Artificial intelligence-based tricuspid annulus positioning and plane tilt angle automatic calculation method and system

Abstract

The application provides a kind of artificial intelligence-based tricuspid annulus accurate positioning and plane inclination automatic calculation method, comprising: obtaining and preprocessing three-dimensional medical image;Based on deep learning model, key structures such as right atrium, tricuspid valve and right ventricle are automatically segmented;Extraction of leaflet junction and annulus reference point, construct initial annulus curve;Through path optimization and resampling, improve curve continuity and point distribution uniformity;Using iterative strategy to optimize the spatial distribution of annulus point set, so as to fit the surface characteristics of valve leaflet;Determine the annulus plane and calculate the normal vector by using mathematical fitting method, and further obtain the included angle with horizontal plane.The method realizes the automatic identification and inclination quantization of tricuspid annulus, improves the accuracy and clinical applicability of morphological measurement, and reduces manual error.

Description

Technical Field

[0001] This invention relates to the fields of medical image processing and artificial intelligence (AI) technology, and in particular to a method and system for automatic calculation of tricuspid valve annulus localization and plane tilt angle based on an AI segmentation model. Technical Background

[0002] The tricuspid valve, located between the right atrium and right ventricle, is crucial for the normal functioning of the cardiovascular system. In clinical practice, tricuspid valve diseases (such as tricuspid regurgitation, tricuspid stenosis, and tricuspid valve prolapse) often require precise anatomical and functional assessment using imaging techniques (such as echocardiography, CT, or MRI). However, traditional tricuspid valve imaging analysis methods rely on manual operation, requiring clinicians to manually mark the tricuspid valve annulus and estimate its plane tilt. This method is not only time-consuming and labor-intensive but also easily influenced by the operator's subjective experience, leading to low accuracy and consistency of the analysis results. Furthermore, the complex morphology of the tricuspid valve and its significant influence from dynamic cardiac motion further increase the difficulty of manual analysis, thus limiting the accuracy and repeatability of imaging analysis.

[0003] Plane tilt angle is a key indicator describing the spatial location and geometry of the tricuspid valve annulus, and it has significant clinical applications. For example, in tricuspid annuloplasty, the plane tilt angle helps determine the optimal implantation position and angle of the annulus support ring to ensure good valve closure; during transcatheter tricuspid valve replacement (TTVR), the plane tilt angle helps select appropriate interventional devices and reduces the risk of displacement and rotation during valve prosthesis implantation; in tricuspid valve repair, the plane tilt angle can serve as an important reference indicator for predicting the recovery of right ventricular function postoperatively. However, most existing methods focus on the segmentation of valve anatomy, lacking a systematic and automated solution for annular spatial location and plane tilt angle calculation.

[0004] To address these challenges, there is an urgent need to develop an AI-based tricuspid valve image analysis technology to achieve automated, precise, and standardized measurement of valve morphology.

[0005] This invention proposes an innovative AI-based tricuspid valve image analysis method. By combining a deep learning segmentation model with precise geometric measurement algorithms, it can automatically identify the tricuspid valve annulus and automatically derive the annulus's planar tilt angle through spatial geometric analysis. This method not only improves the efficiency and accuracy of image analysis but also provides clinicians with more objective and accurate diagnostic evidence, aiding in the precise assessment and surgical planning of tricuspid valve disease. Summary of the Invention

[0006] This invention combines artificial intelligence technology and image processing algorithms to provide an AI-based method for precise localization of the tricuspid valve annulus and automatic calculation of its planar tilt angle, comprising:

[0007] Step 1: Acquire three-dimensional medical image data. Use imaging equipment such as CT, MRI or echocardiography to acquire cardiac images of the patient and perform preprocessing to improve data quality.

[0008] Step 2: Identification and segmentation of the right atrium, tricuspid valve and right ventricle. Based on a deep learning segmentation model, the cardiac anatomical structures are automatically extracted, and the segmentation results are optimized through morphological operations and connected component analysis to improve segmentation accuracy.

