Aortic morphology feature automatic measurement and auxiliary decision method and system based on artificial intelligence

Abstract

The application provides an aorta morphological feature automatic measurement and clinical auxiliary decision-making method and system based on artificial intelligence, which is suitable for three-dimensional medical image analysis of aorta and its main branches. The method comprises the following steps: image data acquisition and preprocessing, automatic blood vessel structure segmentation, center line extraction and key point positioning, multi-parameter automatic measurement, risk assessment and clinical auxiliary decision-making, data standardization output and system feedback optimization. The corresponding system comprises an image acquisition and processing module, a segmentation module, a center line extraction module, a parameter measurement module, an auxiliary decision-making module, a data output module and a verification and optimization module. Through a deep learning model and morphological calculation means, efficient identification and quantitative analysis of blood vessel structure are realized, and individualized risk suggestions and surgical planning are output based on statistical and machine learning models. The application improves the automation, standardization and intelligence level of blood vessel analysis, and enhances the clinical auxiliary decision-making ability under complex cases.

Description

Technical Field

[0001] This invention relates to the fields of medical image processing and artificial intelligence (AI), specifically to a method and system for automatic measurement and clinical decision support of aortic and related vascular morphological features based on artificial intelligence. Technical Background

[0002] In the diagnosis and treatment of cardiovascular diseases, accurate measurement of the morphology of the aorta and its major branches is of great significance for clinical decision-making. For example, in the development of treatment plans for aortic aneurysms, aortic dissections, and complex interventional procedures (such as transcatheter aortic valve replacement, TAVR), accurate information on vessel diameter, curvature, bifurcation angle, and calcification distribution helps physicians assess the difficulty of the surgery, develop personalized surgical approaches, and predict the risk of postoperative complications. However, current technologies often focus only on the automated extraction of single indicators (such as curvature), lacking systematic analysis and comprehensive risk assessment of the morphological characteristics of the aorta as a whole and its multiple branches.

[0003] Currently, manual measurement methods are not only inefficient but also easily influenced by the operator's experience, leading to subjective errors in the measurement results. Furthermore, traditional automated measurement algorithms mostly target only a segment of the aorta or specific morphological parameters, failing to meet the clinical needs for multi-dimensional comprehensive assessment. With the rapid development of deep learning in medical image processing, how to integrate segmentation, centerline extraction, key point localization, multi-index measurement, and clinical decision analysis into a unified "automated measurement + intelligent risk assessment" system has become an urgent technical problem to be solved.

[0004] Based on this, the present invention proposes a method and system that takes automatic multi-structure segmentation as the core, combines centerline extraction and automatic multi-parameter measurement, and finally outputs risk assessment and treatment suggestions through an auxiliary decision analysis module, so as to realize intelligent analysis of the aorta and its main branches throughout the entire process. Summary of the Invention

[0005] This invention provides an artificial intelligence-based method for automatic measurement and decision support of aortic morphological features, overcoming the problems of traditional methods that focus only on a single indicator, rely on manual operation, have insufficient measurement accuracy, and are inefficient. Through automatic segmentation and precise measurement, this invention can extract multi-dimensional vascular morphological information of the aorta and its major branches, and, combined with clinical databases or risk scoring models, provide clinicians with data-driven decision support. The method includes the following steps:

[0006] Step 1: 3D medical image data acquisition and preprocessing, acquiring medical image data containing the complete aorta and related blood vessels, and optimizing the image quality;

[0007] Step 2: AI segmentation of multiple structures in the aortic region. Based on an artificial intelligence model, the aorta and its main branch vessels are automatically identified, and key anatomical structures are extracted.

[0008] Step 3: Extraction of the target vessel centerline. Using computer vision and geometric analysis methods, the vessel centerline is extracted, and the morphological features of the vessel structure are calculated to support subsequent parameter measurement and analysis.

[0009] Step 4: Multi-parameter automated measurement, based on the vessel centerline and segmentation results, quantifies the morphological characteristics of the vessels, providing multi-dimensional data support for clinical analysis;

[0010] Step 5: Decision support analysis. Combining statistical modeling and machine learning methods, analyze vascular parameters, construct a disease risk assessment model, and provide clinical decision support including preoperative planning, surgical plan selection, and interventional treatment risk prediction.

