Method and system for processing and analyzing vascular images by combining magnetic resonance angiography and vessel wall imaging

Abstract

The present invention provides a vascular image processing and analysis method and system for combined magnetic resonance vascular and vascular wall imaging, the system comprising: a data acquisition module for acquiring vascular enhanced images and vascular wall images; an image registration module for registering the vascular wall image to the vascular enhanced image to obtain the registered vascular wall image; a segmentation module for obtaining the vascular lumen segmentation result and the vascular wall segmentation result; a fusion module for weighted fusion of the vascular enhanced image and the registered vascular wall image to obtain a fused image; an analysis module for calculating the intersection area of the vascular lumen segmentation result and the vascular wall segmentation result; a three-dimensional visualization and surface reconstruction module for performing three-dimensional skeleton extraction and surface reconstruction on the three-dimensional connected domain of the suspected occluded blood vessel. The present invention can simultaneously obtain information on the vascular lumen and the vascular wall, making up for the deficiencies of traditional imaging methods in evaluating the distal vascular wall conditions of occluded blood vessels, and significantly shortening the time required for preoperative evaluation through automated processing and analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and medical image processing, and specifically relates to a method and system for processing and analyzing vascular images by combining magnetic resonance angiography and vessel wall imaging. Background Art

[0002] With the development of medical imaging technology, the application of artificial intelligence in medical image processing has become increasingly widespread; especially in the acute endovascular treatment of stroke, time is crucial, and preoperative evaluation needs to be as concise and efficient as possible; while in the elective interventional recanalization treatment of non-acute large artery occlusion, a detailed preoperative evaluation is particularly important, because it can help doctors comprehensively master the details of the patient's pathological mechanism, thereby improving the safety and efficiency of the surgery; multimodal imaging evaluation plays a key role in this process, which can enable doctors to obtain important information such as cerebral infarction and blood perfusion in the brain area supplied by the occluded artery of the patient, the location and extent of the occluded lesion and its cause, and whether there is a thrombus in the occluded segment of the blood vessel.

[0003] Deficiencies of the prior art:

[0004] 1. Limited information: In the case of occlusion of the large arteries in the head and neck, the information provided by conventional preoperative vascular evaluation is limited. Especially when the intracranial artery is occluded, it is often difficult for doctors to obtain detailed information about the occluded segment of the blood vessel.

[0005] 2. Single function: Current post-processing software mainly focuses on one aspect of lumen imaging or vessel wall imaging and cannot provide a comprehensive evaluation tool.

[0006] 3. Manual operation: Existing image processing methods often rely on manual segmentation and analysis, which is not only time-consuming and laborious, but also prone to introducing human errors.

[0007] 4. Lack of integration: There is a lack of a method or software that can integrate vascular lumen and vessel wall images for analysis, making the preoperative evaluation less comprehensive and accurate.

[0008] Therefore, there are deficiencies in the prior art and further improvements are needed. Summary of the Invention

[0009] In view of the problems existing in the prior art, the present invention provides a method and system for processing and analyzing vascular images by combining magnetic resonance angiography and vessel wall imaging.

[0010] To achieve the above object, the specific solutions of the present invention are as follows:

[0011] The present invention provides a system for processing and analyzing vascular images by combining magnetic resonance angiography and vessel wall imaging, and the system includes:

[0012] A data acquisition module, configured to acquire a vascular enhanced image A1 and a vessel wall image B1;

[0013] An image registration module, which realizes the registration between the vessel wall image and the vascular enhanced image, registers the vessel wall image B1 to the vascular enhanced image A1, and obtains a registered vessel wall image A2;

[0014] A segmentation module, which uses a deep learning algorithm to segment the vascular enhanced image A1 and the registered vessel wall image A2, and respectively obtains a vascular lumen segmentation result S1 and a vessel wall segmentation result S2;

[0015] A fusion module, which performs weighted fusion on the vascular enhanced image A1 and the registered vessel wall image A2 to obtain a fused image A3;

[0016] An analysis module, which calculates the intersection area S3 between the vascular lumen segmentation result S1 and the vessel wall segmentation result S2, and sets the label of the intersection voxels to the vascular lumen label;

[0017] Calculate the union S4 of the intersection area S3 and the vessel wall segmentation result S2;

[0018] Perform three-dimensional connected component analysis on S4 to identify occluded vascular regions, and classify and process them according to different connected component label distributions, remove false positive segmentation results, retain normal blood vessels, label stenotic blood vessels, and retain suspected occluded blood vessels;

[0019] A three-dimensional visualization and surface reconstruction module, which is used to perform three-dimensional skeleton extraction and surface reconstruction on the three-dimensional connected components of the suspected occluded blood vessels;

[0020] A storage module, which is used to store the original image data, intermediate processing results and final analysis results.

[0021] Further, the image registration module adopts rigid body registration to improve the registration accuracy.

[0022] Further, the segmentation module uses a pre-trained CNN model, and data augmentation technology is adopted during the training of the model to improve the segmentation accuracy.

[0023] Further, the fusion module further includes a denoising algorithm for further removing small-area noise points.

