardiac CTA images. Image segmentation models are then used to segment the cardiac scan images to extract a segmentation mask for the target detection region.

[0063] For example, training methods for image segmentation models include:

[0064] Obtain sample cardiac images and sample annotation information of the sample cardiac images; the sample annotation information includes the annotation information of the sample aortic region, the sample aortic valve region, the sample mitral valve region, the sample left ventricular region, and the sample right ventricular region of the sample cardiac images; use the sample cardiac images and the sample annotation information to train a deep learning network to determine the image segmentation model.

[0065] The deep learning network in this context is the nnUnet deep learning network. Sample cardiac images refer to sample cardiac CTA images, such as cardiac CTA images of the left ventricular interventricular septum region acquired during historical processing.

[0066] Specifically, sample cardiac images are collected as training data for the model. Historically acquired cardiac CTA images containing the left ventricular septum region can be selected as sample images. The corresponding annotation information for these images is obtained, covering the aortic valve region, mitral valve region, left ventricular region, and right ventricular region. Using the sample cardiac images and their corresponding annotation information, the nnUnet deep learning network is trained in a supervised manner. The nnUnet deep learning network learns the mapping relationship between image features and target detection regions, ultimately obtaining an image segmentation model that can be used for practical segmentation.

[0067] The above scheme can construct an image segmentation model that is accurate and has strong generalization for segmenting multiple target anatomical structures of the heart, providing a reliable structural segmentation basis for subsequent ventricular septum-related measurements and surgical evaluation.

[0068] S220. The moving cube algorithm is used to convert the segmentation mask of the target detection area, and the target 3D point cloud data corresponding to each segmentation mask is determined.

[0069] The Moving Cube algorithm, also known as the MarchingCube algorithm, is a 3D isosurface extraction algorithm. Its core function is to convert 3D voxel data from CT (Computed Tomography Angiography) / MRI (Magnetic Resonance Imaging) into a 3D surface model represented by a triangular mesh. 3D point cloud data, a dataset composed of a large number of 3D spatial coordinate points, can accurately reconstruct the 3D spatial morphology of various structures in the heart, providing suitable input data for subsequent detection.

[0070] Specifically, the moving cube algorithm is used to reconstruct the three-dimensional surface of the target detection area, converting the two-dimensional mask data, which originally existed in the form of voxels, into a discrete set of three-dimensional spatial points, thereby obtaining the three-dimensional point cloud data corresponding to each segmentation mask.

[0071] S230. Based on the target's 3D point cloud data and key point detection model, determine the key point location information of the heart's key points.

[0072] The key points for the heart include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point.

[0073] The keypoint detection model is a pre-trained PointNet++ network model. The PointNet++ network model is a deep neural network model used to process 3D point cloud data. It should be noted that the keypoint detection models include aortic valve detection models, mitral valve detection models, and right ventricle detection models.

[0074] Specifically, based on the 3D point cloud data of each target detection region obtained through the moving cubes algorithm, the 3D point cloud data of each target detection region is input into a trained keypoint detection model. The keypoint detection model identifies and locates the features of the 3D point cloud data, thereby outputting the keypoint location information of the target aortic valve keypoint, the target mitral valve keypoint, and the target right ventricle keypoint. Specifically, the target 3D point cloud data corresponding to the target aortic valve region is input into the aortic valve detection model to determine the keypoint location information of the target aortic valve keypoint; the target 3D point cloud data corresponding to the target mitral valve region is input into the mitral valve detection model to determine the keypoint location information of the target mitral valve keypoint; and the target 3D point cloud data corresponding to the target right ventricle region is input into the right ventricle detection model to determine the keypoint location information of the target right ventricle keypoint. The left ventricular region point cloud data corresponding to the target left ventricular region is determined from the target 3D point cloud data, and the keypoint location information of the target left ventricular keypoint is determined based on the left ventricular region point cloud data.

[0075] For example, the target aortic valve key point is the junction of the left coronary lobes and the non-coronary lobes of the target aortic valve; the target mitral valve key point includes the junction of the two target mitral valve leaflets and the midpoint of the posterior leaflet of the target mitral valve; the target right ventricular key point is the midpoint of the tricuspid valve; and the target left ventricular key point is the apex of the left ventricle.

[0076] For example, methods for determining keypoint detection models include:

[0077] The process involves acquiring the sample 3D point cloud data and its annotation information corresponding to the sample cardiac images. The annotation information includes the cardiac structure labels corresponding to the sample 3D point cloud data, as well as the key point coordinates of key points within the sample 3D point cloud data. These key points include aortic valve key points, mitral valve key points, and right ventricular key points. Based on the sample 3D point cloud data and its annotation information, the point cloud network to be trained is used to determine the key point detection model. The key point detection model includes an aortic valve detection model, a mitral valve detection model, and a right ventricular detection model.

[0078] The cardiac structural labels include aortic valve region labels, mitral valve region labels, left ventricular region labels, and right ventricular region labels.

