Method and system for quantitatively analyzing left atrial appendage occlusion effect based on ct image

Abstract

The application discloses a method for quantitatively analyzing left atrial appendage occlusion effect based on CT images, which comprises the following steps: 1, collecting a heart CTA image; 2, performing three-dimensional reconstruction to obtain a three-dimensional image of the heart, and obtaining a three-dimensional image of the left atrium by using a left atrium segmentation model; 3, obtaining a left atrial appendage occluder region by using a left atrial appendage occluder segmentation model; 4, constructing a three-dimensional model of the left atrial appendage occluder based on the segmented left atrial appendage occluder region, so as to obtain a three-dimensional image of the left atrial appendage occluder; 5, obtaining a three-dimensional plane of the left atrial appendage occluder, an occluder front view and a left atrial appendage opening plane; 6, extracting left atrial appendage occlusion postoperative evaluation parameters, wherein the parameters at least include: the size, position and classification of residual shunt, the position, shoulder position and height of the occluder, the endothelialization degree of the occluder and the position of instrument-related thrombus; and 7, evaluating the left atrial appendage occlusion effect according to the obtained left atrial appendage occlusion postoperative evaluation parameters.

Description

Technical Field

[0001] The present invention relates to the fields of image processing and machine learning, and in particular to a method and system for quantitatively analyzing left atrial appendage occlusion effects based on CT images. Background Art

[0002] Atrial fibrillation (AF) is the most common cardiac arrhythmia in middle-aged and elderly patients. It can lead to thrombosis in the left atrial appendage (LAA), which can subsequently dislodge and cause thromboembolic events such as ischemic stroke. AF patients often take anticoagulants to prevent thrombosis and stroke. However, anticoagulant therapy carries a certain risk of bleeding, and a significant number of AF patients have contraindications to oral anticoagulation. Left atrial appendage occlusion (LAAO) is an interventional procedure for treating AF. It involves implanting an occluder device in the LAA to prevent thrombosis and dislodgement, thereby reducing the risk of embolic events in AF patients. However, after implantation, the occluder requires complete endothelialization, which is the process of forming an endothelial cell layer on the occluder surface. This endothelial cell layer reduces direct contact between blood and the occluder material, thereby reducing the risk of thrombosis and other complications associated with the occluder. Therefore, accurately assessing the occurrence and severity of complications after LAAO is crucial to guide clinical decision-making and improve surgical strategies.

[0003] At present, the quantitative evaluation of complications after left atrial appendage closure surgery, such as residual shunt, incomplete endothelialization of the occluder surface, occluder displacement, and occluder-related thrombosis, has become a research hotspot. More and more researchers and doctors are committed to developing new methods and technologies to achieve accurate quantitative evaluation. However, the methods for evaluating complications after left atrial appendage closure surgery in the prior art mainly rely on the doctor's experience and subjective judgment. This method has certain subjectivity and inconsistency, and has high requirements for the doctor's professional knowledge and experience. Therefore, how to quantitatively evaluate the effect of left atrial appendage surgery to improve the accuracy and repeatability of the evaluation is a problem that needs to be solved urgently.

[0004] In the fields of image processing and computer vision, inventors have explored methods for quantitatively assessing endothelialization on occluder surfaces using CT images, as described in the article "A novel pragmatic paradigm to evaluate device endothelialization: lessons learned from a dual-watchman case." Some of these studies have focused on texture analysis and feature extraction of the occluder surface. For example, these methods assess the degree of endothelialization by calculating texture features of the occluder surface, such as gray-level co-occurrence matrices and wavelet transforms. While these methods can provide some quantitative indicators, they still present certain subjectivity and limitations.

[0005] The rapid development of medical image processing and machine learning technologies in recent years has provided new opportunities to address this problem. Medical image processing technologies can preprocess and enhance CT images to improve image quality and clarity. Feature extraction technologies can extract target-related features from images, such as atrial shape, size, occluder position, and device surface structure. Machine learning technologies can use these features to train models and automatically identify and analyze complications after left atrial appendage surgery. Summary of the Invention

[0006] The present invention provides a method for quantitatively analyzing the left atrial appendage occlusion effect based on CT images, so as to obtain relevant surface parameters of the occluder efficiently, quickly and accurately, and provide a basis for the automatic identification and analysis of complications after left atrial appendage surgery.

[0007] To this end, the present invention provides the following technical solutions:

[0008] A method for quantitatively analyzing left atrial appendage occlusion effects based on CT images, the method comprising:

[0009] Step 1: Acquire CTA images of the patient's heart after left atrial appendage occlusion surgery;

[0010] Step 2: performing three-dimensional reconstruction on the cardiac CTA image sequence to obtain a three-dimensional image of the heart, and obtaining a three-dimensional image of the left atrium using a left atrium segmentation model;

[0011] Step 3: Segment the LAA occluder area in the CTA image using the LAA occluder segmentation model;

[0012] Step 4: Based on the segmented left atrial appendage occluder area, a three-dimensional model of the left atrial appendage occluder is constructed, and a three-dimensional image of the left atrial appendage occluder is obtained through three-dimensional reconstruction;

[0013] Step 5: Based on the three-dimensional image of the left atrial appendage occluder, obtain the three-dimensional plane of the left atrial appendage occluder, the front view of the occluder, and the plane of the left atrial appendage opening;

[0014] Step 6: Extracting evaluation parameters after left atrial appendage occlusion surgery, the parameters including at least: the size, location, and grade of residual shunt, the location, shoulder position, and height of the occluder, the degree of endothelialization of the occluder, and the location of device-related thrombus;

[0015] Step 7: Evaluate the left atrial appendage occlusion effect based on the obtained evaluation parameters after left atrial appendage occlusion surgery.

[0016] Furthermore, the steps of constructing the left atrial appendage occluder segmentation model in step 3 are as follows:

[0017] Step 3.1: Acquire CTA images of patients after left atrial appendage occlusion surgery, segment the left atrial appendage occluder, and obtain a training dataset and corresponding masks;

[0018] Step 3.2: Based on the training data set and mask obtained in step 3.1, a U-net algorithm model is trained, and the trained model is used as the left atrial appendage occluder segmentation model.