[0009] Step 3: Extract key anatomical points of the tricuspid valve. Specifically, based on the anatomical structure and spatial topological relationship of the leaflets, extract the feature point set of the tricuspid valve leaflets, and automatically identify the key points of leaflet junction through distance measurement and anatomical location information.

[0010] Step 4: Construct the initial valve loop. Based on the distribution of key points and the characteristics of the valve leaf region, generate the valve loop curve. Improve the continuity and uniformity of the valve loop curve through path optimization and resampling.

[0011] Step 5: Iteratively optimize the valve annulus localization. By using a localization method based on the surface features of the valve leaflet, the spatial distribution of the valve annulus point set is adjusted, and a smoothing and optimization strategy is adopted to ensure that the valve annulus morphology conforms to the anatomical structure of the valve leaflet.

[0012] Step 6: Determine the tricuspid valve annulus plane. Based on the optimized annulus point set, use mathematical fitting methods to determine the annulus plane and calculate its normal vector.

[0013] Step 7: Calculate the annular plane tilt angle.

[0014] Furthermore, in step 1, the preprocessing includes, but is not limited to, noise removal, normalization, and registration of the image.

[0015] Furthermore, the key structure segmentation in step 2 includes, but is not limited to, the right atrium, the anterior tricuspid valve leaflet, the posterior tricuspid valve leaflet, the septal tricuspid valve leaflet, and the right ventricle. The segmentation method can be manual annotation, traditional image segmentation algorithms, or automatic segmentation models based on deep learning. The automatic segmentation models include, but are not limited to, nnUNet and SwinUNETR, and the analysis results are optimized by combining morphological operations and connected component analysis to remove noise and artifacts and ensure the continuity and accuracy of the segmented region.

[0016] Further, in step 3, based on anatomical structural features, the tricuspid valve region is divided into the anterior leaflet, posterior leaflet, and septal leaflet, and the corresponding segmentation result point sets are extracted for each; then, the Euclidean distance between any two adjacent leaflet point sets is calculated, and the distance is determined within the range that satisfies the condition...

[0017] d(p,q)<α

[0018] Under the condition that α is a preset threshold, the point farthest from the right ventricle is selected from the candidate matching point set, and the key junction points are determined, including the septal-anterior leaflet junction point P. A Anterior-posterior lobe junction point P B and the junction point P between the posterior lobe and the septum. C .

[0019] Furthermore, in step 4, an initial lobe loop is constructed, specifically as follows:

[0020] 4.1: Based on the tricuspid lobe segmentation results, binary segmentation images of the anterior lobe, posterior lobe, and septal lobe are generated sequentially;

[0021] 4.2: Based on the location of the leaflet junction, key point pairs P are constructed sequentially. A -P B P B -P C P C -P A As the initial reference point for the valve ring;

[0022] 4.3: Within the leaflet structure, for each pair of key points, a path P(t) (t∈[0,1]) is found. A skeletonization algorithm combined with dynamic programming is used, and inverse weights based on distance transformation are introduced to obtain an approximate central axis path. This path is the connection (P). A -P B P B -P C P C -P A The centerline path;

[0023] 4.4: Resample the centerline path, focusing on the front and back lobes (P... A -P B PB -P C )Sampling 10 points along the path, for the septum (P C -P A ) 20 points were sampled along the path;

[0024] 4.5: Connect all resampling points sequentially to form a closed initial loop curve, thus obtaining the initial loop point set.

[0025] Furthermore, in step 5, the specific method for optimizing the lobe loop localization includes:

[0026] 5.1: Obtain the three-dimensional voxel data of the tricuspid valve segmentation region and use the surface reconstruction algorithm to convert the leaflet region into a continuous triangular mesh or point cloud; calculate the unit normal vector of each vertex or facet on the leaflet surface and ensure that the normal points to the atrial side to ensure the correctness of the subsequent stepping direction;

[0027] 5.2: For each point in the initial lobe ring point set R0 obtained in step 4 Using nearest-point projection or interpolation methods, determine the nearest point of each point on the leaflet mesh surface. Get The unit normal vector n at the location i This serves as the direction for optimization steps at that point;

[0028] 5.3: An iterative step-by-step update method is adopted for each point. Step size δ = 1 mm, maximum number of iterations k max For each point Perform iterative update operations.