[0079] Understandably, the point cloud network to be trained is determined based on the sample 3D point cloud data and the sample 3D point cloud data corresponding to the aortic valve region, thus determining the aortic valve detection model; the point cloud network to be trained is determined based on the sample 3D point cloud data and the sample 3D point cloud data corresponding to the mitral valve region, thus determining the mitral valve detection model; and the point cloud network to be trained is determined based on the sample 3D point cloud data and the sample 3D point cloud data corresponding to the right ventricular region, thus determining the right ventricular detection model.

[0080] Specifically, the process first acquires the 3D point cloud data and annotation information corresponding to the sample cardiac images. The annotation information includes the cardiac structure labels corresponding to the 3D point cloud data, as well as the coordinates of key points on the aortic valve, mitral valve, and right ventricle. Then, using the PointNet++ network model as the base model, and the sample 3D point cloud data and its annotation information as training data, three keypoint detection models are trained based on the cross-entropy loss function: an aortic valve detection model with one keypoint, a mitral valve detection model with three keypoints, and a right ventricle detection model with one keypoint. During the model prediction stage, a targeted threshold of 0.1 is applied to the prediction results, retaining only positions with a confidence level greater than 0.1 as the final detected keypoints, thereby controlling the number of keypoint location information output by the model.

[0081] The above scheme obtains the corresponding detection model by training a point cloud network based on sample 3D point cloud data containing cardiac structure labels and key point coordinates. It can achieve automatic and accurate positioning of key anatomical points of the heart, providing a reliable coordinate basis for subsequent interventricular septal thickness measurement.

[0082] For example, such as Figure 3 As shown, based on the target 3D point cloud data and the key point detection model, the key point location information of the heart key points is determined, including:

[0083] S2301. Based on the target 3D point cloud data and key point detection model, determine the key point location information of the target aortic valve key point, the target mitral valve key point, and the target right ventricle key point.

[0084] Specifically, the target 3D point cloud data corresponding to the target aortic valve region is input into the aortic valve detection model to determine the 3D coordinates of the target aortic valve key points as the key point location information of the target aortic valve key points; the target 3D point cloud data corresponding to the target mitral valve region is input into the mitral valve detection model to determine the 3D coordinates of the target mitral valve key points as the key point location information of the target mitral valve key points; and the target 3D point cloud data corresponding to the target right ventricle region is input into the right ventricle detection model to determine the 3D coordinates of the target right ventricle key points as the key point location information of the target right ventricle key points.

[0085] S2302. Uniformly sample the target three-dimensional point cloud data corresponding to the target left ventricular region, determine candidate sampling points, and determine the spatial plane formed by the combination of each candidate sampling point, the left ventricular junction point, and the mitral valve center point.

[0086] The left non-coronary junction refers to the junction between the left coronary lobes and the non-coronary lobes of the aortic valve.

[0087] Specifically, the three-dimensional point cloud data corresponding to the left ventricular region is uniformly sampled to obtain multiple evenly distributed candidate sampling points. Each candidate sampling point is then combined with the detected left ventricular junction point and mitral valve center point to construct a spatial plane composed of these three points.

[0088] S2303. Determine the area of ​​the cross section after each spatial plane is tangent to the original left ventricular point cloud in the target three-dimensional point cloud data corresponding to the target left ventricular region.

[0089] Among them, the original left ventricular point cloud refers to the target three-dimensional point cloud data corresponding to the target left ventricular region.

[0090] S2304. Based on the cross-sectional area, the sampling point location information of the candidate sampling points, the boundary point location information of the left ventricular without a boundary point, and the center point location information of the mitral valve center point, determine the target left ventricular key point from the candidate sampling points, and determine the key point location information of the target left ventricular key point.

[0091] For example, a method for determining the key point location information of the target left ventricular key point includes:

[0092] The maximum area is determined from the cross-sectional area, and the candidate sampling point corresponding to the maximum area is identified as the sampling point to be screened. Based on the sampling point location information of the sampling point to be screened, the location information of the left ventricular junction, and the location information of the center point of the mitral valve, the sampling points located below the left ventricular junction and the center point of the mitral valve are identified as the target sampling points. Based on the sampling point location information of the target sampling point and the location information of the left ventricular junction, a first distance between the target sampling point and the left ventricular junction is determined, and based on the sampling point location information of the target sampling point and the location information of the center point of the mitral valve, a second distance between the target sampling point and the center point of the mitral valve is determined. The sum of the first distance and the second distance is taken as the spatial distance between the target sampling point and the left ventricular junction and the center point of the mitral valve. The maximum distance is determined from the spatial distances, and the sampling point location information of the target sampling point corresponding to the maximum distance is identified as the key point location information of the target left ventricular key point.

[0093] It should be noted that the Z-axis, which runs along the long axis of the human heart, can be defined as the image coordinate axis. The direction pointing towards the apex of the heart is the positive direction of the Z-axis, and the negative direction of the Z-axis points to the area below the left ventricular junction and the center of the mitral valve.

