Acute ischemic stroke occlusion section CT blood vessel reconstruction method and system
By combining the generative model and the segmentation model, the problem of insufficient data for simulating occluded blood vessels in the existing technology is solved, and efficient and accurate reconstruction of occluded blood vessels is achieved, supporting personalized medicine and improving diagnostic efficiency.
Patent Information
- Application Number
- CN202510111547.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-09-12
AI Technical Summary
Existing technologies lack reliable methods to generate simulated occluded blood vessel data when diagnosing and treating acute ischemic stroke, resulting in poor data stability and repeatability, making it difficult to support extensive research and accurate diagnosis.
A generative model-based approach is used to learn from non-occluded vascular images and generate realistic simulated occluded vascular images. By training a deep learning model, combining the generative model and the segmentation model, a multi-task learning framework is constructed to achieve occlusion simulation and vascular segmentation.
It improves data diversity and model accuracy, reduces ethical risks, promotes personalized medicine, and enhances the reliability and efficiency of diagnostic tools.
Smart Images
Figure FT_1
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical image processing, and in particular to a method and system for CT blood vessel reconstruction of an occluded segment in acute ischemic stroke. Background Art
[0002] Acute ischemic stroke (AIS) is an ischemic injury to the brain caused by large vessel occlusion, typically manifesting as sudden neurological loss. Timely and accurate assessment of the location and extent of cerebral vascular occlusion is crucial for the diagnosis and treatment of this condition. Traditional stroke diagnostic methods rely on clinical imaging techniques such as CT (computed tomography) and CTA (CT angiography). These technologies provide high-resolution images of cerebral vessels, helping physicians determine the extent of vascular occlusion. However, challenges remain in managing complex occlusion morphologies and accurately assessing them.
[0003] With the development of artificial intelligence (AI) technology, deep learning methods have been widely used in medical image analysis, especially in automated segmentation and reconstruction tasks. Traditional segmentation methods mostly rely on handcrafted features or rule-based algorithms, which have limited efficiency and accuracy. Deep learning models, especially convolutional neural networks (CNNs), have demonstrated excellent performance in image segmentation and reconstruction.
[0004] Traditionally, researchers and physicians have relied on real-world patient data to study and diagnose vascular diseases, such as occluded vessel revascularization tasks. However, obtaining real-world occluded vessel data presents numerous challenges, including but not limited to ethical issues, limited sample sizes, and difficulty obtaining early-stage lesion information. Furthermore, variability in real-world data can lead to unstable or poorly reproducible experimental results. Therefore, a method for reliably generating simulated occluded vessels is urgently needed to support broader research and the development of more accurate diagnostic tools. Summary of the Invention
[0005] To address the above issues, the present invention proposes a generative model-based method that can learn from non-occluded vessel images and generate realistic simulated occluded vessel images. By training a deep learning model to understand the differences between normal vascular structures and occluded vessels, the model can output a corresponding simulated occlusion given a non-occluded vessel input. Specifically, a method for CT vascular reconstruction of occluded segments in acute ischemic stroke includes the following steps: Pre-operative and post-operative CT images of patients with acute ischemic stroke large vessel occlusion who have successfully undergone endovascular treatment for recanalization are obtained; a generative model is constructed, wherein the generative model is used to generate simulated occluded vessels for non-occluded vessels, and the training samples of the generative model are the pre-operative and post-operative CT images; based on the generative model, simulated occluded vessels are generated for non-occluded vessels to obtain simulated occlusion data; a segmentation model is constructed, wherein the segmentation model is used to perform vessel segmentation on the acute ischemic stroke CT images to complete the reconstruction of the occluded CT vessels of the acute ischemic stroke; the training samples of the segmentation model include the simulated occlusion data and the pre-operative and post-operative CT images.
[0006] The method also includes a fusion step, including: combining the generation model and the segmentation model to construct a multi-task learning framework, using the same neural network model to simultaneously perform occlusion simulation and blood vessel segmentation; the image generated in the generation module is passed through the segmentation module to calculate the training loss, and the calculated network gradient is synchronously updated in the generation module and the segmentation module; the performance of multiple models, including segmentation accuracy and the accuracy of generated occlusion, is compared, and the best solution is selected.