[0019] Furthermore, the specific steps of step 5 include:

[0020] Step 5.1: With the center point of the occluder rivet in the left atrial appendage occluder image as the starting point P1, the tip of the occluder as the end point P2, and the path direction P1→P2, the connecting line from P1 to P2 is used as the initial axis of the occluder. The axis of the occluder is continuously corrected through iterative calculation to finally obtain the actual axis of the occluder;

[0021] Step 5.2: The tangent line passing through the center point of the occluder rivet and perpendicular to the initial axis is used as the tangent line of the left atrial surface of the occluder. The tangent line of the left atrial surface of the occluder is corrected during the iterative calculation of the occluder axis, and the actual tangent line of the left atrial surface of the occluder is finally obtained.

[0022] Step 5.3: Fix the obtained tangent of the left atrial surface of the occluder and the actual axis of the occluder at 90 degrees, thereby obtaining the three-dimensional planes of the left atrial appendage occluder: axial section, sagittal plane, and coronal plane;

[0023] Step 5.4: Fix the obtained axial, coronal, and sagittal planes at 90° angles to each other, and adjust the window width and window position of the axial section until the CT image layer of the occluder is displayed, thereby obtaining a frontal view of the left atrial appendage occluder;

[0024] Step 5.5: In the 3D reconstruction, use the orthogonal plane method to place the crosshairs at the level of the proximal circumflex artery.

[0025] Step 5.6: Rotate the crosshairs of the coronal and sagittal planes in sequence, adjust the positions of the crosshairs of the coronal and sagittal planes, and place the crosshairs of the coronal and sagittal planes at the crux and the pulmonary vein ridge, respectively, to determine the position of the left atrial appendage opening, and take the plane corresponding to the crosshairs placed at the crux and the pulmonary vein ridge as the left atrial appendage opening plane.

[0026] Further, step 6 includes the following specific steps:

[0027] Step 6.1: Determine the position, size, and grading of residual shunt;

[0028] Step 6.2: Determine the position of the occluder, including the height of the shoulder and the position of the shoulder;

[0029] Step 6.3: Determine the contrast attenuation thickness HAT of the occluder and the extent of endothelialization;

[0030] Step 6.4: Evaluate HAT, if the HAT thickness is greater than 3 mm, observe whether the HAT is regular, if not, consider it as DRT, and represent the current DRT position by clock method.

[0031] Further, the specific steps of step 6.1 are:

[0032] Step 6.1.1: Rotate the crosshairs of the coronal and sagittal planes, respectively, and in the sagittal or coronal plane, take the channel of contrast agent from the left atrium into the left atrial appendage as the residual shunt;

[0033] Step 6.1.2: Determine the control line at the residual shunt based on the axial section to obtain the cross section of the residual shunt;

[0034] Step 6.1.3: Determine the minimum diameter and area of the residual shunt based on the cross section;

[0035] Step 6.1.4: Grade the residual shunt according to the minimum diameter of the residual shunt, if the diameter is 0-1 mm, it is a trace amount, 1 mm-3 mm is a small amount, 3 mm-5 mm is a moderate amount, and more than 5 mm is a large amount.

[0036] Further, the specific steps of step 6.2 are:

[0037] Step 6.2.1: Rotate the crosshairs of the axial section until the position of the occluder shoulder is determined in the sagittal or coronal plane;

[0038] Step 6.2.2: Take the left atrial appendage opening plane as the starting point and the highest position of the occluder shoulder as the ending point, calculate the Euclidean distance between the two points, and take the distance as the height of the occluder shoulder.

[0039] Further, the specific steps of step 6.3 are:

[0040] Step 6.3.1: Detect the intersection areas of all occluder beams on the three-dimensional image of the left atrial appendage occluder and select the intersection area with the largest volume as the center point of the occluder;

[0041] Step 6.3.2: Start searching near the center of each beam. When the angle between the vector A from the current search point to the next search point and the vector B from the previous search point to the current search point is the largest, the current search point is considered the beam inflection point.

[0042] Step 6.3.3: Calculate the planes that can be constructed using the beam inflection points determined in Step 6.3.2. Select the plane with the largest number of inflection points and use the normal vector of that plane as the vector from the center point to the end of the occluder. If all planes constructed using the beam inflection points have the same number of inflection points, select the plane closest to the center point of the occluder and use the normal vector of that plane as the vector from the center point to the end of the occluder.

[0043] Step 6.3.4: Using the vector from the center point to the end of the occluder determined in Step 6.3.3 as the plane normal, calculate the area of ​​the quasi-circular shape enclosed by the occluder beam on each plane. The plane with the largest area is the maximum cross-section of the occluder.

[0044] Step 6.3.5: Segment the endothelialized area on the surface of the left atrial appendage occluder using the endothelialized area segmentation method;

[0045] Step 6.3.6: Plot the endothelial thickness at each pixel in the endothelialized area using different colors;

[0046] Step 6.3.7: Calculate the endothelialization ratio using the following formula and use the endothelialization ratio as the degree of endothelialization after left atrial appendage occlusion:

[0047] Endothelialization ratio = endothelialization coverage area of ​​the left atrial surface of the occluder / area of ​​the left atrial surface of the occluder 100%.

[0048] Furthermore, the endothelialization area segmentation method in step 6.3.5 is a grayscale threshold segmentation algorithm, and its specific steps are as follows:

[0049] Step 6.3.5.1: Along the two normal vectors of the maximum cross-section of the occluder determined in step 6.3.4, starting from the maximum cross-section of the occluder, search for pixels within the maximum cross-section range in each image layer that meet the endothelialization CT value range;

[0050] Step 6.3.5.2: When a pixel point belonging to the left atrial region is found, the search is stopped, and the found pixel point is used as a pixel point of the endothelialized region, and the region formed by combining these pixel points is used as the endothelialized region.

[0051] Further, the endothelialization area segmentation method in step 6.3.5 is a U-net segmentation algorithm, and the specific steps are as follows:

[0052] Step 1: Based on the CT image data pre-labeled with endothelialization area and audited by doctors in the professional heart discipline field, train an endothelialization area segmentation model with the U-net algorithm;

[0053] Step 2: Obtain the CT image of the current patient, and segment the endothelialization area in the image by using the trained endothelialization area segmentation model.