[0094] Understandably, candidate sampling points corresponding to the largest area are selected based on the cross-sectional area. Then, invalid candidate sampling points above the left ventricular junction and the mitral valve center point are further eliminated by combining the spatial position constraints of the left ventricular junction and the mitral valve center point to determine the target sampling point. Subsequently, the spatial distance is constructed by calculating the sum of the distances from the target sampling point to the left ventricular junction and the mitral valve center point respectively. The target sampling point corresponding to the largest distance is selected as the target left ventricular key point, avoiding interference from non-target area points and achieving accurate positioning of the key point of the left ventricle.

[0095] The above scheme first obtains the location information of the target aortic valve key points, target mitral valve key points, and target right ventricular key points from the target 3D point cloud data through a key point detection model. Then, it uniformly samples the target 3D point cloud data corresponding to the target left ventricular region to obtain candidate sampling points. It then constructs a spatial plane by combining the left non-intersection point and the mitral valve center point, calculates the tangent area, and finally screens out the target left ventricular key points by combining the tangent area and the spatial location information of each key point. This ensures the comprehensiveness of the candidate point distribution. The spatial plane and tangent area are used to achieve quantitative screening of left ventricular key points, which not only improves the accuracy of left ventricular key point positioning but also reduces manual intervention and positioning deviation.

[0096] S240. Based on the key point location information, determine the section position information of the three-dimensional section of the left ventricular interventricular septum region.

[0097] The three-dimensional sections include the short-axis section at the mitral valve level, the long-axis section of the left ventricle, and the four-chamber section.

[0098] For example, based on key point location information, the section position information of the three-dimensional section of the left ventricular septum region is determined, including:

[0099] Using a random sampling consensus algorithm, the short-axis section of the mitral valve is determined based on the target mitral valve key point; the long-axis section of the left ventricle is determined based on the target aortic valve key point, the target mitral valve center point, and the target left ventricle key point; and the four-chamber view is determined based on the center point of the target mitral valve, the target left ventricle key point, and the target right ventricle key point.

[0100] Among them, the Random Sampling Consensus (RANSAC) algorithm uses the target mitral lobe center point as the center point of the target 3D point cloud data corresponding to the target mitral lobe region.

[0101] Specifically, leveraging the robust fitting capability of the RANSAC algorithm, the short-axis section of the mitral valve is first fitted using the target mitral valve key points as the core localization basis. The long-axis section of the left ventricle is then determined by constraining the spatial orientation using the aortic valve key points, the mitral valve center point, and the left ventricle key points. Finally, the four-chamber view is fitted using the spatial relationship between the mitral valve center point, the left ventricle key point, and the right ventricle center point. It should be noted that three non-collinear points can uniquely define a plane; therefore, the plane formed by the aortic valve key points, the mitral valve center point, and the left ventricle key points is the long-axis section of the left ventricle. The plane formed by the mitral valve center point, the left ventricle key point, and the right ventricle center point is the four-chamber view.

[0102] The above scheme can fully utilize the anatomical spatial relationships of key points to achieve automatic positioning of standard sections, ensuring that the section positions of the four-chamber view, the long-axis view of the left ventricle, and the short-axis view at the mitral valve level are highly matched with the actual anatomical structure of the heart. Simultaneously, leveraging the robustness of the RANSAC algorithm, it can effectively eliminate interfering information, improve the stability and accuracy of three-dimensional section positioning, and avoid section offset problems caused by manual positioning or simple geometric calculations, making the obtained three-dimensional sections of the left ventricular septum region more reliable.

[0103] S250. Determine the interventricular septum thickness in the left ventricular septum region based on the section position information of the three-dimensional section.

[0104] For example, on the long-axis and four-chamber views of the left ventricle, measurement points are selected every 15 mm along the left ventricular ventricle wall. Rays are emitted from these points along the normal direction of the view. The second-to-last point where the emitted ray intersects the myocardial wall is the starting point, and the furthest intersection point is the ending point. The distance between the starting and ending points represents the myocardial thickness at the corresponding measurement point. On the short-axis view at the mitral valve level, distance measurement lines are emitted outwards every 30° from the center of the left ventricular ventricle wall, directly measuring the distance between the inner and outer walls of the myocardium. This allows for efficient and accurate measurement of myocardial thickness at different locations and segments within the left ventricular septum on multiple standardized views, providing a reliable preoperative distance measurement basis for myocardial resection surgery.

