Evaluation index detection method and device for evaluating mitral valve opening stenosis degree

Abstract

The invention relates to the technical field of image processing, and discloses an evaluation index detection method and device for evaluating the stenosis degree of a mitral valve orifice, and the method comprises the steps: obtaining a first target echocardiography image which is an image of the horizontal section of a para-sternum minor-axis mitral valve; inputting the first target echocardiography image into a target detection model, and determining a first region of interest where a mitral valve opening is located in the first target echocardiography image and first target information corresponding to the first region of interest, the first target information comprises a center point coordinate of the detection frame and size information of the detection frame; according to the first target information, segmenting the first region of interest by using a mitral valve orifice segmentation model to obtain a mask image of the mitral valve orifice; and according to the number of pixels contained in the mask image of the mitral valve orifice, determining the area of a mitral valve orifice region in a first region of interest where the mitral valve orifice is located.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to a method, apparatus, computer device, storage medium, and computer program product for detecting assessment indicators to evaluate the degree of mitral valve stenosis. Background Technology

[0002] Heart valve stenosis commonly affects the mitral valve, with thickening, calcification, and adhesions at the mitral valve being the main pathological structural changes. These changes can lead to a series of cardiac structural changes and functional failure. Echocardiography, due to its real-time nature, convenience, and cost-effectiveness, has become the preferred auxiliary diagnostic method for this type of disease.

[0003] In related technologies, mitral valve orifice detection methods used to assess the degree of mitral stenosis include: manual valve orifice area recording, which involves manually outlining the mitral valve orifice region in echocardiograms and calculating the mitral valve orifice area using geometric tools. This method is time-consuming, labor-intensive, involves many steps, and is limited by experience. Therefore, there is an urgent need to propose a detection method for assessment indicators of the degree of mitral valve stenosis to improve the detection efficiency and accuracy of parameters such as mitral valve orifice area. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method, apparatus, computer equipment, storage medium, and computer program product for detecting assessment indicators for evaluating the degree of mitral valve stenosis, in order to solve the technical problem of low efficiency and accuracy of existing methods for determining indicators for assessing the degree of mitral valve stenosis.

[0005] According to a first aspect, embodiments of the present invention provide a method for detecting assessment indicators for evaluating the degree of mitral valve stenosis, comprising: acquiring a first target echocardiogram, wherein the first target echocardiogram is an image of a short-axis mitral valve horizontal section parasternally; inputting the first target echocardiogram into a target detection model to determine a first region of interest (ROI) in the first target echocardiogram containing the mitral valve orifice and first target information corresponding to the first ROI, wherein the first target information includes the coordinates of the center point of a detection frame and the size information of the detection frame; segmenting the first ROI using a mitral valve orifice segmentation model based on the first target information to obtain a mask image of the mitral valve orifice; and determining the area of ​​the mitral valve orifice region within the first ROI containing the mitral valve orifice based on the number of pixels contained in the mitral valve orifice mask image.

[0006] Optionally, after determining the area of ​​the mitral valve orifice region within the first region of interest based on the number of pixels contained in the mask image of the mitral valve orifice, the method further includes: obtaining the pixels on the contour boundary of the mitral valve orifice region and the corresponding pixel coordinates; traversing the distance between any two pixels and taking the line connecting the two pixels with the largest distance as the major axis of the mitral valve orifice region; and in the vertical direction of the major axis, according to the traversal operation, taking the line connecting the two endpoints with the largest distance in the vertical direction as the minor axis of the mitral valve orifice region.

[0007] Optionally, the method further includes: when acquiring multiple first target echocardiograms, determining the area of ​​the mitral valve orifice region corresponding to each first target echocardiogram; and constructing area change characterization data corresponding to different first target echocardiograms based on each first target echocardiogram and the area of ​​the corresponding mitral valve orifice region.

[0008] Optionally, the method further includes: determining the largest mitral valve orifice region in all the first target echocardiograms; and displaying the outline of the corresponding mitral valve orifice region, the major axis and the minor axis, and the ratio information of the major axis and the minor axis in the first target echocardiogram corresponding to the largest mitral valve orifice region.

[0009] Optionally, the method further includes: acquiring a second target echocardiogram, wherein the second target echocardiogram is an image of the parasternal long axis section; inputting the second target echocardiogram into a key point detection model to determine the second target information corresponding to the second region of interest where the mitral valve orifice is located in the second target echocardiogram, as well as the positions of the two endpoints of the narrowest part of the mitral valve orifice in the second region of interest and the corresponding endpoint spacing, wherein the second target information includes the coordinates of the center point of the detection box and the size information of the detection box.

[0010] Optionally, the method further includes: when multiple second target echocardiograms are acquired, determining the endpoint spacing corresponding to each second target echocardiogram; and constructing endpoint spacing variation characterization data corresponding to different second target echocardiograms based on each second target echocardiogram and its corresponding endpoint spacing.

[0011] Optionally, the method further includes: determining the maximum endpoint spacing corresponding to the narrowest point of the mitral valve orifice in all the second target echocardiograms; and displaying the corresponding endpoint position and maximum endpoint spacing information in the second target echocardiogram corresponding to the maximum endpoint spacing.

[0012] Optionally, the method further includes: performing contrast enhancement processing on the first target echocardiogram or the second target echocardiogram; and performing connected component analysis processing on the enhanced image using a preset topology filtering algorithm.