[0054] A system for extracting left atrial appendage data parameters based on CT images, which is realized based on the above method, and the system comprises:

[0055] An image sequence acquisition module is configured to acquire CTA images of a heart of a patient after left atrial appendage occlusion surgery;

[0056] An image reconstruction module is configured to perform three-dimensional reconstruction on the CTA image sequence of the heart to obtain a three-dimensional image of the heart;

[0057] A segmentation module comprises,

[0058] A left atrium segmentation unit is configured to segment a three-dimensional image of the left atrium by using a left atrium segmentation model;

[0059] A left atrial appendage occluder segmentation unit is configured to segment a left atrial appendage occluder region in the CTA image by using a left atrial appendage occluder segmentation model;

[0060] A three-dimensional reconstruction module is configured to construct a three-dimensional model of the left atrial appendage occluder based on the segmented left atrial appendage occluder region, and obtain a three-dimensional image of the left atrial appendage occluder by three-dimensional reconstruction;

[0061] A three-dimensional plane acquisition module is configured to obtain a three-dimensional plane of the left atrial appendage occluder, a front view of the occluder, and a left atrial appendage opening plane based on the three-dimensional image of the left atrial appendage occluder;

[0062] An occluder parameter extraction module is configured to extract post-occlusion evaluation parameters of the left atrial appendage, wherein the parameters at least include: size, position and classification of residual shunt, position, shoulder position and height of the occluder, endothelialization degree of the occluder, and position of instrument-related thrombus;

[0063] An evaluation module is configured to evaluate the occlusion effect of the left atrial appendage according to the obtained post-occlusion evaluation parameters of the left atrial appendage.

[0064] Compared with the prior art, the present application has the following beneficial effects:

[0065] First, the present invention intelligently extracts images of the left atrium and left atrial appendage occluder based on CT images, quantitatively analyzes the effect of left atrial appendage occlusion surgery, thereby improving the accuracy and repeatability of left atrial appendage occlusion surgery effect evaluation and providing a basis for postoperative anticoagulation schemes;

[0066] Second, the present invention proposes the POET evaluation principle, which enables clinicians to quickly and accurately evaluate the postoperative effects of left atrial appendage occlusion on patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figures 1a-1b is a diagram showing the location of the residual shunt; Figure 1a For the standard left superior pulmonary vein and mitral valve, aorta; Figure 1b The middle PDL position is 8:00-9:00;

[0068] Figure 2 This is a diagram to evaluate the degree of endothelialization completion;

[0069] Figures 3a-3b is a location map of device-related thrombus; Figure 3b The DRT position is 1:00-3:00

[0070] Figures 4a-4d They are the three-dimensional reconstruction of the left atrium and occluder, the axial section of the left atrial appendage occluder, the sagittal section of the left atrial appendage occluder, and the coronal section of the left atrial appendage occluder;

[0071] Figures 5a-5d They are the three-dimensional reconstruction of the left atrium and occluder, the front view of the left atrial appendage occluder, the sagittal view of the left atrial appendage occluder, and the coronal view of the left atrial appendage occluder;

[0072] Figure 6 For the occluder, the shoulder measurement diagram;

[0073] Figures 7a-7d is a schematic diagram of DRT; Figures 7a-7b The base is irregular. Figures 7c-7d For the base rule;

[0074] Figure 8 Schematic diagram of the ray casting algorithm;

[0075] Figure 9 Build results for the three-dimensional image of the left atrial appendage occluder;

[0076] Figure 10 Schematic diagram of the center point and beam point;

[0077] Figure 11 The results are stained according to the thickness of endothelialization. DETAILED DESCRIPTION

[0078] The present invention applies CT images to the assessment of complications after left atrial appendage surgery, and provides a method for quantitatively analyzing complications after left atrial appendage occlusion surgery based on CT images. A three-dimensional image of the heart is obtained by three-dimensionally reconstructing a coronary CTA image sequence, and then a three-dimensional image of the left atrium and left atrial appendage occluder is obtained using the three-dimensional image of the heart and a pre-established left atrial segmentation model. Then, relevant parameters after left atrial appendage occlusion surgery are extracted from the three-dimensional image. The relevant parameters include the size, position and grade of residual shunt, the position of the occluder, the exposed shoulder position and the adjacent relationship with the surrounding structures, the endothelialization degree of the occluder, and relevant parameters of the position, size and grade of device-related thrombus. The obtained parameters can be used to quickly evaluate the left atrial appendage occlusion effect of the patient.

[0079] 1. Left atrium and left atrial appendage occluder segmentation model

[0080] The present invention adopts the left atrial segmentation model disclosed in the inventor's previous Chinese patent number "CN112244883A" "Method and system for extracting left atrial appendage data parameters based on CT images" to extract the left atrial three-dimensional image from the three-dimensional coronary artery image. The establishment and training process of the model has been disclosed in the document and will not be repeated in the present invention.

[0081] In addition, the present invention also constructs a left atrial appendage occluder segmentation model for extracting a three-dimensional image of the left atrial appendage occluder from a three-dimensional coronary artery image; the steps for constructing the model are as follows: (1) collecting CTA images of patients after left atrial appendage occlusion surgery, manually performing left atrial appendage occluder segmentation, and obtaining a training data set and a corresponding mask required for training the U-net algorithm model; (2) based on the obtained training data set and the corresponding mask, using the U-net algorithm framework, training a left atrial appendage occluder segmentation model.

[0082] Using the trained left atrium segmentation model and the left atrial appendage occluder segmentation model, a 3D left atrium image and a 3D left atrial appendage occluder image can be extracted from the 3D coronary artery image. The specific steps for constructing a 3D left atrial appendage occluder image using the left atrial appendage occluder segmentation model are as follows:

[0083] (1) Based on the CTA images of patients after left atrial appendage occlusion surgery, the left atrial appendage occluder area in the CTA images was segmented using the left atrial appendage occluder segmentation model established above;

[0084] (2) Based on the segmented LAA occluder area, the SSD algorithm was used to construct a three-dimensional model of the occluder;

[0085] (3) Use computer graphics algorithms to construct a three-dimensional image of the left atrial appendage occluder, such as Figure 9 As shown in the figure, the blue area is the constructed three-dimensional model of the occluder.