[0105] In the above method for determining interventricular septal thickness, a cardiac scan image is acquired, and the cardiac scan image is segmented using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan image. The target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region. The segmentation mask of the target detection region is converted using a moving cube algorithm to determine the target three-dimensional point cloud data corresponding to each segmentation mask. Based on the target three-dimensional point cloud data and the key point detection model, the key point location information of the cardiac key points is determined. The cardiac key points include the target aortic valve key points, the target mitral valve key points, the target right ventricle key points, and the target left ventricle key points. Based on the key point location information, the section position information of the three-dimensional section of the left ventricular interventricular septal region is determined. The three-dimensional section includes the mitral valve horizontal short-axis section, the left ventricular long-axis section, and the four-chamber section. Based on the section position information of the three-dimensional section, the interventricular septal thickness of the left ventricular interventricular septal region is determined. This solution addresses the shortcomings of traditional two-dimensional echocardiography-based methods for measuring interventricular septum thickness. These methods fail to fully reflect the three-dimensional morphology of the septum and require physicians to manually sift through numerous two-dimensional tomographic images, leading to inaccurate measurements. The proposed method first acquires cardiac scan images. An image segmentation model is used to obtain segmentation masks for the aortic valve, mitral valve, left ventricular, and right ventricular regions. A moving cubes algorithm then converts these masks into corresponding three-dimensional point cloud data. A key point detection model determines the locations of key cardiac points. Based on these key point locations, three types of three-dimensional sections are determined: a short-axis section at the mitral valve level, a long-axis section at the left ventricular level, and a four-chamber section. The thickness of the left ventricular septum is measured based on the position of these three-dimensional sections. This approach ensures that the determined three-dimensional sections conform to clinical anatomical standards and avoids the errors caused by manual positioning of the sections. Determining the septal thickness through the position of the three-dimensional sections allows for efficient and accurate acquisition of multi-directional septal thickness data, providing precise data for preoperative planning of procedures such as myocardial resection and improving the accuracy of the determined septal thickness.

[0106] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0107] Based on the above scheme, the methods for determining interventricular septal thickness include:

[0108] Sample cardiac images were collected as training data for the model. Historically acquired cardiac CTA images containing the left ventricular interventricular septum region were selected as sample images. The corresponding annotation information for these images was obtained, covering the aortic valve region, mitral valve region, left ventricular region, and right ventricular region. Using the sample cardiac images and their corresponding annotation information, the nnUnet deep learning network was trained in a supervised manner. The nnUnet deep learning network learned the mapping relationship between image features and target detection regions, ultimately obtaining an image segmentation model usable for practical segmentation. High-resolution three-dimensional image data of the left ventricular interventricular septum region, i.e., cardiac scan images, or cardiac CTA images, were acquired using CTA technology. The image segmentation model was used to segment the cardiac scan images to extract the segmentation mask for the target detection region.

[0109] The moving cube algorithm is used to reconstruct the three-dimensional surface of the target detection area, converting the two-dimensional mask data, which originally existed in the form of voxels, into a discrete set of three-dimensional spatial points, thereby obtaining the three-dimensional point cloud data corresponding to each segmentation mask.

[0110] First, the 3D point cloud data and its annotation information corresponding to the sample cardiac images are obtained. The annotation information includes the cardiac structure labels corresponding to the sample 3D point cloud data, as well as the coordinates of key points of the aortic valve, mitral valve, and right ventricle. Then, the PointNet++ network model is used as the base network model, and the sample 3D point cloud data and its annotation information are used as training data. Based on the cross-entropy loss function, three key point detection models are trained: an aortic valve detection model with 1 key point, a mitral valve detection model with 3 key points, and a right ventricle detection model with 1 key point. During the model prediction stage of the key point detection model, a targeted threshold of 0.1 is applied to the model prediction results, meaning only positions with a confidence score greater than 0.1 are retained as the final detected key points, thereby controlling the number of key point location information output by the model. The key point detection models include the aortic valve detection model, the mitral valve detection model, and the right ventricle detection model. The cardiac structure labels include aortic valve region labels, mitral valve region labels, left ventricle region labels, and right ventricle region labels.

[0111] The target aortic valve key point is the junction of the left coronary lobes and the non-coronary lobes of the target aortic valve. The target mitral valve key points include the junction of the two target mitral valve leaflets and the midpoint of the posterior leaflet of the target mitral valve. The target right ventricular key point is the midpoint of the tricuspid valve. The target left ventricular key point is the apex of the left ventricle.

[0112] Based on the 3D point cloud data of each target detection region obtained by the moving cubes algorithm, the 3D point cloud data of each target detection region is input into a trained keypoint detection model. The keypoint detection model identifies and locates the features of the 3D point cloud data, thereby outputting the keypoint location information of the target aortic valve keypoint, the target mitral valve keypoint, and the target right ventricle keypoint. Specifically, the target 3D point cloud data corresponding to the target aortic valve region is input into the aortic valve detection model to determine the keypoint location information of the target aortic valve keypoint; the target 3D point cloud data corresponding to the target mitral valve region is input into the mitral valve detection model to determine the keypoint location information of the target mitral valve keypoint; and the target 3D point cloud data corresponding to the target right ventricle region is input into the right ventricle detection model to determine the keypoint location information of the target right ventricle keypoint. The left ventricular region point cloud data corresponding to the target left ventricular region is determined from the target 3D point cloud data, and the keypoint location information of the target left ventricular keypoint is determined based on the left ventricular region point cloud data.