[0013] Optionally, the method further includes: acquiring multiple first target echocardiogram training data, each of the first target echocardiogram training data having a mitral valve orifice region marked; training a target model using the multiple first target echocardiogram training data to obtain the mitral valve orifice segmentation model, wherein the target model integrates a residual network structure and a spatial attention module.

[0014] According to a second aspect, embodiments of the present invention provide an assessment index detection device for evaluating the degree of mitral valve stenosis, comprising: a first acquisition module for acquiring a first target echocardiogram, wherein the first target echocardiogram is an image of a short-axis mitral valve horizontal section parasternally; a first determination module for inputting the first target echocardiogram into a target detection model to determine a first region of interest (ROI) in the first target echocardiogram and first target information corresponding to the mitral valve orifice, wherein the first target information includes the coordinates of the center point of a detection frame and the size information of the detection frame; a segmentation module for segmenting the first ROI using a mitral valve orifice segmentation model based on the first target information to obtain a mask image of the mitral valve orifice; and a second determination module for determining the area of ​​the mitral valve orifice region within the first ROI region based on the number of pixels contained in the mitral valve orifice mask image.

[0015] According to a third aspect, embodiments of the present invention provide a computer device, including: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the assessment index detection method for assessing the degree of mitral valve stenosis as described in the first aspect or any optional embodiment of the first aspect.

[0016] According to a fourth aspect, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing a computer to perform the assessment index detection method for assessing the degree of mitral valve stenosis as described in the first aspect or any optional embodiment of the first aspect.

[0017] According to a fifth aspect, embodiments of the present invention provide a computer program product, including computer instructions for causing a computer to execute the assessment index detection method for assessing the degree of mitral valve stenosis as described in the first aspect or any optional embodiment of the first aspect.

[0018] The method for detecting assessment indicators for evaluating the degree of mitral valve stenosis provided in this invention determines the region of interest (ROI) of the mitral valve orifice in the target echocardiogram using a target detection model, obtains a mask image of the mitral valve orifice based on the target information of the ROI using a mitral valve orifice segmentation model, and directly determines the area of ​​the mitral valve orifice region based on the number of pixels in the mask image. Compared with the method of determining the mitral valve orifice region area parameters by combining manual operation, the technical solution provided in this application improves the detection efficiency and accuracy of the mitral valve orifice region area parameters. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method for detecting assessment indicators for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0024] Figure 5 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0025] Figure 6 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0026] Figure 7 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0027] Figure 8 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0028] Figure 9 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0029] Figure 10 This is a schematic diagram of the assessment index detection method for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0030] Figure 11 This is a structural block diagram of an assessment index detection device for evaluating the degree of mitral valve stenosis according to an embodiment of the present invention;

[0031] Figure 12 This is a schematic diagram of the hardware structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] According to an embodiment of the present invention, an embodiment of a method for detecting assessment indicators for evaluating the degree of mitral valve stenosis is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0034] This embodiment provides a method for detecting assessment indicators to evaluate the degree of mitral valve stenosis. This method can be applied to an electronic device that is communicatively connected to an ultrasound device used to acquire echocardiographic images. Figure 1 As shown, the process includes the following steps:

[0035] Step S101: Obtain a first target echocardiogram image, which is a parasternal short-axis mitral valve horizontal section image; the target echocardiogram image can be any image from multiple cycles of parasternal short-axis mitral valve horizontal section images acquired by the ultrasound device.

[0036] Step S102: Input the first target echocardiogram into the target detection model to determine the first region of interest where the mitral valve orifice is located in the first target echocardiogram and the first target information corresponding to the first region of interest. The first target information includes the coordinates of the center point of the detection box and the size information of the detection box.

[0037] For example, this target detection model can be trained based on a large number of echocardiogram training samples with regions of interest labeled at the mitral valve orifice, as shown in the labeling results. Figure 2 As shown, the "yellow box" represents the region of interest (ROI) of the mitral valve orifice on the echocardiogram. By inputting the first target echocardiogram image into this target detection model, the confidence level of the ROI of the mitral valve orifice can be predicted. The integrity of the mitral valve orifice structure is judged by the confidence level. When the confidence level is greater than a threshold, the current slice image is a horizontal mitral valve image along the short axis of the target's sternal border. When the confidence level is less than the threshold, the current slice image is an invalid noise image. The model obtains the first ROI of the mitral valve orifice in the horizontal mitral valve image along the short axis of the target's sternal border, as well as the coordinates and size information (such as the length and width of the detection box) of the corresponding ROI detection frame. Specifically, as shown... Figure 3 As shown, the "yellow box" represents the first region of interest where the mitral valve orifice is located, as detected by the current first target echocardiogram using the target detection model. The confidence level (MV) of the target detection result is 0.95.