[0029] Furthermore, step 5.3 specifically includes the following steps:

[0030] 5.3.1: Calculate the current point The nearest point on the leaflet surface To ensure that the path converges to the leaflet surface;

[0031] 5.3.2: Obtaining Local Normals

[0032] 5.3.3: According to the formula Perform iterative updates;

[0033] 5.3.4: Determine the new location If the device is still within the leaflet structure, continue iterating. If it exceeds the leaflet region or the mask value is below a set threshold, stop iterating and record the current state. As the final positioning point;

[0034] 5.3.5 If the number of iterations reaches the maximum value k max If this occurs, the process will be forcibly stopped to prevent abnormal situations from affecting the optimization process.

[0035] 5.4: Add all final iteration points After connection, an optimized tricuspid valve annulus curve is formed. The annulus point set is then smoothed to eliminate noise introduced by discrete iteration. At the same time, the smoothed curve is resampled with equal arc length to ensure uniform distribution of annulus points.

[0036] Furthermore, in step 6, the least squares method is used to fit a plane equation to the set of lobe loop points obtained in step 5.

[0037] Ax + By + Ca + D = 0

[0038] The normalized normal vector n = (A, B, C) is obtained by using singular value decomposition or covariance matrix method.

[0039] Furthermore, in step 7, the angle between the fitting plane normal vector and the horizontal plane is calculated as follows:

[0040] θ = arccos(|C|)

[0041] The calculation results are converted into angles to reflect the tilt of the annular plane.

[0042] This invention also provides an artificial intelligence-based system for automatic tricuspid valve annulus localization and tilt angle calculation, comprising:

[0043] The medical image data processing module is used to acquire and preprocess three-dimensional medical image data, and to remove noise, normalize and register CT, MRI or echocardiogram data to improve data quality.

[0044] The anatomical structure segmentation module is used to automatically segment the right atrium, anterior tricuspid valve leaflet, posterior tricuspid valve leaflet, septal tricuspid valve leaflet, and right ventricle based on a deep learning model, and optimizes the segmentation results by combining morphological operations and connected component analysis.

[0045] The key point extraction module is used to extract leaflet feature points based on the leaflet anatomical structure and identify key points at the leaflet junction through distance measurement and anatomical location information.

[0046] The valve loop construction module is used to generate valve loop curves based on key point distribution, and improve the continuity and uniformity of valve loops through path optimization and resampling processing.

[0047] The valve annulus optimization module is used to adjust the spatial distribution of the valve annulus point set based on the surface features of the valve leaflet, and to use smoothing and optimization strategies to ensure that the valve annulus morphology conforms to the anatomical structure of the valve leaflet.

[0048] The Lobe Ring Plane Calculation Module is used to fit the Lobe Ring Plane based on the optimized Lobe Ring Point Set and calculate its normal vector.

[0049] The tilt angle calculation module is used to calculate the angle between the normal vector of the valve ring plane and the horizontal plane, and to determine the spatial tilt of the valve ring.

[0050] Technical effect

[0051] This invention, through the combination of artificial intelligence and geometric modeling technology, achieves automated and precise analysis of tricuspid valve annulus positioning and planar tilt angle. Its technical effects are as follows:

[0052] 1. High efficiency: The automated process reduces the time required for traditional manual analysis to just a few seconds, significantly improving diagnostic efficiency.

[0053] 2. High precision: Based on deep learning segmentation models and optimization algorithms, it achieves sub-millimeter level spatial positioning accuracy, and the plane tilt angle calculation error is significantly lower than that of traditional methods.

[0054] 3. Consistency: It eliminates the influence of human factors, ensuring a high degree of consistency and reproducibility of analytical results, and is suitable for multicenter studies and clinical applications.

[0055] 4. Adaptability to complex anatomy: Through iterative optimization and spatial geometric analysis, it adapts to the differences in tricuspid valve morphology among different patients, generates personalized valve annulus models, and supports dynamic cardiac motion analysis.