[0086] 2. POET Principle

[0087] Based on the LAAO principle, the present invention proposes a new POET principle for evaluating the effectiveness of left atrial appendage occlusion: P, namely residual shunt (Peri-divice leak, PDL), includes the size, location and grade of PDL; O, namely position (pOsition), includes the shoulder position (placement depth) and shoulder height of the occlusion; E, namely endothelialization, includes the degree of endothelialization; T, namely device-related thrombosis (DRT), includes the size, base size and contrast agent attenuation thickness (HAT) grade of DRT.

[0088] 3. The method proposed by the present invention

[0089] Based on the above-mentioned left atrial segmentation model and POET principle, the present invention proposes a method for quantitatively analyzing the effect of left atrial appendage occlusion based on CT images, comprising the following steps:

[0090] Step 1: Acquire coronary CTA image sequences;

[0091] Step 2: Perform 3D reconstruction on the coronary CTA image sequence to obtain a 3D image of the left atrium and adjacent structures;

[0092] Step 3: Input the three-dimensional image into the pre-trained left atrium segmentation model and left atrial appendage occluder segmentation model to obtain a left atrium three-dimensional image and a left atrial appendage occluder three-dimensional image model;

[0093] Through three-dimensional reconstruction, the axial section, coronary plane, and sagittal plane of the left atrial appendage occluder are obtained, and then the front view of the left atrial appendage occluder and the left atrial appendage opening plane are obtained using the orthogonal method;

[0094] The present invention specifically adopts the following three-dimensional reconstruction method:

[0095] Three-dimensional reconstruction is volume reconstruction. Volume reconstruction directly studies the changes in light when it passes through the three-dimensional volume data field to obtain the final rendering result, so volume reconstruction is also called direct volume rendering. Volume rendering methods are mainly divided into two categories: image space-ordered volume rendering methods and object space-ordered volume rendering methods: (1) The image space-ordered volume rendering method starts from each pixel on the screen and emits a ray according to the viewpoint direction. This ray passes through the three-dimensional data field and performs equidistant sampling along the ray to calculate the opacity and color value of the object at the sampling point. The color and opacity of the sampling points on a ray can be synthesized in two orders, from front to back or from back to front (MaxIP or MinIP takes the maximum or minimum value on the ray), thereby calculating the color value of the pixel on the screen. This method simulates the process of light passing through the object from the opposite direction. (2) The volume rendering method based on the object space first calculates the color and opacity of each data point according to the function value of the point, and then projects each data point onto the image plane according to the given viewing plane and observation direction. The opacity and color are calculated according to the order of occlusion of the data points in space, and finally the image is obtained.

[0096] 1) Ray casting

[0097] The ray casting method casts a ray from each pixel in the image plane to the data field, samples the ray or integrates along line segments to calculate brightness and opacity, and then synthesizes the image in the order of sampling to produce the resulting image. The ray casting method is an image-space-ordered method. It simulates the entire process of light passing through an object in the reverse direction and ultimately calculates the color of the ray after it reaches the data field.

[0098] like Figure 8 As shown, the ray casting algorithm mainly has the following processes (see the drawing process above):

[0099] ① Data preprocessing: including adjustment of sampling grid, data contrast enhancement, etc.;

[0100] ②Data classification and lighting effect calculation: Establish the mapping between field value to color value and opacity respectively, and use the central difference method to calculate the normal vector and calculate the lighting effect;

[0101] ③ Ray casting: From each pixel on the screen, cast a ray along the viewing direction, through the data field, sample each ray, and interpolate to calculate the color value and opacity;

[0102] ④ Synthesis and Rendering: On each ray, the color values ​​of each sampling point are synthesized in order to obtain the pixel color value, and the pixel is displayed. If the Maximum Intensity Projection (MaxIP) or Minimum Intensity Projection (MinIP) method is used, the maximum or minimum value of all sampling points on this ray is directly taken as the final pixel color value.

[0103] 2) Projection method

[0104] The projective volume rendering method is based on the correlation between regions and volumes in the field. It projects voxels onto the image plane, calculates each voxel's contribution to the pixel, and synthesizes the effects of each voxel according to the order of front and back occlusion. This method essentially calculates the effect of light emitted by each voxel in the data field on each pixel on the image plane, and ultimately calculates the image. The main steps of projective volume rendering are as follows:

[0105] ① Voxel traversal: determine the front and back occlusion order of voxels in the data field, and traverse the voxels from front to back or from back to front;

[0106] ② Voxel decomposition: Each voxel is decomposed into a group of sub-voxels, and the projection contours of the sub-voxels must not overlap on the observation plane;

[0107] ③ Projection and synthesis: The sub-voxel is projected onto the image plane to obtain a projected polygon; the values ​​of the projected polygon vertices are calculated, and the brightness contribution of the projected polygon to the covered pixel is calculated by scan conversion, and then synthesized with the original pixel value;

[0108] Step 4: extracting parameters related to the left atrial appendage occluder from the three-dimensional image of the left atrial appendage occluder, the front view of the occluder, and the plane of the left atrial appendage opening, wherein the parameters related to the left atrial appendage occluder include at least the size, location, and grade of residual shunt, the location, shoulder position, and height of the occluder, the degree of endothelialization of the occluder, and the location, size, and grade of device-related thrombus;

[0109] Step 5: The clinician performs a postoperative evaluation of the left atrial appendage occlusion device using the left atrial appendage occlusion device-related parameters obtained from step 4.

[0110] During left atrial appendage occlusion surgery, the effectiveness of the occluder needs to be evaluated postoperatively. The correct extraction of left atrial appendage occluder parameters can provide reference for subsequent clinical evaluation and scientific research analysis.