[0038] Specifically, the network structure corresponding to this object detection model can include, but is not limited to, R-CNN or Faster-R-CNN. This embodiment uses Faster-R-CNN as an example. Faster-R-CNN is an improvement on the classic object detection model R-CNN. Faster-R-CNN introduces a Region Proposal Network (RPN) that shares the convolutional features of the image with the detection network. Faster-R-CNN consists of two modules: a deep convolutional neural network (VGG-16) and a Region Proposal Network (RPN). The deep convolutional neural network (VGG-16) has 13 shared convolutional layers for extracting image features. The features from the last shared convolutional layer are input to the RPN to generate proposals for target regions. The input convolutional feature map is divided into n*n small grid feature maps, and the feature map in each small grid is mapped to a low-dimensional feature, which is then input to two parallel fully connected layers. One fully connected layer outputs the classification probability of the target box, and the other fully connected layer outputs the location information of the target box. The object detection model Faster-R-CNN uses a multi-task loss function, as shown in equation (1) below:

[0039]

[0040] Where P is the predicted probability of the target class, and if it is a positive sample, the true label of the target class is P. * The value is 1; if it is a negative sample, the true label P of the target class is 1. * =0; T is a vector representing the four parametric coordinates of the predicted target bounding box.* The true coordinates of the target bounding box; the target classification loss function L cls It is the cross-entropy logarithmic loss function, and the object detection box loss function L. reg (T,T * ) = R(TT) * R is the error between the predicted bounding box and the ground truth bounding box; P*L reg (T,T * The regression loss of the bounding box is applied only when the predicted target class is positive (P). * The target class loss function is activated when the target class is 1 (P* = 0), and is not activated when the predicted target class is negative (P* = 0). The weights of the target class loss function and the target bounding box loss function are balanced by λ. cls N is the normalization coefficient for classification loss; reg This is the normalized coefficient of the regression loss.

[0041] Step S103: Based on the first target information, the first region of interest is segmented using the mitral valve orifice segmentation model to obtain a mask image of the mitral valve orifice.

[0042] For example, this mitral valve orifice segmentation model can be trained based on a large number of image training samples containing mitral valve orifice masks. The mitral valve orifice mask can be generated manually by outlining the contour along the edge of the mitral valve orifice in echocardiographic images, or it can automatically separate the mitral valve orifice from surrounding tissue by setting a threshold based on the pixel intensity of the region of interest (such as the grayscale value of the echocardiographic image); or it can be generated by identifying the mitral valve orifice boundary using operators such as Canny or Sobel. Specifically, it can be as follows: Figure 4 As shown, the "yellow border" is the labeled mitral valve orifice mask. A trained mitral valve orifice segmentation model can directly segment the first region of interest based on the first target information to obtain the mitral valve orifice mask image.

[0043] Step S104: Determine the area of ​​the mitral valve orifice region in the first region of interest where the mitral valve orifice is located based on the number of pixels contained in the mask image of the mitral valve orifice.

[0044] For example, image processing software can be invoked to calculate the number of pixels within the contour of the mitral valve region and combine the number of pixels with the size of each pixel in the image to determine the area of ​​the mitral valve orifice region. For instance, the Statistics module of a 3D Slicer can be used to determine the number of pixels and the area. This embodiment of the application does not limit the type of image processing software used; those skilled in the art can configure it according to actual needs.

[0045] The assessment index detection method for evaluating the degree of mitral valve stenosis provided in this invention determines the region of interest (ROI) of the mitral valve orifice in the target echocardiogram using a target detection model, and obtains a mask image of the mitral valve orifice using a mitral valve orifice segmentation model based on the target information of the ROI. The area of ​​the mitral valve orifice region is directly determined based on the number of pixels in the mask image. Compared with the method of determining the area of ​​the mitral valve orifice region by combining manual operation, this method improves the detection efficiency and accuracy of the mitral valve orifice region area parameter.

[0046] As an optional implementation of this invention, after step S104, the method further includes: obtaining the pixel points on the contour boundary of the mitral valve orifice region and the corresponding pixel point coordinates; traversing the distance between any two pixel points, and taking the line connecting the two pixel points with the largest distance as the major axis of the mitral valve orifice region; in the vertical direction of the major axis, according to the traversal operation, taking the line connecting the two endpoints with the largest distance in the vertical direction as the minor axis of the mitral valve orifice region.

[0047] For example, the pixels and their coordinates on the contour boundary of the mitral valve orifice region are obtained. The distance between any two pixels on the contour boundary is iterated. Specifically, based on the Euclidean distance between any two pixels, the line connecting the two pixels with the largest distance is taken as the major axis of the mitral valve orifice region. Further, in the vertical direction of this major axis, the pixel distances on the contour boundary are iterated to obtain the two endpoints with the largest vertical distance. The line connecting these two endpoints is taken as the minor axis of the mitral valve orifice region. By obtaining the major and minor axes of the mitral valve orifice region, an evaluation index for the degree of mitral valve stenosis can be provided, improving the accuracy of the mitral valve stenosis detection results.

[0048] As an optional embodiment of the present invention, the method further includes: when acquiring multiple first target echocardiograms, determining the area of ​​the mitral valve orifice region corresponding to each first target echocardiogram; and constructing area change characterization data corresponding to different first target echocardiograms based on each first target echocardiogram and the area of ​​the corresponding mitral valve orifice region.

[0049] For example, based on multiple target echocardiograms acquired over multiple cycles, the area of ​​the mitral valve orifice region corresponding to each target echocardiogram is determined according to the method described in the above embodiment. Specifically, the current target echocardiogram changes over time. After completing the detection of the mitral valve orifice area and the distance between the major and minor axes of the current target echocardiogram, the next frame of the target echocardiogram is extracted as the new current target echocardiogram for further detection of the mitral valve orifice area and the distance between the major and minor axes, until the detection of all frames of echocardiograms is completed. Area change characterization data (such as an area change curve) is constructed based on the area of ​​each target echocardiogram and the corresponding mitral valve orifice region. This facilitates the determination of the temporal changes in mitral valve function during the cardiac cycle through characterization data, overcoming the limitations of single-time-point measurements. Dynamic data support for mitral stenosis procedures is provided through parameters such as the amplitude, rate, and duration of area changes.