[0024] Further, the three-dimensional visualization and surface reconstruction module includes processing with the Laplacian operator for smoothing contour points, and obtaining more accurate central axis point coordinates through ellipse fitting.

[0025] The present invention also provides a method for processing and analyzing vascular images by combining magnetic resonance angiography and vessel wall imaging. Based on the above system, the method includes the following steps:

[0026] S101. Obtain the vascular enhanced image A1 and the vessel wall image B1;

[0027] S102. Register the vessel wall image B1 to the vascular enhanced image A1 by rigid registration to obtain the registered vessel wall image A2;

[0028] S103. Based on the deep learning method, by way of channel splicing, use the vascular enhanced image A1 and the registered vessel wall image A2 data as network inputs to obtain the vessel lumen segmentation result S1 and the vessel wall segmentation result S2;

[0029] S104. Calculate the intersection region S3 of the vessel lumen segmentation result S1 and the vessel wall segmentation result S2, and set the label of the intersection voxels to the vessel lumen label;

[0030] S105. Calculate the union S4 of the intersection region S3 and the vessel wall segmentation result S2, and perform three-dimensional connected component analysis on S4 to identify the occluded vessel region;

[0031] S106. For the three-dimensional connected component of the suspected occluded vessel, use the three-dimensional skeleton extraction method to calculate the initial central axis X1 of the connected component, and correct the central axis point coordinates X2;

[0032] S107. Generate a weighted fusion image A3 based on the vascular enhanced image A1 and the registered vessel wall image A2;

[0033] S108. Based on the corrected central axis point coordinates X2, extract the local region image from the weighted fusion image A3 for surface reconstruction.

[0034] Further, in step S102, the rigid registration is implemented using the iterative closest point algorithm.

[0035] Further, in step S103, the deep learning method uses a convolutional neural network, and data augmentation technology is used during network training.

[0036] Further, in step S104, the calculation of the intersection region S3 further includes removing small-area noise points based on morphological operations.

[0037] Further, in step S106, the three-dimensional skeleton extraction method includes traversing the connected component coordinate points along the central axis, and performing ellipse fitting on the masked contour points within a specified range to obtain the corrected central axis point coordinates X2.

[0038] Adopting the technical solution of the present invention has the following beneficial effects:

[0039] 1. Comprehensive acquisition of pathological information:

[0040] A method and system for combined magnetic resonance angiography and vessel wall imaging are provided, which can simultaneously obtain detailed information of the vessel lumen and the vessel wall, making up for the deficiencies of traditional imaging methods in evaluating the vessel wall condition of the distal occluded vessels.

[0041] 2. Automated segmentation and analysis:

[0042] Through deep learning methods, automatic segmentation of the vessel lumen and the vessel wall is achieved, improving the accuracy and efficiency of segmentation and reducing the need for manual intervention.

[0043] 3. 3D visualization and surface reconstruction:

[0044] 3D visualization and surface reconstruction of the vessel lumen and the vessel wall are realized, enabling doctors to visually observe the lesion degree and vessel morphology of the large vessels at the occlusion site, their distal vessels, and branch vessels, improving the accuracy of preoperative evaluation.

[0045] 4. Improving the diagnostic speed and accuracy:

[0046] Through automated processing and analysis, the time required for preoperative evaluation is significantly shortened, while the accuracy and reliability of the evaluation are improved, which is beneficial for formulating more reasonable treatment plans.

[0047] 5. Exhaustive connected component analysis:

[0048] By performing 3D connected component analysis on the segmentation results, different types of vascular lesions (such as false positive segmentation, normal vessels, stenotic vessels, suspected occluded vessels) can be distinguished, providing more detailed pathological information for doctors.

[0049] 6. Axis correction and local reconstruction:

[0050] The coordinates of the central axis points are corrected through 3D skeleton extraction methods, and surface reconstruction of the local area images is performed based on the corrected coordinates, providing a more refined analysis method for the occluded area and helping to formulate more accurate surgical plans.

[0051] 7. Integrated solution:

[0052] An integrated solution is provided, covering the entire process from data acquisition, image registration, segmentation, analysis to final 3D visualization and surface reconstruction, providing a comprehensive tool for clinical applications.

[0053] 8. Improving surgical safety and efficiency:

[0054] Through exhaustive preoperative evaluation, the uncertainty caused by insufficient information during the operation is reduced, thereby improving the safety and efficiency of the operation and reducing the surgical risk.

[0055] 9. Improved patient management:

[0056] More accurate preoperative assessment helps optimize patient management, ensuring that each patient receives the most suitable treatment plan for their specific situation, thus improving the overall quality of medical services. Brief Description of the Drawings

[0057] Figure 1 is a schematic diagram of the deep learning-based blood vessel lumen and vessel wall segmentation of the present invention;

[0058] Figure 2 is a three-dimensional visualization diagram of the blood vessel lumen and vessel wall segmentation results of the present invention;

[0059] Figure 3 is a surface reconstruction diagram of the suspected occlusion area of the present invention;

[0060] Figure 4 is a system block diagram of the present invention;

[0061] Figure 5 is the overall flowchart of the present invention. Detailed Description of the Embodiments

[0062] The present invention will be further described in detail below with reference to the drawings and embodiments; it can be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention; in addition, it should be noted that, for the sake of description, only parts related to the present invention are shown in the drawings rather than all of them.