[0111] Furthermore, the specific steps of obtaining the left atrial appendage occluder three-dimensional plane, the occluder front view and the left atrial appendage opening plane include:

[0112] Obtain the three-dimensional plane of the left atrial appendage occluder; specifically including:

[0113] (1) Obtain the left atrial appendage occluder anatomical image by using the left atrial appendage segmentation model, as shown in Figure 4a , the occluder axial direction is the direction of the umbrella handle, and the occluder rivet is the umbrella tip;

[0114] (2) Based on the anatomical image, taking the center point (red dot in Figure 5b ) of the occluder rivet in the left atrial appendage occluder image as the starting point P1, and the tip of the occluder as the end point P2, the path direction is P1→P2, and the connecting line from P1 to P2 is taken as the initial axial direction of the occluder (blue line in Figure 5c and Figure 5d ), the axial direction of the occluder is continuously corrected through iterative calculation, and the actual axial direction of the occluder is finally obtained; at the same time, after obtaining the initial axial direction of the occluder, the tangent line passing through the center point of the occluder rivet and perpendicular to the initial axial direction is taken as the tangent line of the left atrial surface of the occluder, and the tangent line of the left atrial surface of the occluder is corrected at the same time as the iterative calculation of the occluder axial direction, and the actual tangent line of the left atrial surface of the occluder is finally obtained;

[0115] (3) The actual tangent line of the left atrial surface of the occluder and the actual axial direction of the occluder are fixed at 90°, so as to obtain the three-dimensional plane of the left atrial appendage occluder (step), as shown in Figures 4b-4d , the yellow line (yellow arrow) in Figure 4c and Figure 4d represents the axial cross-section reference line, Figure 4b and Figure 4c the blue line (blue arrow) represents the coronal reference line, Figure 4b and Figure 4d the green line (green arrow) represents the sagittal reference line.

[0116] (4) Take the left atrial appendage occluder rivet as the starting point, and take the plane perpendicular to the left atrial surface of the occluder at a predetermined interval along the starting point to obtain multiple cross sections. Then determine the left atrial appendage occluder profile and the center of mass point of the occluder for each cross section in turn, and connect the center of mass points on adjacent cross sections to obtain N left atrial appendage occluder CT values. According to the CT value, the gray level histogram of the internal organ is generated; then the gray level distribution Gaussian probability curve is calculated through the gray level histogram, and the profile of the left atrial appendage occluder on each cross section is determined by using the obtained gray level distribution Gaussian probability curve.

[0117] Obtain the front view of the left atrial appendage occluder; specifically including:

[0118] Fix the three planes of the axial cross section, the coronal plane and the sagittal plane at an angle of 90°, and adjust the window width and window level of the axial cross section until the CT image layer of the occluder can be displayed, so as to obtain the front view of the left atrial appendage occluder;

[0119] Through three-dimensional reconstruction, a VR image of the left atrial appendage occluder can also be obtained. The VR image can intuitively show the relationship between the occluder and the surrounding adjacent structures, and can be used to obtain parameters such as PDL, DRT, and the position, size, and height of the occluder's exposed shoulder.

[0120] Obtain the left atrial appendage opening plane; specifically including:

[0121] (1) In the three-dimensional reconstruction of multiple planes, use the orthogonal plane method to place the crosshairs at the horizontal position of the proximal circumflex branch; (2) Rotate the crosshairs of the coronal and sagittal planes in turn, adjust the positions of the crosshairs of the coronal and sagittal planes, and place the crosshairs of the coronal and sagittal planes at the circumflex branch and the pulmonary vein crest respectively to determine the position of the left atrial appendage opening. The corresponding axial section can show the left atrial appendage opening. At this time, the plane corresponding to the crosshairs placed at the circumflex branch and the pulmonary vein crest is the left atrial appendage opening plane, as shown in Figure 2. Figure 6 shown.

[0122] Furthermore, step 4 of extracting the surface parameters of the occluder includes:

[0123] Step 4.1: Determine the location, size, and grade of the residual shunt (PDL);

[0124] Specifically, (1) the crosshairs of the coronal and sagittal planes are rotated respectively, and the channel of the contrast agent from the left atrium to the left atrial appendage in the sagittal or coronal plane is regarded as the residual shunt; (2) the control line of the residual shunt is determined based on the axial section to obtain the cross section of the residual shunt; (3) the minimum diameter and area of ​​the residual shunt are determined based on the cross section;

[0125] When determining the diameter, multiple starting points and end points are determined on the cross section of the residual shunt. The distance between the two points is obtained by calculating the Euclidean distance between the two points, which is used as the diameter of the residual shunt. The minimum diameter and maximum diameter of the residual shunt are obtained by comparison.

[0126] When determining the area, the ROI region was first determined on the axial section of the residual shunt using the ROI regional grid method, and then the area of ​​the ROI region was calculated using the pixel counting method;

[0127] After determining the minimum diameter and area of ​​the residual shunt, the minimum diameter of the residual shunt (PDL) is used for grading according to existing clinical standards. Generally, it is graded according to three cutoff points: 1mm, 3mm, and 5mm. Specifically, 0-1: trace; 1-3: small; 3-5: moderate; >5: large.

[0128] Then, in the three-dimensional reconstructed multi-plane, the axial section of the occluder was displayed in 3D, and the coronal and sagittal crosshairs were adjusted respectively. The axial section plane was adjusted to the mitral valve annulus below, the left upper pulmonary vein below, and the aorta in front. The occluder rivet was used as the clock point, with the left upper pulmonary vein direction at 12 o'clock, the mitral valve annulus direction at 6 o'clock, the aorta direction at 3 o'clock, and the posterior side at 9 o'clock, so that the position of the residual shunt was represented by the "nn" point.