[0050] As an optional embodiment of the present invention, the method further includes: determining the mitral valve orifice region with the largest area in all the first target echocardiograms; and displaying the outline of the corresponding mitral valve orifice region, the major axis and the minor axis, and the ratio information of the major axis and the minor axis in the first target echocardiogram corresponding to the mitral valve orifice region with the largest area.

[0051] For example, by identifying the largest mitral valve orifice region among all acquired first target echocardiograms, the contour, major and minor axes, and the ratio of the major to minor axes of the corresponding mitral valve orifice region are displayed in the first target echocardiogram containing this largest mitral valve orifice region. By identifying the mitral valve orifice region containing the largest area, it can be used to assess the disease grade or severity, and by displaying the contour, major and minor axes, and the ratio of the major to minor axes of the corresponding mitral valve orifice region in the second target echocardiogram corresponding to the largest mitral valve orifice region, quantitative data can be provided intuitively for assessment, specifically as follows... Figure 5 As shown, the "yellow outline" represents the outline of the mitral valve orifice region, the "pink line" represents the long axis of the mitral valve orifice region, and the "blue line" represents the short axis of the mitral valve orifice region. Figure 5 The major and minor axis ratio information is not shown in the image. Those skilled in the art can display it at the target position in the echocardiogram according to actual needs.

[0052] As an optional implementation of this invention, the method further includes:

[0053] Acquire a second target echocardiogram, which is an image of the parasternal long axis section; the second target echocardiogram can be any image from the multi-cycle parasternal long axis section images acquired by the ultrasound equipment.

[0054] The second target echocardiogram is input into the key point detection model to determine the second target information corresponding to the second region of interest where the mitral valve orifice is located in the second target echocardiogram, as well as the positions of the two endpoints of the narrowest part of the mitral valve orifice in the second region of interest and the corresponding endpoint spacing. The second target information includes the coordinates of the center point of the detection frame and the size information of the detection frame.

[0055] For example, this keypoint detection model can be trained based on a large number of labeled training samples. The labeled training samples may include the region of interest where the mitral valve orifice is located, the endpoint position of the narrowest part of the mitral valve orifice within the region, and the line connecting the endpoints, as shown in the example. Figure 6 As shown, the "yellow box" represents the region of interest (ROI) where the mitral valve orifice is located, and the endpoints of the "pink lines" and the connecting lines represent the endpoint positions and the distance between the endpoints of the narrowest part of the mitral valve orifice within the marked region. By using a trained keypoint detection model to detect the input second target echocardiogram, the coordinates of the center point of the detection box corresponding to the second ROI where the mitral valve orifice is located, the size information of the detection box, and the positions and the distance between the two endpoints of the narrowest part of the mitral valve orifice within the second ROI region can be determined. Specifically, as shown... Figure 7 As shown, the "yellow box" represents the region of interest (ROI) where the mitral valve orifice is predicted, and the endpoints of the "pink lines" and the connecting lines represent the endpoints of the narrowest point of the mitral valve orifice within the predicted region, as well as the distance between the endpoints. The confidence level (MV) of the target detection result is 0.95. Specifically, the keypoint detection model predicts the location information of the second ROI where the mitral valve orifice is located, including the coordinates of the center point of the rectangle, its length, and width. Then, within the predicted rectangle, the model's internal regression algorithm calculates the positions of the two endpoints and the distance between the narrowest points of the mitral valve orifice. By determining the distance between the two endpoints of the narrowest point of the mitral valve orifice, an accurate assessment index of the degree of mitral valve stenosis can be provided.

[0056] In this embodiment, the keypoint detection model is based on a network structure that may include, but is not limited to, end-to-end model detection models such as YOLO or CenterNet. This embodiment uses CenterNet as an example. The keypoint detection model is a convolutional neural network model trained on the CenterNet model. CenterNet can directly learn the target center location through regression, instead of the traditional detection framework that uses a process of classification followed by localization, thus reducing computational complexity. The CenterNet model structure uses a lightweight backbone network, such as ResNet or Hourglass, to extract features from the input image. The output of the backbone is a multi-layer feature map, which provides high-level semantic information for each region of the image. A heatmap is predicted at each location in the feature map, and the value of each heatmap represents the probability that the location is the target center. Specific steps may include: determining a high-resolution heatmap, predicting a heatmap for each target category, and regions with higher values ​​representing potential target centers. During training, target centers are labeled using a Gaussian distribution to reinforce the learning of center point information. Besides the center points, the Centernet model predicts the target size using regression methods. The bounding box generated for each pixel location includes width and height, and the width and height of the detection box are output through two independent regression branches. The loss function used by the Centernet model can include positional loss and class loss, and the specific form of the loss function can be expressed as follows:

[0057]

[0058] Among them, L k The representation of the prediction for the target category, N represents the number of target categories, Y xyc The target classification result predicted by the characterization model, The result represents the classification of the real target labeled by humans. α and β represent the parameters of the focus loss function. α represents the weight factor that balances positive and negative samples. β represents the focus parameter that adjusts the weight of easy and difficult samples. x and y represent the coordinate index in the spatial dimension of the heatmap, corresponding to the pixel or feature point position in the width (x-axis) and height (y-axis) directions of the heatmap. c represents the category index in the channel dimension of the heatmap.