[0113] The target 3D point cloud data corresponding to the target aortic valve region is input into the aortic valve detection model to determine the 3D coordinates of key points of the target aortic valve as the key point location information. Similarly, the target 3D point cloud data corresponding to the target mitral valve region is input into the mitral valve detection model to determine the 3D coordinates of key points of the target mitral valve as the key point location information. The target 3D point cloud data corresponding to the target right ventricle region is input into the right ventricle detection model to determine the 3D coordinates of key points of the target right ventricle as the key point location information. The 3D point cloud data corresponding to the left ventricle region is uniformly sampled to obtain multiple evenly distributed candidate sampling points. Each candidate sampling point is then combined with the detected left non-intersection point and mitral valve center point to construct a spatial plane composed of these three points. The tangent area of ​​each spatial plane to the original left ventricle point cloud data in the target 3D point cloud data corresponding to the target left ventricle region is determined. The maximum area is determined from the cross-sectional area, and the candidate sampling point corresponding to the maximum area is identified as the sampling point to be screened. Based on the sampling point location information of the sampling point to be screened, the location information of the left ventricular junction, and the location information of the center point of the mitral valve, the sampling points located below the left ventricular junction and the center point of the mitral valve are identified as the target sampling points. Based on the sampling point location information of the target sampling point and the location information of the left ventricular junction, a first distance between the target sampling point and the left ventricular junction is determined, and based on the sampling point location information of the target sampling point and the location information of the center point of the mitral valve, a second distance between the target sampling point and the center point of the mitral valve is determined. The sum of the first distance and the second distance is taken as the spatial distance between the target sampling point and the left ventricular junction and the center point of the mitral valve. The maximum distance is determined from the spatial distances, and the sampling point location information of the target sampling point corresponding to the maximum distance is identified as the key point location information of the target left ventricular key point.

[0114] Leveraging the robust fitting capability of the RANSAC algorithm, the short-axis section of the mitral valve is first fitted using key points of the target mitral valve as the core localization basis. The long-axis section of the left ventricle is determined by constraining the spatial orientation using key points of the aortic valve, the mitral valve center point, and the left ventricle. Finally, a four-chamber view is fitted using the spatial relationship between the mitral valve center point, the left ventricle center point, and the right ventricle center point. It should be noted that three non-collinear points can uniquely define a plane; therefore, the plane formed by the key points of the aortic valve, the mitral valve center point, and the left ventricle is the long-axis section of the left ventricle. The plane formed by the mitral valve center point, the left ventricle center point, and the right ventricle center point is the four-chamber view.

[0115] In the long-axis and four-chamber views of the left ventricle, measurement points are selected every 15 mm along the left ventricular wall. Rays are emitted from these points along the normal direction of the view. The second-to-last point where the emitted ray intersects the myocardial wall is the starting point, and the furthest intersection point is the ending point. The distance between the starting and ending points represents the myocardial thickness at the corresponding measurement point. In the short-axis view at the mitral valve level, distance measurement lines are emitted outwards every 30° from the center of the left ventricular wall, directly measuring the distance between the inner and outer walls of the myocardium. This allows for efficient and accurate measurement of myocardial thickness at different locations and segments within the left ventricular septum on multiple standardized views, providing reliable preoperative distance measurement data for myocardial resection surgery.

[0116] The above scheme acquires cardiac scan images and performs image segmentation on the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan image; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region; the moving cube algorithm is used to transform the segmentation mask of the target detection region to determine the target three-dimensional point cloud data corresponding to each segmentation mask; based on the target three-dimensional point cloud data and the key point detection model, the key point location information of cardiac key points is determined; the cardiac key points include the target aortic valve key points, the target mitral valve key points, the target right ventricle key points, and the target left ventricle key points; based on the key point location information, the section position information of the three-dimensional section of the left ventricular interventricular septum region is determined; the three-dimensional section includes the mitral valve horizontal short axis section, the left ventricular long axis section, and the four-chamber section; based on the section position information of the three-dimensional section, the interventricular septum thickness of the left ventricular interventricular septum region is determined. This solution addresses the shortcomings of traditional two-dimensional echocardiography-based methods for measuring interventricular septum thickness. These methods fail to fully reflect the three-dimensional morphology of the septum and require physicians to manually sift through numerous two-dimensional tomographic images, leading to inaccurate measurements. The proposed method first acquires cardiac scan images. An image segmentation model is used to obtain segmentation masks for the aortic valve, mitral valve, left ventricular, and right ventricular regions. A moving cubes algorithm then converts these masks into corresponding three-dimensional point cloud data. A key point detection model determines the locations of key cardiac points. Based on these key point locations, three types of three-dimensional sections are determined: a short-axis section at the mitral valve level, a long-axis section at the left ventricular level, and a four-chamber section. The thickness of the left ventricular septum is measured based on the position of these three-dimensional sections. This approach ensures that the determined three-dimensional sections conform to clinical anatomical standards and avoids the errors caused by manual positioning of the sections. Determining the septal thickness through the position of the three-dimensional sections allows for efficient and accurate acquisition of multi-directional septal thickness data, providing precise data for preoperative planning of procedures such as myocardial resection and improving the accuracy of the determined septal thickness.