[0011] Step 6: Data standardization and output, standardized storage of measurement results and analysis data, generation of structured diagnostic reports, and provision of data interfaces to support integration with medical information systems;

[0012] Step 7: System Validation and Feedback Optimization. Validate the applicability of the system in different devices, patient groups and lesion types through multi-source data, and improve the system's generalization ability and clinical reliability based on feedback optimization algorithms.

[0013] Furthermore, in step 1, the three-dimensional medical image data includes medical images that can completely cover the aortic arch and its main branches, specifically including CT and MRI images; preprocessing operations are performed on the original images, including noise reduction, standardization, contrast enhancement and resampling; while ensuring that key anatomical information is completely preserved, the data resolution and spatial orientation are unified to provide standardized input for subsequent segmentation and measurement.

[0014] Furthermore, in step 2, the aorta and its main branches are automatically segmented using a deep learning-based segmentation model, including the mesenteric artery, left and right subclavian arteries, left and right common carotid arteries, left and right renal arteries, brachiocephalic trunk, celiac trunk, and left and right common iliac arteries. Simultaneously, other key structures related to the aorta are identified and measured, including aortic calcification, left ventricle, and femoral head, to provide complete background information for subsequent quantitative analysis. For the segmentation results, boundary smoothing is performed using morphological optimization and connected component analysis methods to ensure the continuity and integrity of structural boundaries.

[0015] Furthermore, in step 3, centerline extraction and key point localization include the following steps:

[0016] Step 3.1: Determine the analysis target, integrate the segmentation results of the aorta and the target vessel, and clarify the centerline extraction area;

[0017] Step 3.2: Perform skeletonization on the segmented 3D mask of blood vessels and initially extract the centerline point set of blood vessels;

[0018] Step 3.3: Based on the preset start and end points, use a path optimization algorithm to extract the optimal centerline and remove noise and irrelevant branches;

[0019] Step 3.4: The centerline is smoothed by Gaussian filtering, and the point set is uniformly resampled based on the cumulative arc length interpolation method. At the same time, key points, including local curvature and bifurcation positions, are automatically identified to provide basic data for subsequent multi-parameter measurements.

[0020] Furthermore, in step 4, the multi-parameter automated measurement includes automatically calculating multiple morphological parameters of the blood vessel based on the centerline and three-dimensional segmentation results, including:

[0021] a) Local curvature calculation and identification of the maximum curvature point

[0022] Using the resampled centerline point set, the velocity and acceleration vectors of each point are calculated. Based on differential geometry formulas and the relationship between the magnitudes of the velocity and acceleration vectors, the curvature values ​​at each point are determined, forming a curvature sequence. The calculation formula is as follows:

[0023]

[0024] Where v i It is the velocity vector of the centerline point, a i κ is the acceleration vector of the centerline point, and κ is the curvature value sequence. Local maxima are identified by a one-dimensional extremum detection method. Several points with the largest curvature are extracted and marked as possible locations of the maximum curvature of the aortic arch or other local curvature peaks.

[0025] b) Calculation of blood vessel diameter and rate of change

[0026] For each point on the centerline, extract the tangent vector t perpendicular to the centerline. k Obtain the cross-section of the blood vessel, acquire the vessel surface region, and calculate the effective diameter D corresponding to that cross-section. k :

[0027] D k =2×R k

[0028] Further calculate the diameter change rate ΔD k :

[0029]

[0030] The diameter change rate is used to detect abnormal dilation or narrowing of blood vessels.

[0031] c) Calculation of vascular torsion angle and bifurcation angle

[0032] Record the adjacent tangential vectors t on the center line k and t k+1 The angle change measures the degree of vascular torsion; at the vascular bifurcation, the angle between the bifurcation vectors v1 and v2 is calculated respectively:

[0033]

[0034] The resulting angle reflects the orientation and connection angle of the aortic branches and provides morphological basis for preoperative planning.

[0035] d) Assessment of the degree of aortic calcification

[0036] Segmentation mask based on blood vessel wall, and calcification mask M calcium Perform intersection statistics to calculate the number of calcified voxels or the volume of calcified cavities:

[0037]

[0038] Further calculations were made of the distribution proportion of calcifications in locations including the ascending aorta, aortic arch, and descending aorta to assess the risk of local intervention. The above measurements were then combined with cardiac function data such as the left ventricle to form a more comprehensive quantitative analysis of vascular morphology.