[0063] Combined with Figures 1 - 5 As shown, the present invention provides a vascular image processing and analysis system combining magnetic resonance angiography and vessel wall imaging, and the system includes:

[0064] A data acquisition module for acquiring a vascular enhanced image A1 and a vessel wall image B1;

[0065] An image registration module for realizing the registration between the vessel wall image and the vascular enhanced image, registering the vessel wall image B1 to the vascular enhanced image A1 to obtain a registered vessel wall image A2;

[0066] A segmentation module for segmenting the vascular enhanced image A1 and the registered vessel wall image A2 using a deep learning algorithm to respectively obtain a blood vessel lumen segmentation result S1 and a vessel wall segmentation result S2;

[0067] A fusion module for performing weighted fusion on the vascular enhanced image A1 and the registered vessel wall image A2 to obtain a fused image A3;

[0068] An analysis module for calculating the intersection area S3 of the blood vessel lumen segmentation result S1 and the vessel wall segmentation result S2 and setting the label of the intersection voxels to the blood vessel lumen label;

[0069] Calculate the union S4 of the intersection region S3 and the blood vessel wall segmentation result S2;

[0070] Perform three-dimensional connected component analysis on S4 to identify occluded blood vessel regions, and classify and process them according to different connected component label distribution situations, remove false positive segmentation results, retain normal blood vessels, mark stenotic blood vessels, and retain suspected occluded blood vessels;

[0071] A three-dimensional visualization and surface reconstruction module, which is used to extract the three-dimensional skeleton and perform surface reconstruction on the three-dimensional connected component of the suspected occluded blood vessel;

[0072] A storage module, which is used to store the original image data, intermediate processing results, and final analysis results.

[0073] The image registration module adopts rigid registration to improve the registration accuracy.

[0074] The segmentation module uses a pre-trained CNN model, and data augmentation technology is adopted during the training of the model to improve the segmentation accuracy.

[0075] The fusion module also includes a denoising algorithm, which is used to further remove small-area noise points.

[0076] The three-dimensional visualization and surface reconstruction module includes processing with the Laplace operator for smoothing contour points, and obtaining more accurate central axis point coordinates through ellipse fitting.

[0077] The present invention also provides a method for processing and analyzing blood vessel images by combining magnetic resonance angiography and blood vessel wall imaging. Based on the above system, the method includes the following steps:

[0078] S101, obtain a blood vessel enhanced image A1 and a blood vessel wall image B1;

[0079] S102, register the blood vessel wall image B1 to the blood vessel enhanced image A1 by rigid registration to obtain the registered blood vessel wall image A2;

[0080] S103, based on a deep learning method, take the blood vessel enhanced image A1 and the registered blood vessel wall image A2 data as network inputs through channel splicing to obtain a blood vessel lumen segmentation result S1 and a blood vessel wall segmentation result S2;

[0081] S104, calculate the intersection region S3 of the blood vessel lumen segmentation result S1 and the blood vessel wall segmentation result S2, and set the label of the intersection voxels to the blood vessel lumen label;

[0082] S105, calculate the union S4 of the intersection region S3 and the blood vessel wall segmentation result S2, and perform three-dimensional connected component analysis on S4 to identify occluded blood vessel regions;

[0083] S106. For the three-dimensional connected domain of the suspected occluded blood vessel, use the three-dimensional skeleton extraction method to calculate the initial central axis X1 of the connected domain, and correct the central axis point coordinates X2.

[0084] S107. Based on the blood vessel enhanced image A1 and the registered blood vessel wall image A2, generate a weighted fusion image A3.

[0085] S108. Based on the corrected central axis point coordinates X2, extract the local region image from the weighted fusion image A3 and perform surface reconstruction.

[0086] In step S102, the rigid registration is implemented using the iterative closest point algorithm.

[0087] In step S103, the deep learning method uses a convolutional neural network, and data augmentation technology is used during network training.

[0088] In step S104, the calculation of the intersection region S3 also includes removing small-area noise points based on morphological operations.

[0089] In step S106, the three-dimensional skeleton extraction method includes traversing the connected domain coordinate points along the central axis and performing elliptical fitting on the masked contour points within a specified range to obtain the corrected central axis point coordinates X2.

[0090] The system of the present invention mainly includes the following modules:

[0091] Data acquisition module:

[0092] Used to acquire the blood vessel enhanced image (MRA) and the blood vessel wall image (vessel wall MRI).

[0093] Image registration module:

[0094] Realize the registration between the blood vessel wall image and the blood vessel enhanced image, register the blood vessel wall image B1 to the blood vessel enhanced image A1, and obtain the registered blood vessel wall image A2.

[0095] Segmentation module:

[0096] Use the deep learning algorithm to segment the blood vessel enhanced image A1 and the registered blood vessel wall image A2, and obtain the blood vessel lumen segmentation result S1 and the blood vessel wall segmentation result S2 respectively.

[0097] Fusion module:

[0098] Perform weighted fusion on the blood vessel enhanced image A1 and the registered blood vessel wall image A2 to obtain the fusion image A3.