[0117] Based on the same inventive concept, this application also provides an interventricular septum thickness determination device for implementing the above-described interventricular septum thickness determination method. The solution provided by this device is similar to the solution described in the above-described method; therefore, the specific limitations in one or more embodiments of the interventricular septum thickness determination device provided below can be found in the limitations of the interventricular septum thickness determination method described above, and will not be repeated here.

[0118] In one embodiment, such as Figure 4 As shown, a device for determining interventricular septal thickness is provided, comprising: an image segmentation module 401, a point cloud data acquisition module 402, a key point determination module 403, a three-dimensional cross-section determination module 404, and an interventricular septal thickness determination module 405, wherein:

[0119] Image segmentation module 401 is used to acquire cardiac scan images and perform image segmentation on the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0120] The point cloud data acquisition module 402 is used to convert the segmentation mask of the target detection area using the moving cube algorithm, and to determine the target three-dimensional point cloud data corresponding to each segmentation mask;

[0121] The key point determination module 403 is used to determine the key point location information of the heart key points based on the target three-dimensional point cloud data and the key point detection model; the heart key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point and the target left ventricle key point;

[0122] The three-dimensional section determination module 404 is used to determine the section position information of the three-dimensional section of the left ventricular interventricular septum region based on the key point position information; the three-dimensional section includes the mitral valve level short axis section, the left ventricular long axis section and the four-chamber section;

[0123] The interventricular septum thickness determination module 405 is used to determine the interventricular septum thickness of the left ventricle septum region based on the section position information of the three-dimensional section.

[0124] For example, the target aortic valve key point is the junction of the left coronary lobes and the non-coronary lobes of the target aortic valve; the target mitral valve key point includes the intersection of two target mitral valve leaflets and the midpoint of the posterior leaflet of the target mitral valve; the target right ventricular key point is the midpoint of the tricuspid valve; and the target left ventricular key point is the apex of the left ventricle.

[0125] For example, the key point determination module 403 is specifically used for:

[0126] Based on the target 3D point cloud data and key point detection model, determine the key point location information of the target aortic valve key point, the target mitral valve key point, and the target right ventricle key point;

[0127] Uniform sampling is performed on the target three-dimensional point cloud data corresponding to the target left ventricular region to determine candidate sampling points, and the spatial plane formed by each candidate sampling point, the left ventricular junction point, and the mitral valve center point is determined.

[0128] Determine the area of ​​the cross section after each spatial plane is tangent to the original left ventricular point cloud in the target three-dimensional point cloud data corresponding to the target left ventricular region;

[0129] Based on the cross-sectional area, the sampling point location information of the candidate sampling points, the boundary point location information of the left non-intersection point, and the center point location information of the mitral valve center point, the target left ventricular key point is determined from the candidate sampling points, and the key point location information of the target left ventricular key point is determined.

[0130] For example, the key point determination module 403 is also specifically used for:

[0131] The largest area is determined from the cross-sectional areas, and the candidate sampling point corresponding to the largest area is determined as the sampling point to be screened;

[0132] Based on the sampling point location information of the sampling point to be screened, the boundary point location information of the left no-boundary point and the center point location information of the mitral valve center point, the sampling point to be screened located below the left no-boundary point and the mitral valve center point is determined, and the sampling point to be screened located below the left no-boundary point and the mitral valve center point is taken as the target sampling point.

[0133] Based on the sampling point location information of the target sampling point and the boundary point location information of the left boundary point, a first distance between the target sampling point and the left boundary point is determined, and based on the sampling point location information of the target sampling point and the center point location information of the mitral valve center point, a second distance between the target sampling point and the mitral valve center point is determined.

[0134] The sum of the first distance and the second distance is taken as the spatial distance between the target sampling point and the left boundary point and the mitral valve center point;

[0135] The maximum distance is determined from the spatial distance, and the sampling point location information of the target sampling point corresponding to the maximum distance is determined as the key point location information of the target left ventricular key point.

[0136] For example, the three-dimensional section determination module 404 is specifically used for:

[0137] The horizontal short axis section of the mitral valve is determined based on the key points of the target mitral valve using a random sampling consensus algorithm.

[0138] Determine the long axis section of the left ventricle based on the target aortic valve key point, the target mitral valve center point, and the target left ventricle key point;

[0139] The four-chamber view is determined based on the center points of the target mitral valve, the target left ventricular key point, and the target right ventricular key point.