[0039] Furthermore, in step 5, the specific method for assisting decision analysis is as follows:

[0040] Step 5.1: Data standardization and feature engineering. The extracted multidimensional vascular morphology indicators are standardized, including normalization or minimum-maximum scaling, to eliminate dimensional differences between different parameters. Cluster analysis, principal component analysis or random forest methods are then used to reduce the dimensionality or screen features of the indicators to extract key parameters that are of high importance to risk assessment and decision support.

[0041] Step 5.2: Risk scoring and clinical decision modeling. The processed indicators are input into a pre-established clinical regression or classification model, which includes Logistic regression, decision tree and XGBoost, to calculate the aortic-related risk score or decision index and compare it with the set reference threshold to generate quantitative assessment results of surgical difficulty, interventional risk or disease grade.

[0042] Step 5.3: Decision Recommendation and Result Visualization. Based on the risk assessment results, the system generates recommendations on the necessity of interventional treatment, the timing of surgery, and the planning of surgical pathways. These recommendations are presented in a visual manner, including an interactive 3D model and a spatial distribution map of key parameters, which helps clinicians to intuitively and efficiently understand the specific patient's condition and supports individualized treatment decisions.

[0043] Furthermore, in step 6, the various morphological indicators and decision results generated in the preceding steps are managed in a unified manner, a structured diagnostic report is output, and an open data interface is provided to support access to external analysis platforms, thereby enabling the sharing and reuse of medical data.

[0044] Furthermore, in step 7, based on test data from multiple centers and different brands of imaging equipment, the stability and generalization ability of the algorithm under different variable conditions are verified. The variables include scan resolution, patient body size, lesion location and lesion severity. Feedback from doctors and patients is collected, and continuous optimization is performed on segmentation accuracy, curvature measurement accuracy and risk prediction effect. While meeting medical compliance and regulatory requirements, the algorithm is continuously iterated and optimized to further improve clinical applicability and intelligence.

[0045] This invention also proposes an automatic measurement and clinical decision support system for aortic morphological features based on artificial intelligence, comprising:

[0046] The medical image data acquisition and preprocessing module is used to acquire three-dimensional medical image data containing the complete aorta and related blood vessels, and to optimize the image quality.

[0047] The automatic vascular structure segmentation module is used to automatically identify and segment the aorta and its major branch vessels based on an artificial intelligence model, and extract key anatomical structures.

[0048] The vascular centerline extraction and key point recognition module is used to extract the vascular centerline and calculate the morphological features of the vascular structure through computer vision and geometric analysis methods, supporting subsequent parameter measurement and analysis.

[0049] The multi-parameter automatic measurement module is used to quantify vascular morphological features based on the vascular centerline and segmentation results, providing multi-dimensional data support for clinical analysis.

[0050] The auxiliary decision analysis module is used to combine statistical modeling and machine learning methods to analyze vascular parameters, construct disease risk assessment models, and provide clinical auxiliary decision support including preoperative planning, surgical plan selection, and interventional treatment risk prediction.

[0051] The data storage and output module is used to standardize the storage of measurement results and analysis data, generate structured diagnostic reports, and provide data interfaces to support integration with medical information systems.

[0052] The system validation and feedback optimization module is used to validate the applicability of the system in different devices, patient groups and lesion types through multi-source data, and improve the system's generalization ability and clinical reliability based on feedback optimization algorithms.

[0053] Technical effect

[0054] This invention provides a general-purpose vascular morphology analysis platform based on deep learning and multi-index automated measurement technology. It can not only automatically analyze aortic structures but also be flexibly applied to various other vascular combinations, achieving efficient automatic extraction and quantitative analysis of multiple morphological indicators. Compared with existing technologies, this invention has the following significant technical advantages:

[0055] 1. Wide range of applications and strong versatility: This invention is not limited to a single vascular structure, but is applicable to the morphological analysis of various vascular structures and their combinations. It constructs a universal vascular analysis framework, which is applicable to the morphological assessment of various vascular structures, thereby improving the adaptability and application scope of the method.

[0056] 2. Significantly improved measurement efficiency and accuracy: Through deep learning models and automated processing, rapid segmentation of vascular structures, centerline extraction, and multi-parameter measurement are achieved, reducing human intervention and operation, and improving measurement consistency and accuracy in clinical procedures. Simultaneously, this method can accurately identify key morphological features, such as points of maximum local curvature and vascular bifurcation angles, avoiding subjective errors caused by manual measurement.