[0056] 5. Clinical applicability: The output results (plane tilt angle) can be directly applied to the planning of valve repair surgery, transcatheter replacement surgery and repair surgery, improving the success rate of surgery and the ability to predict postoperative recovery.

[0057] In summary, this invention significantly improves the diagnostic efficiency and treatment level of tricuspid valve-related diseases, provides reliable support for clinical decision-making, and has important application value. Attached Figure Description

[0058] Figure 1 This is a flowchart of an artificial intelligence-based method for tricuspid valve annulus localization and automatic plane tilt calculation.

[0059] Figure 2 This is a schematic diagram of the division of the right atrium, tricuspid valve, and right ventricle as described in step 2.

[0060] Figure 3 This is a schematic diagram of the key anatomical points of the tricuspid valve extracted as described in step 3.

[0061] Figure 4 This is a schematic diagram of the initial lobe ring constructed as described in step 4.

[0062] Figure 5 This is a schematic diagram of the final lobe loop after iterative optimization as described in step 5.

[0063] Figure 6 This is a schematic diagram of the lobe loop iteration process described in step 5.

[0064] Figure 7 This is a schematic diagram of the tricuspid valve annulus plane determined in step 6.

[0065] Figure 8 This is a schematic diagram of the final tricuspid valve annulus plane tilt angle obtained as described in step 7. Detailed Implementation

[0066] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that the embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. Furthermore, it should be understood that after reading the disclosure of this invention, those skilled in the art can make various modifications or alterations to the invention, and these equivalent forms also fall within the scope of protection defined by this invention.

[0067] This invention combines artificial intelligence technology with image processing algorithms to provide an AI-based method and system for precise localization of the tricuspid valve annulus and automatic calculation of its planar tilt angle. The method's flow is as follows: Figure 1 As shown, the specific steps include:

[0068] Step 1: Scan the patient's heart using high-precision medical imaging equipment (such as CT, MRI, or echocardiography) to obtain its three-dimensional image data. To improve the accuracy of subsequent segmentation and calculation, the raw data is preprocessed, including but not limited to:

[0069] Noise Removal: Filtering algorithms (such as Gaussian filtering or median filtering) are used to reduce random noise introduced during imaging while preserving important anatomical boundary information.

[0070] Normalization: Maps image grayscale values ​​to a uniform range ([0,1]) to eliminate differences caused by different devices or scanning parameters and ensure the consistency of model input.

[0071] Registration: For multimodal image data (such as joint analysis of CT and MRI), rigid and non-rigid registration techniques are used to align images from different sources to achieve spatial consistency.

[0072] Step 2: This step addresses the needs of tricuspid valve image analysis by automatically identifying and segmenting the tricuspid valve and its surrounding key anatomical structures that can form connections. It focuses on extracting the spatial relationships of structures such as the right atrium, tricuspid valve, and right ventricle to provide a complete basis for valve annulus localization. The segmentation results are shown below. Figure 2As shown. In this step, a deep learning-based segmentation algorithm, such as nnUnet or SwinUNETR, can be used to automatically generate accurate label maps for each anatomical structure based on the segmentation results of the deep learning model. Combined with morphological optimization and connected component analysis, a certain smoothing process is performed to ensure the continuity and integrity of the boundaries of each structure.

[0073] Step 3: Extract key anatomical points of the tricuspid valve. Specifically, based on the segmentation results of Step 2, automatically identify the key anatomical points of the anterior, posterior, and septal leaflets of the tricuspid valve. The results are as follows: Figure 3 As shown. Specifically, it includes the following sub-steps:

[0074] Step 3.1: Leaflet Division

[0075] Based on the anatomical structure, the tricuspid valve region is subdivided into the anterior leaflet, posterior leaflet, and septal leaflet, and their spatial point sets are extracted respectively: M1, M2, and M3.

[0076] Step 3.2: Neighborhood Matching and Threshold Determination

[0077] Calculate the Euclidean distance between adjacent leaflet point sets.