[0140] For example, the above-mentioned interventricular septum thickness determination device further includes a first model training module, specifically used for:

[0141] Obtain sample cardiac images and sample annotation information of the sample cardiac images; the sample annotation information includes annotation information of the sample aortic region, sample aortic valve region, sample mitral valve region, sample left ventricular region, and sample right ventricular region of the sample cardiac images;

[0142] The deep learning network is trained using the heart-related information of the samples and the sample annotation information to determine the image segmentation model.

[0143] For example, the above-mentioned interventricular septum thickness determination device further includes a second model training module, specifically used for:

[0144] Acquire the sample three-dimensional point cloud data corresponding to the sample cardiac image and the annotation information of the sample three-dimensional point cloud data; the annotation information of the sample three-dimensional point cloud data includes the cardiac structure label corresponding to the sample three-dimensional point cloud data, and the key point coordinates of the sample key points in the sample three-dimensional point cloud data; the sample key points include the sample aortic valve key points, the sample mitral valve key points and the sample right ventricle key points.

[0145] Based on the sample 3D point cloud data and the annotation information of the sample 3D point cloud data, the point cloud network to be trained is trained to determine the key point detection model; the key point detection model includes an aortic valve detection model, a mitral valve detection model and a right ventricle detection model.

[0146] Each module in the aforementioned interventricular septum thickness determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0147] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for determining interventricular septal thickness. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or a key vector, trackball, or touchpad set on the computer device casing, or an external key vector disk, touchpad, or mouse, etc.

[0148] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does 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 those shown in the figure, or combine certain components, or have different component arrangements.

[0149] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0150] Step 1: Acquire cardiac scan images and segment the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0151] Step 2: Use the moving cube algorithm to transform the segmentation mask of the target detection region, and determine the target 3D point cloud data corresponding to each segmentation mask;

[0152] Step 3: Based on the target 3D point cloud data and key point detection model, determine the key point location information of the heart key points; the heart key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point;

[0153] Step 4: Based on the key point location information, determine the section position information of the three-dimensional section of the left ventricular septum region; the three-dimensional section includes the mitral valve level short-axis section, the left ventricular long-axis section, and the four-chamber section;

[0154] Step 5: Determine the interventricular septum thickness in the left ventricular septum region based on the section position information of the three-dimensional section.

[0155] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0156] Step 1: Acquire cardiac scan images and segment the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0157] Step 2: Use the moving cube algorithm to transform the segmentation mask of the target detection region, and determine the target 3D point cloud data corresponding to each segmentation mask;

[0158] Step 3: Based on the target 3D point cloud data and key point detection model, determine the key point location information of the heart key points; the heart key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point;

[0159] Step 4: Based on the key point location information, determine the section position information of the three-dimensional section of the left ventricular septum region; the three-dimensional section includes the mitral valve level short-axis section, the left ventricular long-axis section, and the four-chamber section;

[0160] Step 5: Determine the interventricular septum thickness in the left ventricular septum region based on the section position information of the three-dimensional section.

[0161] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0162] Step 1: Acquire cardiac scan images and segment the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region;

[0163] Step 2: Use the moving cube algorithm to transform the segmentation mask of the target detection region, and determine the target 3D point cloud data corresponding to each segmentation mask;

[0164] Step 3: Based on the target 3D point cloud data and key point detection model, determine the key point location information of the heart key points; the heart key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point;

[0165] Step 4: Based on the key point location information, determine the section position information of the three-dimensional section of the left ventricular septum region; the three-dimensional section includes the mitral valve level short-axis section, the left ventricular long-axis section, and the four-chamber section;

[0166] Step 5: Determine the interventricular septum thickness in the left ventricular septum region based on the section position information of the three-dimensional section.

[0167] 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 related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0168] 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. When executed, the computer program 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). 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.

[0169] 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.

[0170] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining interventricular septum thickness, characterized in that, include: Acquire cardiac scan images and segment the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region; The moving cube algorithm is used to transform the segmentation mask of the target detection region, and the target 3D point cloud data corresponding to each segmentation mask is determined respectively; Based on the target 3D point cloud data and key point detection model, determine the key point location information of the heart key point; The cardiac key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point; Based on the key point location information, determine the section location information of the three-dimensional section of the left ventricular septum region; The three-dimensional sections include the mitral valve level short-axis section, the left ventricular long-axis section, and the four-chamber section; Based on the section position information of the three-dimensional section, the interventricular septum thickness in the left ventricular septum region is determined.

2. The method according to claim 1, characterized in that, The target aortic valve key point is the junction of the left coronary lobes and the non-coronary lobes of the target aortic valve. The target mitral valve key point includes the intersection of two target mitral valve leaflets and the midpoint of the posterior leaflet of the target mitral valve. The target right ventricular key point is the midpoint of the tricuspid valve. The target left ventricular key point is the apex of the left ventricle.