[0059]

[0060] Among them, L off The model represents the prediction of keypoint offsets, where N represents the number of keypoints, P represents the predicted keypoint positions, and R represents the output stride. * O represents the position of the actual keypoints labeled by humans, while O represents the offset of the predicted keypoints.

[0061]

[0062] Among them, L size The representation of the prediction of the target center point, N represents the number of key points, S k The target center point location predicted by the characterization model, It represents the actual center point location of the target, which is manually labeled.

[0063] L det =L k +λ size L size +λ off L off (6)

[0064] Where, λ size The weighting parameter representing the prediction of the target center point can be set to 0.1, λ. off This is a weight parameter for position offset; for example, it can be set to 1, through L. det To determine the effectiveness of model training.

[0065] As an optional embodiment of the present invention, the method further includes: when multiple second target echocardiograms are acquired, determining the endpoint spacing corresponding to each second target echocardiogram; and constructing endpoint spacing variation characterization data corresponding to different second target echocardiograms based on each second target echocardiogram and the corresponding endpoint spacing.

[0066] For example, based on multiple second target echocardiograms acquired over multiple cycles, the endpoint spacing at the narrowest point of the mitral valve orifice corresponding to each second target echocardiogram is determined according to the method described in the above embodiment. Specifically, the current second target echocardiogram changes over time. After completing the endpoint spacing detection at the narrowest point of the mitral valve orifice in the current second target echocardiogram, the next frame of the second target echocardiogram is extracted as a new target echocardiogram, and the endpoint spacing at the narrowest point of the mitral valve orifice is detected again, until the detection of all frames of echocardiograms is completed. Endpoint spacing change characterization data (such as a dynamic spacing change curve) is constructed based on each second target echocardiogram and the corresponding endpoint spacing detection at the narrowest point of the mitral valve orifice. This facilitates the determination of the temporal changes in mitral valve function during the cardiac cycle through the spacing change characterization data, overcoming the limitations of single-time-point measurements and providing dynamic data support for mitral stenosis procedures.

[0067] As an optional embodiment of the present invention, the method further includes: determining the maximum endpoint spacing corresponding to the narrowest point of the mitral valve orifice in all the second target echocardiograms; and displaying the corresponding endpoint position and maximum endpoint spacing information in the second target echocardiogram corresponding to the maximum endpoint spacing.

[0068] For example, by determining the maximum inter-endpoint distance corresponding to the narrowest point of the mitral valve orifice in all acquired second-target echocardiograms, the corresponding endpoint positions and maximum inter-endpoint distance information are displayed in the second-target echocardiogram corresponding to the maximum inter-endpoint distance at the narrowest point of the mitral valve orifice. By determining the maximum inter-endpoint distance corresponding to the narrowest point of the mitral valve orifice, it can be used to assess the disease grade or severity, and displaying the corresponding endpoint positions and maximum inter-endpoint distance information in the corresponding second-target echocardiogram can provide intuitive quantitative data for assessment.

[0069] As an optional embodiment of the present invention, before the first target echocardiogram data or the second target echocardiogram data is input into the corresponding model, the method further includes: performing contrast enhancement processing on the first target echocardiogram image or the second target echocardiogram image; and performing connected component analysis processing on the enhanced image using a preset topology filtering algorithm.

[0070] For example, applying a contrast enhancement algorithm can effectively reduce the impact of noise in an image. Simultaneously, using a pre-defined topology filtering algorithm (such as a dual-tree topology filtering algorithm) can perform connected component analysis on the echocardiogram image, preserving the valve topology while removing speckle noise. Specifically, in this embodiment, the contrast enhancement algorithm employs an adaptive contrast enhancement (ACE) mechanism, and the specific formula used is as follows:

[0071] G(x)=C / (1+σ x (7)

[0072] Where G(x) is the dynamic gain coefficient, C is the gain coefficient, and σ x The standard deviation is the coefficient in the low contrast region (σ). x Small) enhance signal amplitude, while in the high contrast region (σ) x (Large) Suppress over-enhancement to avoid noise amplification.

[0073] I enhanced =m x +G(X)×(I original -m x (8)

[0074] Among them, I enhanced For the enhanced image, I original For the original image, m xFor local mean, through high frequency components (I original -m x Adjustments can be made to enhance the distinction between the target area and the background.

[0075] The steps of performing connected component analysis on the enhanced image using a dual-tree topological filtering algorithm based on max-tree and min-tree are as follows:

[0076] (1) Decomposition of gray-level connected components: For echocardiogram I: Ω→{0,1,…,L max Max-trees construct hierarchical connected components from highest to lowest level. Each node N k This indicates that the grayscale value satisfies T k ≥T k-1 Connected components (Max-tree) or T k ≤T k-1 (Min-tree).

[0077] (2) Noise suppression mechanism of dual-tree collaboration: Max-tree dominates the high signal-to-noise ratio region, through threshold T valve Retain nodes that conform to anatomical structures to filter out high-grayscale areas (such as calcified valves):

[0078] N valve ={N k |T k >T valve Area(N) k )>θ min}(9)

[0079] Where, θ min Filter out small connected regions (e.g., areas < 10 pixels) generated by noise.