[0078] d(p,q)=||pq||

[0079] When a certain pair of points satisfies

[0080] d(p,q)<α

[0081] Where α is a preset threshold, it is considered that the point has an anatomical proximity relationship with p and q.

[0082] Step 3.3: Identify key handover points

[0083] Among candidate matching points satisfying proximity, points farther from the right ventricle (obtained through segmentation) in each group are selected according to the principle of "farthest from the right ventricle". Key points of tricuspid valve leaflet junction are determined, including the septal-anterior leaflet junction point P. A Anterior-posterior lobe junction point P B The junction point P between the posterior lobe and the septal lobe C .

[0084] Step 4: Construct the initial valve annulus. Specifically, based on the key points extracted in Step 3, construct the initial morphology of the tricuspid valve annulus. The result is as follows: Figure 4 As shown. Specifically, it includes the following sub-steps:

[0085] Step 4.1: Generate a binary segmentation map

[0086] Based on the tricuspid lobe segmentation results, a binary image B(x,y,z) is generated, where:

[0087]

[0088] Furthermore, the anterior lobe B can be obtained separately. ant (x,y,z), posterior lobe B post (x,y,z) and septal B sep Binary segmentation map of (x,y,z).

[0089] Step 4.2: Construct keypoint pairs

[0090] To describe the anatomical structure of the tricuspid valve, the following three pairs of key points are defined:

[0091] Anterior lobe to posterior lobe path point pair: B ant-post =(P A -P B ), indicating the beginning and end points of the anterior lobe.

[0092] Path point pair from posterior lobe to septum: P post-sep =(P B -P C ), indicating the beginning and end points of the posterior lobe.

[0093] Path point pair from septum to anterior lobe: P sep-ant =(P C -P A ), indicating the start and end points of the septum.

[0094] Step 4.3: Find the centerline path

[0095] For each pair of key points, an approximate central axis path is found within the corresponding leaflet structure. The specific method is as follows:

[0096] Step 4.3.1: Perform skeletonization on the binary segmentation image B(x,y,z) of each lobe and extract its topological centerline. The skeletonization algorithm preserves the core paths of connected regions through iterative erosion operations, generating a single-pixel-wide skeleton structure S(x,y,z).

[0097] Step 4.3.2: Calculate the distance transformation field D(x,y,z) of the segmentation map, defined as follows:

[0098]

[0099] Here, (u,v,w) represents the coordinates of the background region. The distance transform field reflects the distance from each foreground pixel to the nearest background pixel.

[0100] Step 4.3.3: Combining the distance transformation field D(x,y,z) and the skeletonization result S(x,y,z), a dynamic programming algorithm is used to search for keypoint pairs (P... A -P B The optimal path is found. The objective function for this path is:

[0101]

[0102] Where, p i Let be the i-th point on the path, and n be the path length. By minimizing the objective function C(path), a smooth path that is close to the center of the leaflet can be obtained. This ultimately forms a "centerline path" connecting key point pairs.

[0103] Step 4.4: Resampling Path

[0104] To standardize the representation of different lobe paths and ensure a uniform distribution of path points, the acquired centerline paths are resampled. Specifically, the paths of the anterior and posterior lobes (P...) are resampled. ant-post and P post-sep Uniform sampling is performed, selecting 10 points. For the path P of the septum... sep-ant Uniform sampling is used, selecting 20 points. The sampling method is based on path arc length parameterization, defining an arc length accumulation function:

[0105]

[0106] Where, q j For the j-th point on the path, the total arc length L(m) is divided into K equal segments according to the number of target points K, and interpolation is performed in each segment to obtain new sampling points.

[0107] Step 4.5: Connect the three sets of resampling points in sequence to form a closed initial valve loop curve, ensuring the uniformity of the valve loop point distribution and the continuity of the curve.