[0057] 3. Standardized and highly scalable data output: This system can not only automatically calculate and output various morphological indicators, but also provide standardized data reports and open data interfaces, supporting seamless integration with medical systems and scientific research platforms, building personalized diagnostic and prognostic prediction models, and improving data sharing efficiency and system integration capabilities.

[0058] 4. Enhanced Clinical Decision-Making and Research Capabilities: Leveraging the comprehensive and precise morphological quantitative data provided by this invention, clinicians and researchers can more easily conduct risk stratification, disease prognosis assessment, and treatment planning. This invention provides rich foundational data support for the subsequent construction of accurate prognostic prediction models and risk scoring systems, thereby promoting the precision and objectivity of clinical diagnosis and treatment, and advancing the development of cardiovascular disease research. Attached Figure Description

[0059] Figure 1 This is a flowchart of an artificial intelligence-based method for automatic measurement and decision support of aortic morphological features.

[0060] Figure 2 This is a schematic diagram of the segmentation results of the complete aorta and its main branch vessels described in step 2.

[0061] Figure 3 A is a schematic diagram of the centerline extraction results from the aortic root to the abdominal aortic segment described in step 3.

[0062] Figure 3 B is a schematic diagram of the centerline extraction results from the aortic root to the brachiocephalic trunk as described in step 3.

[0063] Figure 4 This is a schematic diagram of the interface of the decision support analysis module described in step 5.

[0064] Figure 5 This is a flowchart illustrating the standardized data output and iterative model optimization process within the system. Detailed Implementation

[0065] 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. For those skilled in the art, equivalent substitutions or modifications made without departing from the core idea of ​​the present invention should be covered within the scope of protection of the present invention.

[0066] This invention proposes an automatic measurement and decision support system for aortic morphological features based on artificial intelligence. The overall process is as follows: Figure 1 As shown, it includes the following steps:

[0067] Step 1: 3D Medical Image Data Acquisition and Preprocessing. In this step, 3D medical image data covering the aorta and related vascular structures is acquired. This 3D medical image data can be derived from various medical imaging techniques, including computed tomography (CT), computed tomography angiography (CTA), and magnetic resonance imaging (MRI). To ensure the accuracy of subsequent analysis, the imaging results need to cover all relevant structures of the patient's heart. Furthermore, the image data needs to undergo preprocessing, including noise reduction and standardization, to ensure data quality. These processing steps can be implemented using medical image processing software.

[0068] Furthermore, to improve image quality and computational accuracy, the acquired raw images need to undergo preprocessing operations, including denoising, intensity normalization, and spatial resampling, to unify data resolution, reduce artifact interference, and enhance image edge sharpness. Unifying data resolution and spatial orientation, while ensuring the complete preservation of key anatomical information, lays the foundation for subsequent segmentation and measurement.

[0069] Step 2: AI Segmentation of the Aortic Region. This step, based on the medical images obtained in Step 1, utilizes an artificial intelligence segmentation model to automatically identify and segment multiple structures within the aortic region. Specifically, this includes segmenting major vascular structures such as the intact aortic arch, renal arteries, subclavian artery, and common carotid artery, as well as identifying and segmenting related auxiliary structures such as calcified areas of the aorta and the left ventricle. The segmentation results are shown below. Figure 2 As shown, the specific steps include the following:

[0070] Step 2.1: Input the downsampled CTA medical image I(x,y,z) into a deep learning-based segmentation network to achieve automatic segmentation of the aortic region and output a binarized aortic mask M(x,y,z). The segmentation model can adopt advanced network architectures such as SAM2 and MedNeXt, and is trained and optimized using labeled data to ensure the anatomical accuracy and consistency of the segmentation results. The basic segmentation relationships are as follows:

[0071] M(x,y,z)=Segmentation(I(x,y,z))

[0072]

[0073] The system can simultaneously output mask images corresponding to multiple structures, including:

[0074] Aortic Mask M Cartery (x,y,z), calcification region mask M calcium (x,y,z),

[0075] Carotid branch mask M aorat (x,y,z)...