[0099] Analysis module:

[0100] Calculate the intersection region S3 of the blood vessel lumen segmentation result S1 and the blood vessel wall segmentation result S2, and set the labels of the intersecting voxels to the blood vessel lumen labels.

[0101] Perform three-dimensional connected component analysis to identify occluded blood vessel regions, and classify and process them according to different connected component label distributions (such as removing false positive segmentation results, retaining normal blood vessels, marking stenotic blood vessels, and retaining suspected occluded blood vessels).

[0102] Three-dimensional visualization and surface reconstruction module:

[0103] Extract the three-dimensional skeleton of the three-dimensional connected component of the suspected occluded blood vessel, calculate the central axis, and correct the central axis point coordinates.

[0104] Extract local region images and perform surface reconstruction to achieve a detailed display of the occluded region.

[0105] User interface module (optional):

[0106] Provide a user interaction interface to allow doctors to view and operate on the segmentation results and reconstructed images for preoperative evaluation.

[0107] Storage module:

[0108] Store the original image data, intermediate processing results, and final analysis results.

[0109] These modules work together to jointly achieve the precise reconstruction and analysis of the head and neck blood vessel lumen and wall, thereby assisting doctors in analyzing the lesion degree and blood vessel morphology of the large blood vessels at the occlusion site and their distal blood vessels and branch blood vessels, and completing the preoperative evaluation decision more quickly and accurately.

[0110] The method of the present invention includes the following steps:

[0111] Obtain data: Obtain the blood vessel enhanced image (MRA) A1 and the blood vessel wall image B1.

[0112] Image registration: By means of rigid registration, register the blood vessel wall image B1 to the blood vessel enhanced image A1 to obtain the registered blood vessel wall image A2.

[0113] Deep learning-based blood vessel lumen and blood vessel wall segmentation:

[0114] Use deep learning methods, and through the method of channel splicing, use the blood vessel enhanced image A1 and the registered blood vessel wall image A2 data as network inputs to obtain the blood vessel lumen segmentation result S1 and the blood vessel wall segmentation result S2.

[0115] Segmentation result processing:

[0116] Calculate the intersection region S3 of the blood vessel lumen segmentation result S1 and the blood vessel wall segmentation result S2, and set the labels of the intersecting voxels to the blood vessel lumen label.

[0117] Calculate the union S4 of the intersection region S3 and the blood vessel wall segmentation result S2, and perform three-dimensional connected component analysis on S4 to identify occluded blood vessel regions.

[0118] Connected component analysis and classification:

[0119] Perform three-dimensional connected component analysis on S4, count the distribution of segmentation labels within each connected component, and classify them as follows:

[0120] Remove false positive segmentation results (connected components with only blood vessel wall segmentation labels).

[0121] Retain normal blood vessels (connected components with only blood vessel lumen segmentation labels).

[0122] Mark stenotic blood vessels (connected components with both blood vessel lumen and blood vessel wall segmentation labels, but the volume of the blood vessel wall segmentation is less than 20% of the connected component volume).

[0123] Retain suspected occluded blood vessels (connected components with both blood vessel lumen and blood vessel wall segmentation labels, and the volume of the blood vessel wall segmentation accounts for more than 20% of the connected component volume).

[0124] Three-dimensional visualization and surface reconstruction:

[0125] For the three-dimensional connected component of the suspected occluded blood vessel, use the three-dimensional skeleton extraction method to calculate the initial central axis X1 of the connected component, and correct the central axis point coordinates X2.

[0126] Based on the corrected central axis point coordinates X2, extract the corresponding local region image and perform surface reconstruction.

[0127] These steps together constitute a blood vessel image processing and analysis method for combined magnetic resonance angiography and vessel wall imaging, aiming to assist doctors in analyzing the lesion degree and vessel morphology of large blood vessels and their distal and branch vessels at the occlusion site, in order to more quickly and accurately complete preoperative evaluation and decision-making.

[0128] Principle of operation

[0129] 1. Data acquisition

[0130] Obtain blood vessel enhanced image (MRA): First, obtain the blood vessel enhanced image (MRA) of the patient. This is a commonly used vascular imaging technique that can clearly show the morphology and structure of blood vessels.

[0131] Obtain blood vessel wall image (vessel wall MRI): At the same time, obtain the image of the blood vessel wall (vessel wall MRI). This kind of image can provide detailed information about the blood vessel wall, including changes in the intima, media, and adventitia.

[0132] 2. Image Registration

[0133] Rigid Registration: The vascular wall image (B1) is registered onto the vascular enhancement image (A1) through rigid registration to obtain the registered vascular wall image (A2). This process is to ensure the spatial correspondence between the two images for subsequent processing.

[0134] 3. Deep Learning-based Segmentation

[0135] Input to the Deep Learning Network: The vascular enhancement image (A1) and the registered vascular wall image (A2) are input into a pre-trained deep learning network through channel concatenation.

[0136] Output of the Segmentation Results: The network outputs the vascular lumen segmentation result (S1) and the vascular wall segmentation result (S2). The segmentation results use labels to distinguish different tissue structures, where the label for the vascular lumen is 1 and the label for the vascular wall is 2.