[0108] Step 5: Iterative optimization of lobe ring localization, specifically: based on the initial lobe ring obtained in Step 4, the final lobe ring is obtained through iterative optimization, as shown in the following figure. Figure 5 As shown. Specifically, it includes the following sub-steps:

[0109] Step 5.1: Obtain 3D voxel data of the tricuspid valve segmentation region and use a surface reconstruction algorithm (such as Marching Cubes or Poisson Surface Reconstruction) to convert the leaflet region into a continuous triangular mesh or point cloud. Calculate the unit normal vector for each vertex or facet on the leaflet surface, with the normal direction pointing towards the atrium to ensure the correctness of subsequent stepping directions.

[0110] Step 5.2: For each point in the initial lobe ring obtained in Step 4 The nearest point on the leaflet mesh surface is determined using nearest point projection or interpolation methods. And obtain the point The unit normal vector n at the location i This serves as a direction for optimizing the stepping.

[0111] Step 5.3: Iterative step update, such as... Figure 6 As shown, for each point Set the step size δ = 1 mm and the maximum number of iterations k. max For each point The specific steps for iterative optimization and updates are as follows:

[0112] Step 5.3.1: Calculate the nearest point on the leaflet surface to the current point. To ensure that the path converges to the leaflet surface;

[0113] Step 5.3.2: Obtain the local normal To guide the direction of optimization;

[0114] Step 5.3.3: Perform iterative updates, through... Calculate the new point position;

[0115] Step 5.3.4: Determine the new position Is it still within the leaflet structure? If within the leaflet, continue iteration; if outside the leaflet region or the mask value is below a set threshold, forcibly stop iteration and record the final location point:

[0116] Step 5.3.5: If the number of iterations reaches the maximum value k max If this occurs, the process will be forcibly stopped to prevent abnormal situations from affecting the optimization process.

[0117] Step 5.4: Complete all final iteration points Connect the points to form the optimized tricuspid valve annulus curve. Smooth the annulus point set using methods such as cubic spline interpolation or moving average filtering to eliminate noise introduced by discrete iterations. Resample the smoothed curve using equal arc lengths to ensure a uniform distribution of the annulus points.

[0118] Step 6: Determine the tricuspid valve annulus plane. Specifically, based on the optimized annulus point set obtained in Step 5, the least squares method is used to fit the annulus plane to obtain the spatial orientation of the annulus. The result is as follows: Figure 7 As shown.

[0119] The final set of lobe ring points is:

[0120]

[0121] The plane equation fitted using the least squares method is:

[0122] Ax + By + Cz + D = 0

[0123] Constrained by the objective function:

[0124]

[0125] Singular Value Decomposition (SVD) is used, and the eigenvectors corresponding to the smallest eigenvalues ​​are taken as the plane normal vectors n = (A, B, C).

[0126] Step 7: Calculate the angle between the plane normal vector of the fitted plane in Step 6 and the horizontal plane to quantify the spatial tilt of the lobe ring plane. The result is as follows: Figure 8 As shown. Specific calculation method: The horizontal plane is defined as z = 0, and its normal vector is n. h = (0,0,1), the angle between the horizontal planes

[0127] θ = arccos(|n·n h |)=arrccos(|C|)

[0128] Convert the calculation results to degrees.

[0129]

[0130] The resulting angle θ reflects the degree of inclination of the annular plane relative to the horizontal plane, providing key geometric parameters for preoperative assessment and surgical planning.

[0131] This invention also provides an artificial intelligence-based system for tricuspid valve annulus localization and automatic plane tilt calculation, comprising:

[0132] The medical image data processing module is used to acquire and preprocess three-dimensional medical image data, receive CT, MRI or echocardiography (3DE) data, and perform noise removal, normalization and registration to improve data quality and image consistency.

[0133] The anatomical structure segmentation module is used to automatically segment the right atrium, anterior tricuspid valve leaflet, posterior tricuspid valve leaflet, septal tricuspid valve leaflet, and right ventricle based on a deep learning model. It also combines morphological operations and connected component analysis to optimize the segmentation results, thereby improving segmentation accuracy and the continuity of the segmented regions.

[0134] The key point extraction module is used to extract leaflet feature points based on the leaflet anatomical structure and identify key points at the leaflet junction through distance measurement and anatomical location information to determine the starting point for the construction of the valve annulus.