[0076] Where I(x,y,z) is the original medical image (CTA), and M(x,y,z) is the segmented blood vessel mask.

[0077] Step 2.2: Further generate an accurate label map of the aortic region from the segmentation results, and combine image morphology optimization and connected component analysis algorithms to smooth the segmentation boundaries to ensure the integrity of each structure and the continuity and completeness of the boundaries.

[0078] Step 3: Target vessel centerline extraction and key point localization. This step is used to extract the centerline of the aorta and its branches based on the segmentation results, and to locate geometric key points. An example of the structural process is shown below. Figure 3 A and Figure 3 As shown in B, the specific sub-steps include:

[0079] Step 3.1: Tag merging for the target blood vessel

[0080] The segmentation masks of the aorta and its target branch are merged to obtain the composite mask image M′(x′,y′,z′), which is represented as follows:

[0081] M′(x′,y′,z′)=M aorat (x,y,z)∪M target (x,y,z)

[0082] Step 3.2: Skeletonization and Preliminary Centerline Extraction

[0083] The segmented aortic mask is skeletonized, a preliminary centerline point set is extracted, and redundant branches are removed. The optimization objective is to find the continuous line segment with the minimum path weight within the mask region, as shown below:

[0084]

[0085] Among them, C raw It is the set of centerline points obtained from the initial skeletonization.

[0086] Step 3.3: Optimal Centerline Path Optimization

[0087] Considering potential pseudo-branches and curve offsets in the skeleton path, a cost function is used to optimize the centerline path. Let:

[0088] The distance accumulation term is:

[0089] d(x′,y′,z′)=d(x′ prev ,y′ prev ,z′ prev )+||(x′,y′,z′)-(x′ prev ,y′ prev ,z′ prev )||

[0090] The normalized image intensity is:

[0091]

[0092] The overall cost function is:

[0093] W(x′,y′,z′)=d(x′,y′,z′)+α·(1-I′(x′,y′,z′))

[0094]

[0095] Among them, C * This is the final optimized aortic centerline.

[0096] Step 3.4: Centerline smoothing and uniform resampling

[0097] Gaussian filtering is applied to the centerline coordinate sequence, and cumulative arc length interpolation is used to resample the centerline to a fixed point spacing, ensuring that the spacing between adjacent points is uniformly 1 mm. The specific kernel function is expressed as follows:

[0098]

[0099] s0=0,s i =s i-1 +||C smooth,i -C smooth,i-1 ||

[0100] s j ′=j×1mm,j=0,1,2,...,M

[0101] C resampled,j =Interp(C smooth ,s j ′)

[0102] Where s i It is the cumulative length of the centerline, Interp(·) represents interpolation along the arc length, C resampled This represents the center line for resampling (1mm spacing between adjacent points). Along this uniformly resampled center line, key locations such as bifurcation points and localized abnormal bends can be automatically detected.

[0103] Step 4: Automated Multi-Parameter Measurement. This step, based on the centerline coordinate sequence obtained in Step 3, measures multiple morphological parameters of the aorta and its branches. Specifically, it includes the following sub-steps:

[0104] Step 4.1: Curvature Calculation and Identification of the Maximum Curvature Point

[0105] Based on the centerline point set obtained by resampling in step 3, calculate the velocity vector (obtained from the position difference between adjacent points) and acceleration vector of each point. According to the differential geometry formula, use the relationship between the magnitudes of the velocity and acceleration vectors to calculate the curvature value at each point, thus forming a complete centerline curvature sequence.

[0106] For the i-th centerline point, the velocity vector is represented as:

[0107]

[0108] The acceleration vector is represented as:

[0109]

[0110] The formula for calculating curvature value is:

[0111]

[0112] Among them, κi κ is the curvature value of the i-th point on the centerline, and κ is the curvature value sequence. Several points with the largest curvature are obtained through one-dimensional extremum detection (such as finding local maxima) and marked as possible locations of maximum aortic arch curvature or other local curvature peak locations.

[0113] Step 4.2: Calculation of vessel diameter and rate of change

[0114] For each point on the centerline, extract the tangent vector t perpendicular to the centerline. k The cross-section is obtained to acquire the corresponding blood vessel region, and the effective diameter D corresponding to this cross-section is calculated. k The calculation formula is:

[0115] D k =2×R k

[0116] Furthermore, the rate of change of diameter ΔD between adjacent sections is calculated. k Used to assess local trends in changes in blood vessel diameter:

[0117]

[0118] The rate of change in diameter can be used to identify abnormal dilation or narrowing of blood vessels.