[0137] 4. Processing of the Segmentation Results

[0138] Intersection Calculation: Calculate the intersection region (S3) between the vascular lumen segmentation result (S1) and the vascular wall segmentation result (S2), and set the label of the intersection voxels to the vascular lumen label (1).

[0139] Union Calculation and 3D Connected Component Analysis: Calculate the union (S4) of the intersection region (S3) and the vascular wall segmentation result (S2), and perform 3D connected component analysis on S4 to count the distribution of segmentation labels within each connected component.

[0140] 5. Connected Component Analysis and Classification

[0141] Classification Processing:

[0142] False Positive Segmentation Results: When only the vascular wall segmentation label exists within the connected component, it is regarded as an incorrect segmentation result and removed.

[0143] Normal Blood Vessels: When only the vascular lumen segmentation label exists within the connected component, it is regarded as a normal blood vessel and retained.

[0144] Narrow Blood Vessels: When both the vascular lumen and vascular wall segmentation labels exist within the connected component, but the volume of the vascular wall segmentation is less than 20% of the volume of the connected component, it is regarded as a narrow blood vessel and retained.

[0145] Suspected Occluded Blood Vessels: When both the vascular lumen and vascular wall segmentation labels exist within the connected component, and the volume of the vascular wall segmentation exceeds 20% of the volume of the connected component, it is regarded as a suspected occluded blood vessel and retained.

[0146] 6. 3D Visualization and Surface Reconstruction

[0147] 3D skeleton extraction and central axis correction: For the 3D connected region of the suspected occluded blood vessel, use the 3D skeleton extraction method to calculate the initial central axis (X1) of the connected region, and by traversing the central axis coordinate points of the connected region, perform elliptical fitting on the mask contour points within the specified range to obtain the corrected central axis point coordinates (X2).

[0148] Fusion image generation: Perform weighted fusion on the vessel-enhanced image (A1) and the registered vessel wall image (A2) to obtain the fusion image (A3).

[0149] Local region image extraction and surface reconstruction: Based on the fusion image (A3) and the corrected central axis point coordinates (X2), extract the corresponding local region image to achieve the surface reconstruction of the suspected occluded region.

[0150] Through the above steps, the method and system of the present invention can achieve the precise reconstruction and analysis of the lumen and wall of the head and neck blood vessels, assisting doctors to complete preoperative assessment decisions more quickly and accurately.

[0151] Example 1:

[0152] Acute endovascular treatment for stroke is a battle, and time is brain. The preoperative assessment should be as concise as possible. When performing elective interventional recanalization for non-acute large artery occlusion, it is a positional battle, and the operation is complex and difficult. A detailed preoperative assessment can provide key information for a safe and efficient operation. In particular, advanced multimodal imaging assessment enables the surgeon to comprehensively master the details of the patient's pathological mechanism.

[0153] Compared with carotid artery stenosis, the cases of large artery occlusion in the head and neck are much more complex. When the lumen of the large artery in the head and neck is in an occluded state, especially when the intracranial artery is occluded, the information obtained from routine preoperative vascular assessment is limited. Interventional doctors usually face a "dark room" without windows. During the operation, they rely on operation feedback and angiography to continuously understand the changes in the vascular conditions of the occluded segment of the large artery and adjust the treatment strategy. It is inevitable that there is uncertainty. Therefore, in order to safely and efficiently implement interventional recanalization treatment for non-acute occlusion of large arteries in the head and neck, it is necessary to understand its pathological mechanism as much as possible before surgery, including the cerebral infarction and blood perfusion conditions in the blood supply area of the occluded artery, the location, scope and cause of the large artery occlusion lesion, whether there are stenosis-occlusion lesions in the extracranial and intracranial branch vessels of the occluded large artery, and whether there is thrombus in the lumen of the occluded segment of the large artery. The cerebral infarction and blood perfusion conditions in the blood supply area of the occluded artery can be determined based on the method of volume mismatch between the cerebral infarction area and the hypoperfused area in cerebral perfusion examination. Due to the complete occlusion of the middle cerebral artery, traditional vascular assessment methods such as CTA or MRA and other imaging techniques cannot accurately judge the relevant information of the vascular wall conditions and their pathological characteristics at the distal end of the occluded blood vessel, while wall MRI can provide information of potential value. Magnetic resonance vessel wall imaging can display the intima, subintimal structure and wall changes in the blood vessel through images, helping doctors observe the blood vessel conditions.

[0154] Currently, there are only post-processing and analysis software for lumen imaging or vessel wall imaging alone. The former is mainly used for three-dimensional reconstruction and visualization of blood vessels, identification of lesions such as vascular stenosis and aneurysms, and the latter mainly focuses on the assessment of the pathological nature and stability of plaques. There is a lack of a method or software that can simultaneously analyze and process the lumen and vessel wall images of blood vessels, realize three-dimensional and curved surface reconstruction and visualization of the lumen and vessel wall of blood vessels, and assist doctors in analyzing the lesion degree and vascular morphology of the large blood vessel at the occlusion site and its distal blood vessels and branch blood vessels, so as to complete the preoperative assessment and decision-making more quickly and accurately.