[0080] Min-tree suppresses low signal-to-noise ratio speckle noise, and the local noise variance is dynamically calculated using the following formula:

[0081]

[0082] Where, N noise is the interval for calculating the noise variance; p is the pixel index of the image; I(p) is the pixel value of the image; u noise It is the mean of local noise; It is the variance of local noise.

[0083] (3) Dual-tree topology constraints: Filtering constraints are defined based on anatomical features (such as valve aspect ratio and edge continuity).

[0084]

[0085] Where, N k It refers to any connected component; Length(N) k Width(N) is the length of the connected component; k ) refers to the width of the connected component; γ aspect The aspect ratio threshold (e.g., 2.5-3.5 for the mitral valve orifice); Hausdorff(N) k ) describes the maximum distance between the least matching points in two sets of points, used to quantify edge breaks; ∈ edge Retain(N) measures the degree of edge breakage. k ) is a dual-tree topology filtering function; this constraint can preserve connected components that conform to the valve anatomy.

[0086] As an optional embodiment of the present invention, the method further includes: acquiring a plurality of first target echocardiogram training data, each of the first target echocardiogram training data having a mitral valve orifice region marked; training a target model using the plurality of first target echocardiogram training data to obtain the mitral valve orifice segmentation model, wherein the target model integrates a residual network structure and a spatial attention module.

[0087] For example, a target model is trained using multiple first-target echocardiogram training data to obtain a mitral valve orifice segmentation model. The target model integrates a residual network structure and a spatial attention module. Specifically, the model structure of the mitral valve orifice segmentation model includes, but is not limited to, Vnet, Unet, etc. This embodiment takes Unet as an example. The fully convolutional neural network Unet has a U-shaped structure with encoder-decoder symmetry. The encoder is used to progressively downsample the image and extract features; the decoder is used to progressively upsample the image, recombine features, and finally generate an output of the same size as the input image. Unet's skip connections allow the decoder to directly connect to the corresponding layer of the encoder, so as to quickly fuse high-level features containing semantic information and low-level features containing spatial information. This allows Unet to retain rich details while effectively utilizing contextual information for more accurate pixel-level segmentation.

[0088] Obesity, lung emphysema, or chest wall deformities can cause blurred echocardiographic images due to sound wave attenuation, making automatic segmentation of the target region's edge contours difficult. In this embodiment, a residual network structure and a spatial attention module are added to the Unet network structure to improve the model's segmentation performance. For example... Figure 8As shown, the learning mechanism of the residual network structure is to introduce skip connections, which directly transmit the input signal to the output of the network module of that layer. During backpropagation, shallow gradients can bypass intermediate layers and be directly transmitted, effectively avoiding gradient vanishing. Figure 8 In this context, "Weight layer" represents the weight layer, "X" represents the input features of the residual block, "F(x)" represents the transformed features obtained after processing by the weight layer, and "F(x)+X" represents the output features of the residual block. For example... Figure 9 As shown, the feature x of the decoding layer l The dimension is F l ×Hx×Wx×Dx and the corresponding features g from the previous layer's encoding are used as input to the spatial attention module, which has dimensions Fg×Hg×Wg×Dg. The module then uses a self-attention mechanism to... l Feature fusion is performed with g, and the weight matrix is ​​normalized using a sigmoid function to increase the feature response of the target region of interest and suppress irrelevant regions. Specifically, the features from path A and path B are added element-wise and passed through path C to obtain the fused features, with dimensions Fint×Hg×Wg×Dg. These features are then passed through path D to the compression convolution module, where a 1×1×1 convolution ψ is used to compress the channels, resulting in compressed channels "Hg×Wg×Dg". This compressed channel is then passed through path E to the activation function module "Sigmoid(σ2)", where the spatial attention weights are obtained through the sigmoid activation function σ2. Finally, these weights are passed through path F to the spatial resampler module "Resampler", where Hg×Wg×Dg is adjusted to Hx×Wx×Dx to obtain the weight α. Based on the weight α, the weighted output features are obtained. The residual mechanism optimizes gradient propagation through skip connections, while the spatial attention module enhances feature selection through dynamic weight allocation. The combination of these two approaches improves the model's accuracy in segmenting the edges of target regions. For example... Figure 10 The diagram shows the segmentation results after incorporating the residual and spatial attention mechanisms into the network. The "green line" represents the ground truth labels, the "blue line" represents the prediction results of the Unet model, and the "red line" represents the prediction results of the network with the residual and spatial attention mechanisms. It can be seen that the network prediction results are more accurate after incorporating the residual and spatial attention mechanisms. The loss function Dice of the mitral valve orifice segmentation model Unet with integrated spatial attention module in this embodiment is as follows:

[0089]

[0090] Where P is the result predicted by the model, p * For real labels.

[0091] The assessment index detection method for evaluating the degree of mitral valve stenosis provided in this application embodiment can automatically detect the mitral valve orifice area, the distance between the long and short axes and their ratio on the parasternal short-axis-mitral valve horizontal section, and can automatically detect the opening distance at the narrowest point of the mitral valve orifice on the parasternal long-axis section. This provides a basis for the diagnosis of whether mitral valve stenosis is present, can reduce the variability problems that may be caused by human factors, and thus better assist doctors in making accurate diagnoses and improve doctors' efficiency.