[0119] Step 4.3: Calculation of vascular torsion angle and bifurcation angle

[0120] By recording the adjacent tangential vectors t on the center line k and t k+1 The angle change measures the degree of torsion of the blood vessel; at the bifurcation point, the angle between the bifurcation vectors v1 and v2 is calculated, expressed as:

[0121]

[0122] The obtained angle θ can reflect the direction and connection angle of the aortic branches, which can help with preoperative planning.

[0123] Step 4.4: Quantitative analysis of calcified areas

[0124] Based on the segmentation mask of the blood vessel wall, and the calcification mask M calcium Perform intersection statistics to calculate the number of calcified voxels or the volume of calcified cavities:

[0125]

[0126] Furthermore, based on spatial location information, the number of voxels in the calcified area in the ascending aorta, aortic arch, and descending aorta can be counted separately, and their proportions can be calculated to help assess the local interventional risk of each site.

[0127] Step 5: Decision support analysis module, such as... Figure 4 As shown, the decision support analysis module is used to conduct risk assessment and clinical auxiliary judgment for patients based on multiple morphological parameters and statistical models, and specifically includes the following sub-steps:

[0128] Step 5.1: Data Standardization and Feature Selection

[0129] Multiple variables, including peak curvature, maximum vessel diameter, torsion angle, bifurcation angle, calcification volume, and other auxiliary indicators (such as ventricular ejection fraction), are combined into a feature vector.

[0130] x = [x1, x2, ..., x n ];

[0131] The feature vector is processed to remove outliers and normalize it to meet the input requirements of subsequent classification or regression models, ensuring that each indicator has a uniform dimension and numerical range.

[0132] Step 5.2: Risk Assessment Model and Results Output

[0133] Combining known data from multiple centers, models such as Logistic Regression or XGBoost are employed. The risk of a patient's target blood vessel is predicted as follows:

[0134]

[0135] Where r represents the predicted risk score. If r exceeds the set high-risk threshold, the system will prompt that there may be outliers and the weights of each indicator, and output an intuitive prompt.

[0136] Step 5.3: Visualization and Interactive Decision Support

[0137] The system displays the segmentation results, centerline, and key parameter distribution of the aorta and its major branches in a 3D visualization interface, with high-risk areas marked by different colors or icons. Based on the model evaluation results, the system can generate risk assessment level prompts and allows users to fill in more patient characteristic information (such as basic medical history and physiological parameters) to adjust the weights of the risk model, assisting doctors in making further clinical judgments and individualized treatment plans.

[0138] Step 6: Data standardization and output, such as... Figure 5As shown, this step integrates the morphological quantitative indicators, risk scoring results, and 3D visualization results obtained from the preceding steps into a structured PDF or online report. Simultaneously, the system supports data output based on a RESTful API interface or DICOM SR (structured report) format, enabling seamless integration with hospital information systems (HIS), image archiving and communication systems (PACS), or third-party research data platforms, meeting the needs of clinical applications across multiple scenarios.

[0139] Step 7: System verification and feedback optimization, such as... Figure 5 As shown, to ensure the applicability and stability of the system in actual clinical environments, this invention also includes a system verification and iterative optimization process, specifically including the following sub-steps:

[0140] Step 7.1: Multicenter Clinical Trial

[0141] Clinical trials were conducted in hospitals from different regions using different brands of CT equipment. The segmentation accuracy and measurement stability under different imaging parameters (such as low-dose / high-dose scanning, different contrast agent concentrations) were evaluated to verify its generalization ability across devices and populations.

[0142] Step 7.2: Feedback Collection and Continuous Optimization

[0143] Regularly collect clinicians' subjective assessments or postoperative verification results on the accuracy of the system's quantitative results and risk scores; update the segmentation model and algorithm parameters based on feedback information to improve the suitability for complex cases (such as severe calcification and variant anatomical structures).

[0144] Step 7.3: Compliance Assurance and Functionality Expansion

[0145] The system development and updates strictly adhere to the relevant regulatory requirements for medical software. After each version iteration, stability and compliance must be reconfirmed through a testing and verification process. Once the main functions are verified and perfected, the system can be extended to other large blood vessels (such as the carotid artery and lower limb arteries) or special cardiovascular structures to provide auxiliary support for related interventional treatments.