[0155] Provide a fully automatic, rapid and accurate volume reconstruction of the lumen and wall of the head and neck blood vessels, and combine the morphology of the vessel wall to perform curved surface reconstruction on the suspected occluded blood vessel area, assisting doctors in analyzing the lesion degree and vascular morphology of the large blood vessel at the occlusion site and its distal blood vessels and branch blood vessels, so as to complete the preoperative assessment and decision-making more quickly and accurately:

[0156] (1) Using the enhanced vascular (MRA) as the main image A1, and registering the vessel wall image B1 to the enhanced vascular image through rigid registration to obtain image A2.

[0157] (2) Such as Figure 1As shown, based on the deep learning method, the data of the three-dimensional vascular enhancement and vascular wall image sequences (A1, A2) are used as the network input through channel splicing, and the segmentation results of the regions of interest corresponding to the data are used as the network output targets (the segmentation result label of the vascular lumen is 1, and the segmentation result label of the vascular wall is 2). The segmentation network is repeatedly iteratively trained until the network segmentation result is consistent with the region of interest.

[0158] (3) Calculate the intersection region S3 = S1 ∩ S2 of the vascular lumen segmentation result S1 and the vascular wall segmentation result S2, and set the segmentation label of the intersection voxels to 1, that is, regard the intersection voxels as lumen voxels. Calculate the union S4 = S3 ∪ S2 of the intersection region and the vascular wall segmentation result, and output the three-dimensional visualization diagram of the intersection, as Figure 2 shown. Then, perform three-dimensional connected component analysis on S4, count the segmentation label distribution in each connected component, and divide it into the following four cases:

[0159] ① False positive segmentation result: When only the vascular wall segmentation label exists in the connected component, regard the connected component as an incorrect segmentation result and remove it.

[0160] ② When only the vascular lumen segmentation label exists in the connected component, regard the connected component as a normal blood vessel and retain it;

[0161] ③ When both the vascular lumen and vascular wall segmentation labels exist in the connected component, but the volume of the vascular wall segmentation is less than 20% of the volume of the connected component, regard the connected component as a stenotic blood vessel and retain it.

[0162] ④ When both the vascular lumen and vascular wall segmentation labels exist in the connected component, and the volume of the vascular wall segmentation accounts for > 20% of the volume of the connected component, regard the connected component as a suspected occluded blood vessel and retain it.

[0163] (4) For the three-dimensional connected components suspected of having occluded blood vessels, use the three-dimensional skeleton extraction method to calculate the initial central axis X1 of the connected component. Along the z-axis direction, traverse the central axis coordinate points of the connected component from the bottom to the brain direction. Taking the central axis coordinate point region as the center, extract the connected component mask submask_D within the range of 80x80 pixels, extract the contour points of the mask, and perform ellipse fitting. Using the center point coordinates of the fitted ellipse, obtain the corrected central axis point coordinates X2. Weightedly fuse the vascular enhancement image A1 and the registered vascular wall image A2 to obtain the fused image A3 = 0.7 * A1 + 0.3 * A2. Based on the fused image A3 and the corrected central axis point coordinates X2, extract the corresponding local region image to realize the surface reconstruction of the corresponding suspected occlusion region, as Figure 3 shown.

[0164] Example 2: Vascular lumen and vascular wall segmentation and analysis

[0165] Step 1: Obtain data

[0166] Obtain the vascular enhanced image (MRA) A1:

[0167] Use a 3T MRI scanner to acquire the vascular enhanced image A1 of the patient's head, ensuring that the image covers the entire head and neck vascular region.

[0168] File format: DICOM format, Resolution: 0.5mm × 0.5mm × 2mm.

[0169] Obtain the vessel wall image (wall MRI) B1:

[0170] Use a high-resolution wall MRI scanner to acquire the vessel wall image B1, ensuring that the image clearly shows the vessel wall structure.

[0171] File format: NIFTI format, Resolution: 0.3mm × 0.3mm × 0.3mm.

[0172] Step 2: Image registration

[0173] Rigid registration:

[0174] Use a rigid registration algorithm based on mutual information (MI) to register the vessel wall image B1 to the vascular enhanced image A1, obtaining the registered vessel wall image A2.

[0175] Ensure that the registration error is less than 1mm.

[0176] Step 3: Deep learning-based segmentation

[0177] Input of the deep learning network:

[0178] Use a pre-trained convolutional neural network (CNN), and take the vascular enhanced image A1 and the registered vessel wall image A2 as the network input by channel concatenation.

[0179] Input size: 256×256×16 (XYZ dimensions).

[0180] Output of the segmentation result:

[0181] The network outputs the vessel lumen segmentation result S1 (labeled 1) and the vessel wall segmentation result S2 (labeled 2).

[0182] The segmentation accuracy reaches over 95%.

[0183] Step 4: Processing of the segmentation result

[0184] Intersection calculation:

[0185] Calculate the intersection region S3 of the blood vessel lumen segmentation result S1 and the blood vessel wall segmentation result S2, and set the labels of the intersection voxels to the blood vessel lumen label (1).

[0186] The volume of the intersection region S3 accounts for about 5% of the total segmentation region.