[0092] This embodiment also provides an assessment index detection device for evaluating the degree of mitral valve stenosis. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. The term "module" used below can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0093] This embodiment provides a device for detecting assessment indicators to evaluate the degree of mitral valve stenosis, such as... Figure 11 As shown, it includes:

[0094] The first acquisition module 201 is used to acquire a first target echocardiogram, wherein the first target echocardiogram is an image of the parasternal short-axis mitral valve horizontal section.

[0095] The first determining module 202 is used to input the first target echocardiogram into the target detection model, and determine the first region of interest where the mitral valve orifice is located in the first target echocardiogram and the first target information corresponding to the first region of interest. The first target information includes the coordinates of the center point of the detection frame and the size information of the detection frame.

[0096] The segmentation module 203 is used to segment the first region of interest according to the first target information using a mitral valve orifice segmentation model to obtain a mask image of the mitral valve orifice.

[0097] The second determining module 204 is used to determine the area of ​​the mitral valve orifice region within the first region of interest, based on the number of pixels contained in the mask image of the mitral valve orifice. For detailed explanation, please refer to the relevant descriptions in the above method embodiments; they will not be repeated here.

[0098] In this embodiment, the assessment index detection device for evaluating the degree of mitral valve stenosis is presented in the form of a functional unit. Here, a unit refers to an ASIC circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0099] The assessment index detection device for evaluating the degree of mitral valve stenosis provided in this embodiment determines the region of interest (ROI) of the mitral valve orifice in the target echocardiogram using a target detection model. Based on the target information of the ROI, a mitral valve orifice segmentation model is used to obtain a mask image of the mitral valve orifice. The area of ​​the mitral valve orifice region is directly determined based on the number of pixels in the mask image. Compared with the method of determining the area of ​​the mitral valve orifice region by combining manual operation, this method improves the detection efficiency and accuracy of the mitral valve orifice region area parameter.

[0100] As an optional embodiment of the present invention, the device further includes: a second acquisition module, used to acquire pixels on the contour boundary of the mitral valve orifice region and the coordinates of the corresponding pixels; a first traversal module, used to traverse the distance between any two pixels and take the line connecting the two pixels with the largest distance as the major axis of the mitral valve orifice region; and a second traversal module, used to, according to the traversal operation, take the line connecting the two endpoints with the largest distance in the vertical direction as the minor axis of the mitral valve orifice region in the vertical direction of the major axis.

[0101] As an optional embodiment of the present invention, the device further includes: a third determining module, configured to determine the area of ​​the mitral valve orifice region corresponding to each of the first target echocardiograms when multiple first target echocardiograms are acquired; and a first constructing module, configured to construct area change characterization data corresponding to different first target echocardiograms based on each first target echocardiogram and the area of ​​the corresponding mitral valve orifice region.

[0102] As an optional embodiment of the present invention, the device further includes: a fourth determining module, configured to determine the mitral valve orifice region with the largest area in all the first target echocardiograms; and a first display module, configured to display the outline of the corresponding mitral valve orifice region, the major axis and the minor axis, and the ratio information of the major axis and the minor axis in the first target echocardiogram corresponding to the mitral valve orifice region with the largest area.

[0103] As an optional embodiment of the present invention, the device further includes: a third acquisition module, used to acquire a second target echocardiogram, wherein the second target echocardiogram is an image of the parasternal long axis section; and a fifth determination module, used to input the second target echocardiogram into a key point detection model to determine the second target information corresponding to the second region of interest where the mitral valve orifice is located in the second target echocardiogram, as well as the positions of the two endpoints of the narrowest part of the mitral valve orifice in the second region of interest and the corresponding endpoint spacing, wherein the second target information includes the coordinates of the center point of the detection frame and the size information of the detection frame.

[0104] As an optional embodiment of the present invention, the device further includes: a sixth determining module, configured to determine the endpoint spacing corresponding to each of the second target echocardiograms when multiple second target echocardiograms are acquired; and a second constructing module, configured to construct endpoint spacing variation characterization data corresponding to different second target echocardiograms based on each second target echocardiogram and the corresponding endpoint spacing.

[0105] As an optional embodiment of the present invention, the device further includes: a seventh determining module, configured to determine the maximum endpoint spacing corresponding to the narrowest point of the mitral valve orifice in all the second target echocardiograms; and a second display module, configured to display the corresponding endpoint position and the maximum endpoint spacing information in the second target echocardiogram corresponding to the maximum endpoint spacing.

[0106] As an optional embodiment of the present invention, the device further includes: a first processing module, used to perform contrast enhancement processing on the first target echocardiogram or the second target echocardiogram; and a second processing module, used to perform connected component analysis processing on the enhanced image using a preset topology filtering algorithm.

[0107] As an optional embodiment of the present invention, the device further includes: a fourth acquisition module, used to acquire multiple first target echocardiogram training data, each of the first target echocardiogram training data having a mitral valve orifice region marked; and a training module, used to train the target model using the multiple first target echocardiogram training data to obtain the mitral valve orifice segmentation model, wherein the target model integrates a residual network structure and a spatial attention module.

[0108] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0109] In this embodiment, the assessment index detection device for evaluating the degree of mitral valve stenosis is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0110] This invention also provides a computer device having the above-described features. Figure 11 The device shown is a test device for assessing the degree of mitral valve stenosis.