[0135] The valve annulus construction module is used to generate valve annulus curves based on key point distribution, and improves the continuity and uniformity of valve annulus curves through path optimization, skeletonization and resampling processing to ensure that the valve annulus conforms to anatomical features.

[0136] The valve ring optimization module is used to adjust the spatial distribution of the valve ring point set based on the surface features of the valve leaflet. It uses iterative stepping, surface normal calculation and smoothing to ensure that the valve ring morphology conforms to the anatomical structure of the valve leaflet and improves the stability of the model.

[0137] The Lobe Ring Plane Calculation Module is used to determine the Lobe Ring Plane based on the optimized Lobe Ring Point Set using least squares fitting, singular value decomposition (SVD), or covariance matrix methods, and to calculate its normal vector to obtain the spatial orientation of the Lobe Rings.

[0138] The tilt calculation module is used to calculate the angle between the normal vector of the valve annulus plane and the horizontal plane, determine the spatial tilt of the valve annulus, and provide support for preoperative planning and surgical optimization.

[0139] This system can automatically execute the entire process from medical image input to valve annulus tilt angle calculation, reducing manual operation, improving the accuracy and stability of valve annulus morphology analysis, and providing precise imaging support for tricuspid valve repair, valve repair and transcatheter valve replacement.

Claims

1. A method for precise positioning of the tricuspid valve annulus and automatic calculation of its planar tilt angle based on artificial intelligence, characterized in that, Includes the following steps: Step 1: Acquire three-dimensional medical image data. Use CT, MRI or echocardiography imaging equipment to acquire cardiac images of the patient and perform preprocessing to improve data quality. Step 2: Identification and segmentation of the right atrium, tricuspid valve and right ventricle. Based on a deep learning segmentation model, the cardiac anatomical structures are automatically extracted, and the segmentation results are optimized through morphological operations and connected component analysis to improve segmentation accuracy. Step 3: Extract key anatomical points of the tricuspid valve. Specifically, based on the anatomical structure and spatial topological relationship of the leaflets, extract the feature point set of the tricuspid valve leaflets, and automatically identify the key points of leaflet junction through distance measurement and anatomical location information. Step 4: Construct the initial valve ring, and construct key point pairs sequentially based on the leaflet junction positions. - , - , - As the initial reference point for the valve annulus; within the valve leaflet structure, find a path for each pair of key points. The centerline path is obtained by combining skeletonization algorithm with dynamic programming algorithm. This path is the connection. The centerline path is determined; the centerline path is resampled; all resampled points are connected sequentially to form a closed initial loop curve, thus obtaining the initial loop point set. ; Step 5: Iteratively optimize the lobe ring localization, obtain 3D voxel data of the tricuspid valve segmentation region, and use a surface reconstruction algorithm to convert the leaflet region into a continuous triangular mesh or point cloud; calculate the unit normal vector of each vertex or facet on the leaflet surface; and apply the initial lobe ring point set obtained in Step 4. Each point in Using nearest-point projection or interpolation methods, determine the nearest point of each point on the leaflet mesh surface. , obtain Unit normal vector at the location This serves as the direction for the optimization step at that point; an iterative step update method is adopted for each point. Step size δ, maximum number of iterations For each point Perform iterative update operations; connect all final iteration points to form the optimized tricuspid valve annulus curve; Step 6: Determine the tricuspid valve annulus plane. Based on the optimized annulus point set, use mathematical fitting methods to determine the annulus plane and calculate its normal vector. Step 7: Calculate the annular plane tilt angle.

2. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to claim 1, characterized in that, In step 1, the preprocessing includes, but is not limited to, noise removal, normalization, and registration of the image.

3. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to claim 2, characterized in that, The key structure segmentation in step 2 includes, but is not limited to, the right atrium, the anterior tricuspid valve leaflet, the posterior tricuspid valve leaflet, the septal tricuspid valve leaflet, and the right ventricle. The segmentation method can be manual annotation, traditional image segmentation algorithms, or automatic segmentation models based on deep learning. The automatic segmentation models include, but are not limited to, nnUNet and SwinUNETR, and the analysis results are optimized by combining morphological operations and connected component analysis to remove noise and artifacts and ensure the continuity and accuracy of the segmented region.

4. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to claim 3, characterized in that, In step 3, based on anatomical structural features, the tricuspid valve region is divided into the anterior leaflet, posterior leaflet, and septal leaflet, and the corresponding segmentation result point sets are extracted for each. Then, the Euclidean distance between any two adjacent leaflet point sets is calculated, and the following conditions are met: Where α is a preset threshold, the point farthest from the right ventricle is selected from the candidate matching point set to determine the key junction points, including the septal-anterior leaflet junction point. Anterior-posterior lobe junction and the junction of the posterior lobe and septum. .

5. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to claim 1, characterized in that, In step 4, when constructing the initial lobe loop, a reverse weight based on distance transformation is introduced during the process of obtaining the centerline path using a skeletonization algorithm combined with a dynamic programming algorithm; furthermore, when resampling the centerline path, the front lobe and the back lobe ( - , - ) Sample 10 points along the path, and perform sampling on the septum ( - ) 20 points were sampled along the path.

6. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to claim 1, characterized in that, In step 5, when calculating the unit normal vector of each vertex or surface element on the leaflet surface, it is ensured that the normal points towards the atrial side; and when using the iterative step update method, the step size is... =1mm; After connecting all the final iteration points to form the optimized tricuspid valve annulus curve, the annulus point set is smoothed, and the smoothed curve is resampled with equal arc length to ensure that the annulus points are evenly distributed.

7. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to claim 6, characterized in that, Step 5.3 specifically includes the following steps: 5.3.1: Calculate the current point The nearest point on the leaflet surface This ensures that the path converges to the leaflet surface; 5.3.2: Obtaining Local Normals ; 5.3.3: According to the formula Perform iterative updates; 5.3.4: Determine the new location If the device is still within the leaflet structure, continue iterating. If it exceeds the leaflet region or the mask value is below a set threshold, stop iterating and record the current state. As the final location point; 5.3.5 If the number of iterations reaches the maximum value If the condition is abnormal, the process will be forcibly stopped to prevent abnormal situations from affecting the optimization process.

8. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to claim 1, characterized in that, In step 6, the least squares method is used to fit a plane equation to the set of lobe loop points obtained in step 5. The normalized normal vector is obtained by using singular value decomposition or covariance matrix methods. .

9. The method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle according to any one of claims 1-8, characterized in that, In step 7, the angle between the fitting plane normal vector and the horizontal plane is calculated as follows: The calculation results are converted into angles to reflect the tilt of the annular plane.

10. A system for implementing the method for precise positioning of the tricuspid valve annulus and automatic calculation of the plane tilt angle as described in any one of claims 1-9, characterized in that, include: The medical image data processing module is used to acquire and preprocess three-dimensional medical image data, and to remove noise, normalize and register CT, MRI or echocardiogram data to improve data quality. The anatomical structure segmentation module is used to automatically segment the right atrium, anterior tricuspid valve leaflet, posterior tricuspid valve leaflet, septal tricuspid valve leaflet, and right ventricle based on a deep learning model, and optimizes the segmentation results by combining morphological operations and connected component analysis. The key point extraction module is used to extract leaflet feature points based on the leaflet anatomical structure and identify key points at the leaflet junction through distance measurement and anatomical location information. The valve loop construction module is used to generate valve loop curves based on key point distribution, and improve the continuity and uniformity of valve loops through path optimization and resampling processing. The valve annulus optimization module is used to adjust the spatial distribution of the valve annulus point set based on the surface features of the valve leaflet, and to use smoothing and optimization strategies to ensure that the valve annulus morphology conforms to the anatomical structure of the valve leaflet. The Lobe Ring Plane Calculation Module is used to fit the Lobe Ring Plane based on the optimized Lobe Ring Point Set and calculate its normal vector. The tilt angle calculation module is used to calculate the angle between the normal vector of the valve ring plane and the horizontal plane, and to determine the spatial tilt of the valve ring.

Images