[0187] Union calculation and three-dimensional connected component analysis:

[0188] Calculate the union S4 of the intersection region S3 and the blood vessel wall segmentation result S2, and identify the occluded blood vessel regions based on three-dimensional connected component analysis.

[0189] Statistically analyze the segmentation label distribution in each connected component and classify them:

[0190] Remove false positive segmentation results (connected components with only blood vessel wall segmentation labels).

[0191] Retain normal blood vessels (connected components with only blood vessel lumen segmentation labels).

[0192] Mark stenotic blood vessels (connected components with both blood vessel lumen and blood vessel wall segmentation labels, but the volume of the blood vessel wall segmentation is less than 20% of the volume of the connected component).

[0193] Retain suspected occluded blood vessels (connected components with both blood vessel lumen and blood vessel wall segmentation labels, and the volume of the blood vessel wall segmentation accounts for more than 20% of the volume of the connected component).

[0194] Step 5: Three-dimensional visualization and surface reconstruction

[0195] Three-dimensional skeleton extraction and central axis correction:

[0196] For the three-dimensional connected component of the suspected occluded blood vessel, use the three-dimensional skeleton extraction method to calculate the initial central axis X1 of the connected component, and correct the central axis point coordinates X2 by ellipse fitting.

[0197] The error of central axis correction is controlled within 0.5 mm.

[0198] Fusion image generation:

[0199] Perform weighted fusion of the blood vessel enhanced image A1 and the registered blood vessel wall image A2 with a weight ratio of 0.7:A1 + 0.3:A2 to obtain the fusion image A3.

[0200] Local region image extraction and surface reconstruction:

[0201] Based on the fusion image A3 and the corrected central axis point coordinates X2, extract the local region image within the range of 80x80 pixels and perform surface reconstruction.

[0202] The surface reconstruction accuracy reaches the sub-millimeter level.

[0203] Example 2: System Integration and User Interface

[0204] Step 1: Data Acquisition

[0205] Obtain enhanced vascular image (MRA) A1:

[0206] Use a 1.5T MRI scanner to acquire the enhanced vascular image A1 of the patient's head.

[0207] File format: DICOM format, Resolution: 1mm × 1mm × 3mm.

[0208] Obtain vessel wall image (wall MRI) B1:

[0209] Use a high-resolution wall MRI scanner to acquire the vessel wall image B1.

[0210] File format: NIFTI format, Resolution: 0.5mm × 0.5mm × 0.5mm.

[0211] Step 2: Image Registration

[0212] Rigid registration:

[0213] Use a rigid registration algorithm based on mutual information (MI) to register the vessel wall image B1 to the enhanced vascular image A1, obtaining the registered vessel wall image A2.

[0214] Ensure that the registration error is less than 2mm.

[0215] Step 3: Deep Learning-based Segmentation

[0216] Input of the deep learning network:

[0217] Use a pre-trained deep residual network (ResNet), and take the enhanced vascular image A1 and the registered vessel wall image A2 as the network input through channel concatenation.

[0218] Input size: 512×512×16 (XYZ dimensions).

[0219] Output of the segmentation result:

[0220] The network outputs the vessel lumen segmentation result S1 (labeled 1) and the vessel wall segmentation result S2 (labeled 2).

[0221] The segmentation accuracy reaches over 90%.

[0222] Step 4: Processing of the Segmentation Result

[0223] Intersection calculation:

[0224] Calculate the intersection area S3 of the blood vessel lumen segmentation result S1 and the blood vessel wall segmentation result S2, and set the labels of the intersection voxels to the blood vessel lumen label (1).

[0225] The volume of the intersection area S3 accounts for about 10% of the total segmentation area.

[0226] Union calculation and three-dimensional connected component analysis:

[0227] Calculate the union S4 of the intersection area S3 and the blood vessel wall segmentation result S2, and identify the occluded blood vessel area based on three-dimensional connected component analysis.

[0228] Statistically analyze the segmentation label distribution in each connected component and classify for processing:

[0229] Remove false positive segmentation results (connected components with only blood vessel wall segmentation labels).

[0230] Retain normal blood vessels (connected components with only blood vessel lumen segmentation labels).

[0231] Mark stenotic blood vessels (connected components with both blood vessel lumen and blood vessel wall segmentation labels, but the volume of the blood vessel wall segmentation is less than 20% of the volume of the connected component).

[0232] Retain suspected occluded blood vessels (connected components with both blood vessel lumen and blood vessel wall segmentation labels, and the volume of the blood vessel wall segmentation accounts for more than 20% of the volume of the connected component).

[0233] Step 5: Three-dimensional visualization and surface reconstruction

[0234] Three-dimensional skeleton extraction and central axis correction:

[0235] For the three-dimensional connected component of the suspected occluded blood vessel, use the three-dimensional skeleton extraction method to calculate the initial central axis X1 of the connected component, and correct the central axis point coordinates X2 by ellipse fitting.

[0236] The central axis correction error is controlled within 1 mm.

[0237] Fusion image generation:

[0238] Perform weighted fusion of the blood vessel enhanced image A1 and the registered blood vessel wall image A2 with a weight ratio of 0.6:A1 + 0.4:A2 to obtain the fusion image A3.