3. The method according to claim 2, characterized in that, Based on the target 3D point cloud data and key point detection model, the key point location information of the heart key points is determined, including: Based on the target 3D point cloud data and key point detection model, determine the key point location information of the target aortic valve key point, the target mitral valve key point, and the target right ventricle key point; Uniform sampling is performed on the target three-dimensional point cloud data corresponding to the target left ventricular region to determine candidate sampling points, and the spatial plane formed by each candidate sampling point, the left ventricular junction point, and the mitral valve center point is determined. Determine the area of ​​the cross section after each spatial plane is tangent to the original left ventricular point cloud in the target three-dimensional point cloud data corresponding to the target left ventricular region; Based on the cross-sectional area, the sampling point location information of the candidate sampling points, the boundary point location information of the left non-intersection point, and the center point location information of the mitral valve center point, the target left ventricular key point is determined from the candidate sampling points, and the key point location information of the target left ventricular key point is determined.

4. The method according to claim 3, characterized in that, Based on the cross-sectional area, the sampling point location information of the candidate sampling points, the boundary point location information of the left ventricular non-boundary point, and the center point location information of the mitral valve center point, the target left ventricular key point is determined from the candidate sampling points, and the key point location information of the target left ventricular key point is determined, including: The largest area is determined from the cross-sectional areas, and the candidate sampling point corresponding to the largest area is determined as the sampling point to be screened; Based on the sampling point location information of the sampling point to be screened, the boundary point location information of the left no-boundary point and the center point location information of the mitral valve center point, the sampling point to be screened located below the left no-boundary point and the mitral valve center point is determined, and the sampling point to be screened located below the left no-boundary point and the mitral valve center point is taken as the target sampling point. Based on the sampling point location information of the target sampling point and the boundary point location information of the left boundary point, a first distance between the target sampling point and the left boundary point is determined, and based on the sampling point location information of the target sampling point and the center point location information of the mitral valve center point, a second distance between the target sampling point and the mitral valve center point is determined. The sum of the first distance and the second distance is taken as the spatial distance between the target sampling point and the left boundary point and the mitral valve center point; The maximum distance is determined from the spatial distance, and the sampling point location information of the target sampling point corresponding to the maximum distance is determined as the key point location information of the target left ventricular key point.

5. The method according to claim 1, characterized in that, Based on the key point location information, the section position information of the three-dimensional section of the left ventricular septum region is determined, including: The horizontal short axis section of the mitral valve is determined based on the key points of the target mitral valve using a random sampling consensus algorithm. Determine the long axis section of the left ventricle based on the target aortic valve key point, the target mitral valve center point, and the target left ventricle key point; The four-chamber view is determined based on the center points of the target mitral valve, the target left ventricular key point, and the target right ventricular key point.

6. The method according to claim 1, characterized in that, Also includes: Obtain sample cardiac images and sample annotation information of the sample cardiac images; the sample annotation information includes annotation information of the sample aortic region, sample aortic valve region, sample mitral valve region, sample left ventricular region, and sample right ventricular region of the sample cardiac images; The deep learning network is trained using the heart-related information of the samples and the sample annotation information to determine the image segmentation model.

7. The method according to claim 1, characterized in that, Also includes: Obtain the sample three-dimensional point cloud data corresponding to the sample cardiac image and the annotation information of the sample three-dimensional point cloud data; The annotation information of the sample three-dimensional point cloud data includes the cardiac structure label corresponding to the sample three-dimensional point cloud data, and the key point coordinates of the sample key points in the sample three-dimensional point cloud data; the sample key points include the sample aortic valve key points, the sample mitral valve key points and the sample right ventricle key points. Based on the sample 3D point cloud data and the annotation information of the sample 3D point cloud data, the point cloud network to be trained is trained to determine the key point detection model; the key point detection model includes an aortic valve detection model, a mitral valve detection model and a right ventricle detection model.

8. A device for determining interventricular septum thickness, characterized in that, The interventricular septum thickness determination device includes: The image segmentation module is used to acquire cardiac scan images and perform image segmentation on the cardiac scan images using an image segmentation model to determine the segmentation mask of the target detection region in the cardiac scan images; the target detection region includes the target aortic valve region, the target mitral valve region, the target left ventricle region, and the target right ventricle region; The point cloud data acquisition module is used to convert the segmentation mask of the target detection area using the moving cube algorithm, and to determine the target 3D point cloud data corresponding to each segmentation mask; The key point determination module is used to determine the key point location information of cardiac key points based on the target 3D point cloud data and the key point detection model; the cardiac key points include the target aortic valve key point, the target mitral valve key point, the target right ventricle key point, and the target left ventricle key point; The three-dimensional section determination module is used to determine the section position information of the three-dimensional section of the left ventricular interventricular septum region based on the key point position information; the three-dimensional section includes the mitral valve level short axis section, the left ventricular long axis section, and the four-chamber section; The interventricular septum thickness determination module is used to determine the interventricular septum thickness in the left ventricular septum region based on the section position information of the three-dimensional section.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images