[0129] Among them, the "nn" point is a clock notation. Since the PDL refers to the residual channel between the left atrium and the left atrial appendage, it is a three-dimensional structure, so its cross-section (i.e., the image displayed on the axial section) is a range. For example, the range of residual shunt can be 7 o'clock to 9 o'clock, such as Figure 1b As shown in the figure, the white dotted line is the PDL, the yellow line is the 8 o'clock position, and the remaining blue lines are the 9 o'clock, 6 o'clock, 3 o'clock, and 12 o'clock positions respectively;

[0130] Step 4.2: Extract the occluder position (pOsition), including the shoulder-exposed height and shoulder-exposed position;

[0131] Specifically, (1) in the 3D reconstructed multi-plane, the crosshairs of the axial section are rotated until the position of the occluder shoulder is determined in the sagittal or coronal plane; (2) with the left atrial appendage opening plane obtained in step 3.3 as the starting point and the highest position of the occluder shoulder as the end point, the Euclidean distance between the two is calculated and the distance is used as the height of the occluder shoulder, as follows: Figure 6 As shown in the figure, the red line is the left atrial appendage opening plane, the blue line is the position of the occluder shoulder, and the distance between the two lines is the height of the occluder shoulder;

[0132] In the three-dimensional reconstruction of the multi-plane, the axial section of the occluder is displayed in 3D, and the coronal and sagittal crosshairs are adjusted respectively. The axial section plane is adjusted to the mitral valve ring below, the left upper pulmonary vein below, and the aorta in front. The occluder rivet is used as the clock point, the left upper pulmonary vein direction is 12 o'clock, the mitral valve ring direction is 6 o'clock, the aorta direction is 3 o'clock, and the back side is 9 o'clock. The position where the occluder is exposed is represented by the "nn" point.

[0133] Step 4.3: Extract the endothelialization ratio of the occluder;

[0134] (1) Based on the three-dimensional image of the left atrial appendage occluder, the center point of the occluder is determined on the three-dimensional image of the left atrial appendage occluder, and a vector is drawn between the center point of the occluder and the end of the occluder;

[0135] The specific steps include:

[0136] ① Determine the center point of the occluder: Detect the intersection area of ​​all occluder beams on the 3D image of the left atrial appendage occluder and select the intersection area with the largest volume as the center point of the occluder. If the intersection area is difficult to detect or inaccurate due to compression and deformation after implantation, the doctor will manually determine the center point.

[0137] ② Determine the inflection point of each occluder beam. Start searching from the position close to the center point of each beam. When the angle between the vector A formed by the current search point to the next search point and the vector B formed by the previous search point to the current search point is the largest, the current search point is taken as the inflection point of the beam, such as Figure 10 As shown in the figure, the red point is the center point of the occluder, and the blue point is the inflection point of the beam;

[0138] ③ Calculate the planes that can be formed by the inflection points of the beams determined in the previous step, select the plane with the largest number of inflection points, and use the normal vector of this plane as the vector from the center point to the end of the occluder; if the number of inflection points in each plane that can be formed by the inflection points of the beams is the same, select the plane closest to the center point of the occluder, and use the normal vector of this plane as the vector from the center point to the end of the occluder;

[0139] Likewise, the vector from the center point to the end of the occluder can also be determined manually.

[0140] (2) Determine the maximum cross-section of the occluder

[0141] Using the vector from the center point to the end of the occluder as the normal vector of the plane, calculate the area of ​​the quasi-circular shape enclosed by the occluder beam on each plane. The plane with the largest area is the maximum cross-section of the occluder. The normal vector of the maximum cross-section of the occluder is also the vector from the center point to the end of the occluder.

[0142] (3) Segmentation of the endothelialized area on the surface of the left atrial appendage occluder

[0143] Specific segmentation methods include grayscale threshold segmentation and U-net algorithm segmentation;

[0144] ① Grayscale threshold segmentation

[0145] Step 1: Along the two normal vectors of the occluder's maximum cross-section, starting from the maximum cross-section of the occluder, search for pixels within the maximum cross-section of each image slice that meet the endothelialization CT value (pixel value) range. The range of endothelialization CT values ​​is generally determined by cardiologists. In clinical practice, the CT value of normal myocardium is less than 100HU, or the CT value of HAT / left atrial CT value range is 0.10-0.82.

[0146] Step 2: When a pixel point belonging to the left atrium area is found, the search is stopped, and the found pixel point is used as the pixel point of the endothelialization area, and the area formed by combining these pixel points is used as the endothelialization area.

[0147] ②U-net algorithm segmentation of endothelialization area

[0148] Step 1: Based on CCTA image data with pre-labeled endothelialization areas and reviewed by cardiology professionals, a U-net algorithm is used to train an endothelialization area segmentation model.

[0149] Step 2: Obtain the CCTA image of the current patient and use the trained endothelial region segmentation model to segment the endothelial region in the image.

[0150] (4) Calculation of endothelial thickness

[0151] On the maximum cross-section of the occluder, the endothelialization thickness at each pixel is plotted with different colors. For example, when the HAT thickness is 0≤HAT<1mm, it is represented by purple, and when the HAT thickness is 1≤HAT<2mm, it is represented by blue. Figure 11 As shown;

[0152] (5) Calculation of the degree of endothelialization

[0153] The complete endothelialization ratio (En Ratio) was calculated by the following formula and used as the degree of complete endothelialization after left atrial appendage occlusion:

[0154] Endothelialization ratio = endothelialization coverage area of ​​the left atrial surface of the occluder / area of ​​the left atrial surface of the occluder 100%;

[0155] The main steps for calculating the endothelialization ratio are shown in Table 1.

[0156]

[0157] Step 4.4: Extract the location and size of the device-related thrombus;

[0158] Specifically, the thickness of the HAT calculated in step 4.3 is judged. If the HAT thickness is greater than 3 mm, observe whether the base of the HAT is regular. If not (i.e., the diameter is inconsistent when starting from the occluder surface and / or continuing with the left atrium), it is considered as DRT (device-related thrombosis). Figures 7a-7d As shown, Figure 7a 、 Figure 7b For irregular, Figure 7c 、 Figure 7dFor specific judgment methods, please refer to the literature "Cardiac Computed Tomography Following Watchman FLXImplantation" and the literature "Detection of Device-Related Thrombosis Following Left Atrial Appendage Occlusion".

[0159] In the three-dimensional reconstruction of the multi-plane, the axial section of the occluder is displayed in 3D, and the coronal and sagittal crosshairs are adjusted respectively. The axial section plane is adjusted to the mitral valve annulus below, the left upper pulmonary vein below, and the aorta in front. The occluder rivet is used as the clock point, the left upper pulmonary vein direction is 12 o'clock, the mitral valve annulus direction is 6 o'clock, the aorta direction is 3 o'clock, and the posterior side is 9 o'clock. The position of the device-related thrombus is represented by the "nn" point.

[0160] Accordingly, an embodiment of the present invention also provides a system for quantitatively analyzing complications after left atrial appendage occlusion surgery based on CT images.