[0239] Local region image extraction and surface reconstruction:

[0240] Based on the fusion image A3 and the corrected central axis point coordinates X2, extract the local region image within the range of 80x80 pixels and perform surface reconstruction.

[0241] The surface reconstruction accuracy reaches the sub-millimeter level.

[0242] User Interface:

[0243] Integrate a user interface module in the system to display the processing results and allow doctors to interact, view, and analyze the segmentation results and reconstructed images.

[0244] The user interface supports zooming, rotation, and panning functions, facilitating doctors to observe the vascular structure from different angles.

[0245] The above embodiments demonstrate the specific applications of the present invention under different device configurations, ensuring the generality and flexibility of the method. Through these detailed steps, doctors can obtain comprehensive and accurate vascular information, thus better performing preoperative evaluation and decision-making.

[0246] The above are only the preferred embodiments of the present invention, and thus do not limit the scope of the present invention. Any equivalent structural transformation made under the inventive concept of the present invention by using the content of the specification and drawings of the present invention, or direct / indirect application in other related technical fields, is included in the protection scope of the present invention.

Claims

1. A vascular image processing and analysis system for combined magnetic resonance angiography and vessel wall imaging, characterized in that The system includes: A data acquisition module for acquiring a vascular enhanced image A1 and a vessel wall image B1; An image registration module that realizes the registration between the vessel wall image and the vascular enhanced image, registers the vessel wall image B1 to the vascular enhanced image A1, and obtains the registered vessel wall image A2; A segmentation module that uses a deep learning algorithm to segment the vascular enhanced image A1 and the registered vessel wall image A2, and respectively obtains a vascular lumen segmentation result S1 and a vessel wall segmentation result S2; A fusion module that performs weighted fusion on the vascular enhanced image A1 and the registered vessel wall image A2 to obtain a fused image A3; An analysis module that calculates the intersection region S3 between the vascular lumen segmentation result S1 and the vessel wall segmentation result S2, and sets the label of the intersection voxels as the vascular lumen label; Calculate the union S4 of the intersection region S3 and the vessel wall segmentation result S2; Perform three-dimensional connected component analysis on S4, identify the occluded vessel regions, and classify and process them according to different connected component label distribution situations, remove false positive segmentation results, retain normal vessels, mark stenotic vessels, and retain suspected occluded vessels; A three-dimensional visualization and surface reconstruction module for performing three-dimensional skeleton extraction and surface reconstruction on the three-dimensional connected components of the suspected occluded vessels; A storage module for storing the original image data, intermediate processing results, and final analysis results.

2. The system according to claim 1, wherein The image registration module adopts rigid body registration to improve the registration accuracy.

3. The system according to claim 1, characterized in that The segmentation module uses a pre-trained CNN model, and data augmentation technology is adopted during the training of the model to improve the segmentation accuracy.

4. The system according to claim 1, wherein The fusion module further includes a denoising algorithm for further removing small-area noise points.

5. The system according to claim 1, characterized in that, The three-dimensional visualization and surface reconstruction module includes processing with the Laplacian operator for smoothing contour points, and obtaining more accurate central axis point coordinates through ellipse fitting.

6. A method for processing and analyzing vascular images by combining magnetic resonance angiography and vessel wall imaging, based on the system according to any one of claims 1-5, characterized in that The method includes the following steps: S101, acquire a vascular enhanced image A1 and a vessel wall image B1; S102, register the vessel wall image B1 to the vascular enhanced image A1 by means of rigid body registration to obtain the registered vessel wall image A2; S103, based on the deep learning method, through the way of channel splicing, use the data of the vascular enhanced image A1 and the registered vessel wall image A2 as the network input to obtain a vascular lumen segmentation result S1 and a vessel wall segmentation result S2; S104, calculate the intersection region S3 between the vascular lumen segmentation result S1 and the vessel wall segmentation result S2, and set the label of the intersection voxels as the vascular lumen label; S105, calculate the union S4 of the intersection region S3 and the vessel wall segmentation result S2, and perform three-dimensional connected component analysis on S4 to identify the occluded vessel regions; S106, for the three-dimensional connected components of the suspected occluded vessels, use the three-dimensional skeleton extraction method to calculate the initial central axis X1 of the connected components and correct the central axis point coordinates X2; S107, generate a weighted fusion image A3 based on the vascular enhanced image A1 and the registered vessel wall image A2; S108, based on the corrected central axis point coordinates X2, extract the local region image from the weighted fusion image A3 for surface reconstruction.

7. The method according to claim 6, wherein In step S102, rigid registration is implemented using the iterative closest point algorithm.

8. The method according to claim 6, wherein In step S103, the deep learning method uses a convolutional neural network, and data augmentation techniques are used during network training.

9. The method according to claim 6, wherein In step S104, the calculation of the intersection region S3 also includes removing small-area noise points based on morphological operations.

10. The method according to claim 6, characterized in that In step S106, the three-dimensional skeleton extraction method includes traversing the connected domain coordinate points along the central axis and performing elliptical fitting on the mask contour points within a specified range to obtain the corrected central axis point coordinates X2.

Images