[0111] Please see Figure 12 , Figure 12This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 12 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 12 Take a processor 10 as an example.

[0112] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0113] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0114] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0115] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0116] The computer device also includes an input device 30 and an output device 40. The processor 10, memory 20, input device 30, and output device 40 can be connected via a bus or other means. Figure 12 Taking the example of a connection between China and Israel via a bus.

[0117] Input device 30 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 40 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.

[0118] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0119] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0120] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various edits and modifications without departing from the spirit and scope of the invention, all of which fall within the scope defined by the appended claims.

Claims

1. A method for detecting assessment indicators to evaluate the degree of mitral valve stenosis, characterized in that, include: Acquire a first target echocardiogram, which is a parasternal short-axis mitral valve level section image; The first target echocardiogram is input into the target detection model to determine the first region of interest where the mitral valve orifice is located in the first target echocardiogram and the first target information corresponding to the first region of interest. The first target information includes the coordinates of the center point of the detection box and the size information of the detection box. Based on the first target information, the first region of interest is segmented using the mitral valve orifice segmentation model to obtain a mask image of the mitral valve orifice. The area of ​​the mitral valve orifice region in the first region of interest is determined based on the number of pixels contained in the mask image of the mitral valve orifice.

2. The method according to claim 1, characterized in that, After determining the area of ​​the mitral valve orifice region within the first region of interest based on the number of pixels contained in the mask image of the mitral valve orifice, the method further includes: Obtain the pixel points and corresponding pixel coordinates on the contour boundary of the mitral valve orifice region; The distance between any two pixels is traversed, and the line connecting the two pixels with the largest distance is taken as the major axis of the mitral valve orifice region. In the direction perpendicular to the major axis, according to the traversal operation, the line connecting the two endpoints with the largest vertical distance is taken as the minor axis of the mitral valve orifice region.

3. The method according to claim 2, characterized in that, The method further includes: When multiple echocardiograms of the first target are acquired, the area of ​​the mitral valve orifice region corresponding to each echocardiogram of the first target is determined. Based on each first target echocardiogram and the area of ​​the corresponding mitral valve orifice region, area change characterization data corresponding to different first target echocardiograms are constructed.

4. The method according to claim 3, characterized in that, The method further includes: Identify the mitral valve orifice region with the largest area in all of the first target echocardiograms; In the first target echocardiogram corresponding to the largest mitral valve orifice region, the outline of the corresponding mitral valve orifice region, the major axis and minor axis, and the ratio information of the major axis and minor axis are displayed.

5. The method according to claim 1, characterized in that, The method further includes: Acquire a second target echocardiogram, which is an image of the parasternal long axis section; The second target echocardiogram is input into the key point detection model to determine the second target information corresponding to the second region of interest where the mitral valve orifice is located in the second target echocardiogram, as well as the positions of the two endpoints of the narrowest part of the mitral valve orifice in the second region of interest and the corresponding endpoint spacing. The second target information includes the coordinates of the center point of the detection frame and the size information of the detection frame.

6. The method according to claim 5, characterized in that, The method further includes: When multiple echocardiograms of the second target are acquired, the endpoint spacing corresponding to each echocardiogram of the second target is determined; Based on each second target echocardiogram and its corresponding endpoint spacing, endpoint spacing variation characterization data corresponding to different second target echocardiograms are constructed.

7. The method according to claim 6, characterized in that, The method further includes: Determine the maximum inter-endpoint distance corresponding to the narrowest point of the mitral valve orifice in all the second target echocardiograms; The corresponding endpoint positions and maximum endpoint spacing information are displayed in the second target echocardiogram corresponding to the maximum endpoint spacing.

8. The method according to claim 5, characterized in that, The method further includes: Contrast enhancement processing is performed on the first target echocardiogram or the second target echocardiogram; The enhanced image is processed by connected component analysis using a preset topology filtering algorithm.

9. The method according to claim 1, characterized in that, The method further includes: Multiple training data sets of first-target echocardiograms were acquired, each of which included a mitral valve orifice region: The target model is trained using the training data of the multiple first target echocardiograms to obtain the mitral valve orifice segmentation model. The target model integrates a residual network structure and a spatial attention module.

10. A device for detecting assessment indicators to evaluate the degree of mitral valve stenosis, characterized in that, include: The first acquisition module is used to acquire a first target echocardiogram, wherein the first target echocardiogram is an image of the parasternal short-axis mitral valve level section; The first determining module is used to input the first target echocardiogram into the target detection model, determine the first region of interest where the mitral valve orifice is located in the first target echocardiogram and the first target information corresponding to the first region of interest, wherein the first target information includes the coordinates of the center point of the detection box and the size information of the detection box; The segmentation module is used to segment the first region of interest based on the first target information using a mitral valve orifice segmentation model to obtain a mask image of the mitral valve orifice. The second determining module is used to determine the area of ​​the mitral valve orifice region in the first region of interest where the mitral valve orifice is located, based on the number of pixels contained in the mask image of the mitral valve orifice.

11. A computer device, characterized in that, include: The device includes a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the assessment index detection method for evaluating the degree of mitral valve stenosis as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the assessment index detection method for evaluating the degree of mitral valve stenosis as described in any one of claims 1-9.

Images