[0161] In this embodiment, the system includes the following modules:

[0162] An image sequence acquisition module, used for acquiring coronary CTA image sequences;

[0163] An image reconstruction module, configured to perform three-dimensional reconstruction on the coronary CTA image sequence to obtain a three-dimensional coronary image;

[0164] a segmentation module, configured to obtain a three-dimensional image of the left atrium and the left atrial appendage occluder using the three-dimensional coronary artery image and a pre-established left atrial segmentation model;

[0165] An occluder parameter extraction module is used to extract occluder surface parameters from a three-dimensional image of the left atrial appendage occluder, wherein the parameters include at least: the size, location, and grade of residual shunt, the location, shoulder position, and shoulder height of the occluder, the degree of endothelialization of the occluder, and the location and size of device-related thrombus;

[0166] The evaluation module is used to evaluate the left atrial appendage occlusion effect based on the relevant parameters obtained after the left atrial appendage occlusion operation.

[0167] Example

[0168] Patient A, who had a history of atrial fibrillation and stroke, underwent percutaneous left atrial appendage occlusion and underwent coronary CTA examination 3 months after the operation. We can obtain the original images of the patient's coronary CTA or left atrial CTA through the PACS system, determine the axis of the occluder and the plane of the left atrial appendage opening, and perform three-dimensional reconstruction and segmentation of the left atrium and left atrial appendage occluder. According to the POET principle, when obtaining the residual shunt parameters, in the three-dimensional reconstructed image, rotate the sagittal or coronal plane to find the residual channel between the left atrium and the left atrial appendage, place the axial section reference line at the residual shunt, and measure the size of the residual shunt in the axial section. Use VR mode to display the heart structure, adjust the image so that the mitral valve annulus is below, the aorta is in front, and the left superior pulmonary vein is above. The position of the residual shunt is expressed using the clock method and expressed as Figures 1a-1b In the 3D multi-plane reconstruction, the crosshairs of the axial section were rotated to locate the exposed shoulder of the occluder. The Euclidean distance between the left atrial appendage opening plane and the highest point of the exposed shoulder was calculated as the height of the exposed shoulder. Using the endothelialization quantitative analysis function, the HAT area on the occluder surface was delineated, and the endothelialization ratio was calculated using the endothelialization calculation formula. Finally, if the patient had no DRT and the endothelialization ratio was ≥90%, the patient was considered to have a good postoperative outcome.

[0169] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or apparatuses.

[0170] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0171] The embodiments of the present invention are described in detail above. Specific implementation methods are used herein to illustrate the present invention. The description of the above embodiments is only used to help understand the method and system of the present invention. They are only embodiments of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention, and the content of this specification should not be understood as limiting the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for quantitatively analyzing the effect of left atrial appendage occlusion based on CT images, characterized in that: The method comprises: Step 1: Acquire CTA images of the patient's heart after left atrial appendage occlusion surgery; Step 2: performing three-dimensional reconstruction on the cardiac CTA image sequence to obtain a three-dimensional image of the heart, and obtaining a three-dimensional image of the left atrium using a left atrium segmentation model; Step 3: Segment the LAA occluder area in the CTA image using the LAA occluder segmentation model; Step 4: Based on the segmented left atrial appendage occluder area, a three-dimensional model of the left atrial appendage occluder is constructed, and a three-dimensional image of the left atrial appendage occluder is obtained through three-dimensional reconstruction; Step 5: Based on the three-dimensional image of the left atrial appendage occluder, obtain the three-dimensional plane of the left atrial appendage occluder, the front view of the occluder, and the plane of the left atrial appendage opening; Step 6: Extracting evaluation parameters after left atrial appendage occlusion surgery, the parameters including at least: the size, location, and grade of residual shunt, the location, shoulder position, and height of the occluder, the degree of endothelialization of the occluder, and the location of device-related thrombus; Step 7: Evaluate the left atrial appendage occlusion effect based on the obtained evaluation parameters after left atrial appendage occlusion surgery; The specific steps of step 5 include: Step 5.1: With the center point of the occluder rivet in the LAA occluder image as the starting point P1, the tip of the occluder as the end point P2, and the path direction P1→P2, the connecting line from P1 to P2 is used as the initial axis of the occluder. The axis of the occluder is continuously corrected through iterative calculation to finally obtain the actual axis of the occluder; Step 5.2: The tangent line passing through the center point of the occluder rivet and perpendicular to the initial axis is used as the tangent line of the left atrial surface of the occluder. The tangent line of the left atrial surface of the occluder is corrected during the iterative calculation of the occluder axis, and the actual tangent line of the left atrial surface of the occluder is finally obtained. Step 5.3: Fix the obtained tangent of the left atrial surface of the occluder and the actual axis of the occluder at 90 degrees, thereby obtaining the three-dimensional planes of the left atrial appendage occluder: axial section, sagittal plane, and coronal plane; Step 5.4: Fix the obtained axial, coronal, and sagittal planes at 90° angles to each other, and adjust the window width and window position of the axial section until the CT image layer of the occluder is displayed, thereby obtaining a frontal view of the left atrial appendage occluder; Step 5.5: In the 3D reconstruction, use the orthogonal plane method to place the crosshairs at the level of the proximal circumflex artery. Step 5.6: Rotate the coronal and sagittal crosshairs in sequence, adjust their positions, and place them at the circumflex artery and pulmonary vein crest, respectively, to determine the location of the left atrial appendage opening. The plane corresponding to the crosshairs at the circumflex artery and pulmonary vein crest is used as the left atrial appendage opening plane.

2. The method for quantitatively analyzing left atrial appendage occlusion effect based on CT images according to claim 1, characterized in that: Step 3: The steps for constructing the left atrial appendage occluder segmentation model are as follows: Step 3.1: Acquire CTA images of patients after left atrial appendage occlusion surgery, segment the left atrial appendage occluder, and obtain a training dataset and corresponding masks; Step 3.2: Based on the training data set and mask obtained in step 3.1, a U-net algorithm model is trained, and the trained model is used as the left atrial appendage occluder segmentation model.