Claims

1. A method for automatic measurement and decision support of aortic morphological features based on artificial intelligence, characterized in that, Includes the following steps: Step 1: 3D medical image data acquisition and preprocessing, acquiring medical image data containing the complete aorta and related blood vessels, and optimizing the image quality; Step 2: AI segmentation of multiple structures in the aortic region. Based on an artificial intelligence model, the aorta and its major branches are automatically identified, and key anatomical structures are extracted. Specifically, a deep learning-based segmentation model is used to automatically segment the aorta and its major branches, including the mesenteric artery, left and right subclavian arteries, left and right common carotid arteries, left and right renal arteries, brachiocephalic trunk, celiac trunk, and left and right common iliac arteries. Simultaneously, other key structures related to the aorta are identified and segmented, including aortic calcification, the left ventricle, and the femoral head. For the segmentation results, morphological optimization and connected component analysis methods are used for boundary smoothing to ensure the continuity and integrity of structural boundaries. Step 3: Target vessel centerline extraction. Using computer vision and geometric analysis methods, the vessel centerline is extracted, and the morphological features of the vessel structure are calculated to support subsequent parameter measurement and analysis. This includes: determining the analysis target; integrating the segmentation results of the aorta and the target vessel to clarify the centerline extraction area; performing skeletonization on the segmented 3D mask of the vessel to initially extract the vessel centerline point set; extracting the optimal centerline based on preset start and end points using a path optimization algorithm, and removing noise and irrelevant branches; smoothing the centerline using Gaussian filtering, and uniformly resampling the point set based on cumulative arc length interpolation, while automatically identifying key points, including local curvature and bifurcation locations, to provide basic data for subsequent multi-parameter measurements. Step 4: Multi-parameter automated measurement. Based on the vessel centerline and segmentation results, quantify the morphological features of the vessels to provide multi-dimensional data support for clinical analysis. This multi-parameter automated measurement includes: automatically calculating multiple morphological parameters of the vessels based on the centerline and three-dimensional segmentation results, including assessing the degree of calcification. Specifically, this involves using a segmentation mask of the vessel wall and a calcification mask. Perform intersection statistics to calculate the number of calcified voxels or the volume of calcified cavities: Further calculations were made of the distribution proportion of calcifications in locations including the ascending aorta, aortic arch, and descending aorta to assess the risk of local intervention; the above measurements were combined with left ventricular cardiac function data to form a more comprehensive quantitative analysis of vascular morphology. Step 5: Decision support analysis. Combining statistical modeling and machine learning methods, analyze vascular parameters, construct a disease risk assessment model, and provide clinical decision support including preoperative planning, surgical plan selection, and interventional treatment risk prediction. Step 6: Data standardization and output, standardized storage of measurement results and analysis data, generation of structured diagnostic reports, and provision of data interfaces to support integration with medical information systems; Step 7: System Validation and Feedback Optimization. Validate the applicability of the system in different devices, patient groups and lesion types through multi-source data, and improve the system's generalization ability and clinical reliability based on feedback optimization algorithms.

2. The automatic measurement and decision support method for aortic morphological features according to claim 1, wherein in step 1, the three-dimensional medical image data includes medical images that can completely cover the aortic arch and its main branches, specifically including CT and MRI images; preprocessing operations are performed on the original images, including noise reduction, standardization, contrast enhancement and resampling; while ensuring the complete preservation of key anatomical information, the data resolution and spatial orientation are unified to provide standardized input for subsequent segmentation and measurement.