3. The method for quantitatively analyzing left atrial appendage occlusion effect based on CT images according to claim 1, characterized in that: Step 6 includes the following specific steps: Step 6.1: Determine the location, size, and grade of the residual shunt; Step 6.2: Determine the position of the occluder, including the height and position of the shoulder; Step 6.3: Determine the contrast attenuation thickness (HAT) and endothelialization degree of the occluder; Step 6.4: Assess the HAT. If the HAT thickness is greater than 3 mm, observe whether the HAT is regular. If not, consider it as a device-related thrombus (DRT) and indicate the current DRT position using the clock method.

4. The method for quantitatively analyzing left atrial appendage occlusion effect based on CT images according to claim 3, characterized in that: The specific steps of step 6.1 are: Step 6.1.1: Rotate the crosshairs in the coronal and sagittal planes, respectively, and define the path of contrast medium from the left atrium to the left atrial appendage as the residual shunt in the sagittal or coronal plane. Step 6.1.2: Determine the control line at the residual shunt based on the axial section and obtain the cross section of the residual shunt; Step 6.1.3: Determine the minimum diameter and area of ​​the residual shunt based on the cross section; Step 6.1.4: Classify the residual shunt according to its minimum diameter: if the diameter is 0-1 mm, it is trace; if the diameter is 1 mm-3 mm, it is small; if the diameter is 3 mm-5 mm, it is medium; if the diameter is greater than 5 mm, it is large.

5. The method for quantitatively analyzing left atrial appendage occlusion effect based on CT images according to claim 4, characterized in that: The specific steps of step 6.2 are: Step 6.2.1: Rotate the crosshairs of the axial section until the occluder shoulder is located in the sagittal or coronal plane. Step 6.2.2: Calculate the Euclidean distance between the left atrial appendage opening plane as the starting point and the highest point of the occluder shoulder as the end point. This distance is used as the height of the occluder shoulder.

6. The method for quantitatively analyzing left atrial appendage occlusion effect based on CT images according to claim 5, characterized in that: The specific steps of step 6.3 are: Step 6.3.1: Detect the intersection areas of all occluder beams on the three-dimensional image of the left atrial appendage occluder and select the intersection area with the largest volume as the center point of the occluder; Step 6.3.2: Start searching near the center of each beam. When the angle between the vector A from the current search point to the next search point and the vector B from the previous search point to the current search point is the largest, the current search point is considered the beam inflection point. Step 6.3.3: Calculate the planes that can be constructed using the beam inflection points determined in Step 6.3.

2. Select the plane with the largest number of inflection points and use the normal vector of that plane as the vector from the center point to the end of the occluder. If all planes constructed using the beam inflection points have the same number of inflection points, select the plane closest to the center point of the occluder and use the normal vector of that plane as the vector from the center point to the end of the occluder. Step 6.3.4: Using the vector from the center point to the end of the occluder determined in Step 6.3.3 as the plane normal, calculate the area of ​​the quasi-circular shape enclosed by the occluder beam on each plane. The plane with the largest area is the maximum cross-section of the occluder. Step 6.3.5: Segment the endothelialized area on the surface of the left atrial appendage occluder using the endothelialized area segmentation method; Step 6.3.6: Plot the endothelial thickness at each pixel in the endothelialized area using different colors; Step 6.3.7: Calculate the endothelialization ratio using the following formula and use the endothelialization ratio as the degree of endothelialization after left atrial appendage occlusion: Endothelialization ratio = endothelialization coverage area of ​​the left atrial surface of the occluder / area of ​​the left atrial surface of the occluder 100%.

7. The method for quantitatively analyzing left atrial appendage occlusion effect based on CT images according to claim 6, characterized in that: The endothelialization area segmentation method in step 6.3.5 is the grayscale threshold segmentation algorithm, and its specific steps are as follows: Step 6.3.5.1: Starting from the maximum cross-section of the occluder, search for pixels within the maximum cross-section range in each image layer that meet the endothelialization CT value range along the two normal vectors of the maximum cross-section of the occluder determined in Step 6.3.4; Step 6.3.5.2: When a pixel point belonging to the left atrial region is found, the search is stopped, and the found pixel point is used as a pixel point of the endothelialized region, and the region formed by combining these pixel points is used as the endothelialized region.

8. The method for quantitatively analyzing left atrial appendage occlusion effect based on CT images according to claim 7, characterized in that: The endothelialization region segmentation method in step 6.3.5 is the U-net segmentation algorithm, and its specific steps are as follows: Step 1: Based on CT image data with pre-labeled endothelialization areas and reviewed by cardiology professionals, a U-net algorithm is used to train an endothelialization area segmentation model. Step 2: Obtain a CT image of the current patient and use the trained endothelialization region segmentation model to segment the endothelialization region in the image.

9. A system for extracting left atrial appendage data parameters based on CT images, implemented based on the method of claim 1, characterized in that: The system comprises: An image sequence acquisition module is used to acquire CTA images of the patient's heart after left atrial appendage occlusion surgery; An image reconstruction module performs three-dimensional reconstruction on the cardiac CTA image sequence to obtain a three-dimensional cardiac image; Segmentation module, including, a left atrium segmentation unit, configured to segment the left atrium into a three-dimensional image using a left atrium segmentation model; a left atrial appendage occluder segmentation unit, configured to segment the left atrial appendage occluder region in the CTA image using a left atrial appendage occluder segmentation model; A three-dimensional reconstruction module is used to construct a three-dimensional model of the left atrial appendage occluder based on the segmented left atrial appendage occluder area, and obtain a three-dimensional image of the left atrial appendage occluder through three-dimensional reconstruction; A three-dimensional plane acquisition module is used to acquire a three-dimensional plane of the left atrial appendage occluder, a front view of the occluder, and a left atrial appendage opening plane based on a three-dimensional image of the left atrial appendage occluder; An occluder parameter extraction module is used to extract evaluation parameters after left atrial appendage occlusion surgery, including at least the size, location, and grade of residual shunt, the location, shoulder position, and height of the occluder, the degree of endothelialization of the occluder, and the location of device-related thrombus; The evaluation module is used to evaluate the left atrial appendage occlusion effect based on the obtained evaluation parameters after the left atrial appendage occlusion operation.

Images