3. The automatic measurement and decision support method for aortic morphological features according to claim 2, wherein step 4, based on the centerline and three-dimensional segmentation results, further includes automatically calculating multiple morphological parameters of the blood vessel: a) Local curvature calculation and identification of the maximum curvature point Using the resampled centerline point set, the velocity and acceleration vectors of each point are calculated. Based on differential geometry formulas and the relationship between the magnitudes of the velocity and acceleration vectors, the curvature values ​​at each point are determined, forming a curvature sequence. The calculation formula is as follows: in It is the velocity vector of the centerline point. It is the acceleration vector of the centerline point. It is a curvature value sequence. Local maxima are identified by a one-dimensional extreme value detection method. Several points with the largest curvature are extracted and marked as possible locations of the maximum curvature of the aortic arch or other local curvature peak locations. b) Calculation of blood vessel diameter and rate of change For each point on the center line, extract the tangent vector perpendicular to the center line. Obtain the cross-section of the blood vessel, acquire the vessel surface region, and calculate the effective diameter corresponding to that cross-section. : Further calculation of diameter change rate : The diameter change rate is used to detect abnormal dilation or stenosis of blood vessels; c) Calculation of vascular torsion angle and bifurcation angle Record adjacent tangential vectors on the center line and The angle change is used to measure the degree of vascular torsion; at the vascular bifurcation points, the bifurcation vector is calculated respectively. and The included angle: The resulting angle reflects the orientation and connection angle of the aortic branches and provides morphological basis for preoperative planning.

4. The automatic measurement and decision support method for aortic morphological characteristics according to claim 3, wherein the specific method for decision support analysis in step 5 is as follows: Step 5.1: Data standardization and feature engineering. The extracted multidimensional vascular morphology indicators are standardized, including normalization or minimum-maximum scaling, to eliminate dimensional differences between different parameters. Cluster analysis, principal component analysis or random forest methods are then used to reduce the dimensionality or screen features of the indicators to extract key parameters that are of high importance to risk assessment and decision support. Step 5.2: Risk scoring and clinical decision modeling. The processed indicators are input into a pre-established clinical regression or classification model, which includes Logistic regression, decision tree and XGBoost. The risk score or decision index related to the aorta is calculated and compared with the set reference threshold to generate quantitative assessment results of surgical difficulty, interventional risk or disease grade. Step 5.3: Decision Recommendation and Result Visualization. Based on the risk assessment results, the system generates recommendations on the necessity of interventional treatment, the timing of surgery, and the planning of surgical pathways. These recommendations are presented in a visual manner, including a three-dimensional interactive model and a spatial distribution map of key parameters, which facilitates clinicians' intuitive and efficient understanding of the patient's specific condition and supports individualized treatment decisions.

5. The automatic measurement and decision support method for aortic morphological characteristics according to claim 4, in step 6, the various morphological indicators and decision results generated in the preceding steps are uniformly managed, a structured diagnostic report is output, and an open data interface is provided to support the access of external analysis platforms, so as to realize the sharing and reuse of medical data.

6. The automatic measurement and decision support method for aortic morphological features according to any one of claims 1-5, in step 7, based on test data from multiple centers and different brands of imaging equipment, the stability and generalization ability of the algorithm under different variable conditions are verified, the variables including scan resolution, patient body size, lesion location and lesion severity; feedback from doctors and patients is collected, and continuous optimization is performed on segmentation accuracy, curvature measurement accuracy, and risk prediction effect; while meeting medical compliance and regulatory requirements, the algorithm is continuously iterated and optimized to further improve clinical applicability and intelligence level.

7. An AI-based automatic measurement and clinical decision support system for aortic morphological features, implementing the method of any one of claims 1-6, comprising: The medical image data acquisition and preprocessing module is used to acquire three-dimensional medical image data containing the complete aorta and related blood vessels, and to optimize the image quality. The automatic vascular structure segmentation module is used to automatically identify and segment the aorta and its major branch vessels based on an artificial intelligence model, and extract key anatomical structures. The vascular centerline extraction and key point recognition module is used to extract the vascular centerline and calculate the morphological features of the vascular structure through computer vision and geometric analysis methods, supporting subsequent parameter measurement and analysis. The multi-parameter automatic measurement module is used to quantify vascular morphological features based on the vascular centerline and segmentation results, providing multi-dimensional data support for clinical analysis. The auxiliary decision analysis module is used to combine statistical modeling and machine learning methods to analyze vascular parameters, construct disease risk assessment models, and provide clinical auxiliary decision support including preoperative planning, surgical plan selection, and interventional treatment risk prediction. The data storage and output module is used to standardize the storage of measurement results and analysis data, generate structured diagnostic reports, and provide data interfaces to support integration with medical information systems. The system validation and feedback optimization module is used to validate the applicability of the system in different devices, patient groups and lesion types through multi-source data, and improve the system's generalization ability and clinical reliability based on feedback optimization algorithms.

Images