[0007] A second aspect of the present invention provides an acute ischemic stroke occluded segment CT vascular reconstruction system, comprising: An acquisition unit acquires pre- and post-operative CT images of patients with acute ischemic stroke large vessel occlusion who have successfully undergone endovascular treatment for recanalization; a generative model construction unit, wherein the generative model is used to generate simulated occluded vessels for non-occluded vessels, and the training samples of the generative model are pre- and post-operative CT images; a data amplification unit, based on the generative model, generates simulated occluded vessels for non-occluded vessels to obtain simulated occlusion data; a segmentation unit, wherein the segmentation model is used to perform vessel segmentation on the acute ischemic stroke CT images to complete the reconstruction of the occluded CT vessels of the acute ischemic stroke; the training samples of the segmentation model include the simulated occlusion data and the pre- and post-operative CT images.
[0008] A third aspect of the present invention provides an electronic device comprising: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the electronic device executes the above-mentioned method for CT vascular reconstruction of the occluded segment of acute ischemic stroke.
[0009] A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored therein, which, when executed on a computer, enables the computer to execute the above-mentioned method for CT vascular reconstruction of an occluded segment of acute ischemic stroke.
[0010] The present invention has the following beneficial effects: The model using the solution of the present invention does not consider the information of the blood vessel segments, thus achieving complete self-adaptation.
[0011] 1. Improve R&D efficiency: Generative models can quickly generate a large amount of high-quality simulated occluded blood vessel data, greatly accelerating model development.
[0012] 2. Increased data diversity: The model can create occlusion conditions of varying severity and types, increasing the diversity of data available for analysis.
[0013] 3. Reduce ethical risks: Since there is no need to collect occlusive samples directly from patients, the inconvenience and potential risks to patients are reduced.
[0014] 4. Enhance model accuracy: Using large amounts of simulated data to train occluded vessel detection and segmentation reconstruction models can help improve the generalization ability and accuracy of these model algorithms.
[0015] 5. Promote personalized medicine: The generative model can generate customized occlusion scenarios based on the location, course, and characteristics of non-occluded blood vessels in individual patients, helping to promote personalized medical solutions. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 Schematic diagram of the method for simulating occlusive angiogenesis based on the conditional diffusion generation model. DETAILED DESCRIPTION
[0017] The terms "first," "second," "third," "fourth," and so on (if any) in the description and claims of the present invention and in the accompanying drawings are used to distinguish similar items and are not necessarily used to describe a particular order or sequential sequence. It should be understood that the terms used in this manner are interchangeable under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that shown or described herein. In addition, the terms "including," "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.
[0018] The large blood vessels of the present invention refer to the internal carotid artery, the middle cerebral artery M1-2 segment, the anterior cerebral artery A1 segment and the vertebral basilar artery.
[0019] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 The first embodiment of the method for CT vascular reconstruction of an occluded segment of acute ischemic stroke according to the present invention includes: To obtain preoperative and postoperative CT images of patients with acute ischemic stroke and large vessel occlusion who were successfully recanalized by endovascular treatment; Constructing a generative model, wherein the generative model is used to simulate occluded blood vessels for non-occluded blood vessels, and the training samples of the generative model are pre-operative and post-operative CT images; generating a simulated occluded blood vessel for a non-occluded blood vessel based on the generation model to obtain simulated occlusion data; A segmentation model is constructed, which is used to perform blood vessel segmentation on acute ischemic stroke CT images and complete the reconstruction of CT blood vessels in the occluded segment of acute ischemic stroke; the training samples of the segmentation model include the simulated occlusion data and pre- and post-operative CT images.
[0020] Traditional vessel segmentation and occlusion simulation methods typically rely on real images of occluded vessels for training. This proposal, however, proposes a generative model-based approach that simulates occluded vessels by generating non-occluded vessels. This means the model can learn to generate vessels of varying degrees of occlusion from normal vessels without the need for occluded vessel images. This approach generates richer and more diverse occluded vessel data, addressing the clinical challenge of insufficient occluded lesion samples.
[0021] Specifically, the present invention employs a method for simulating occluded blood vessels based on a conditional diffusion generative model. This method uses a three-channel image as input and introduces specific conditional information to guide the generation process, ensuring that the generated occluded blood vessel images not only have a realistic appearance but also accurately reflect the set image features. The specific application steps of this method are as follows: 1. Select the target blood vessel segment: First, select a target blood vessel segment from the blood vessel segmentation mask of the normal data as the area to generate the occlusion effect.
[0022] 2. Set Occlusion Imaging Factors: For a selected target vessel segment, users can set occlusion imaging factors based on their needs. These factors determine the degree and appearance of occlusion in the resulting image, such as partial occlusion, complete occlusion, or varying stages of stenosis. This parameter allows physicians or researchers to adjust disease severity to better simulate real-world pathology.
[0023] 3. Background Removal: To improve the performance of the generative model, irrelevant background information in the target occluded vessel region is removed during input image processing. This preprocessing step helps the model focus on structural changes within the target region, resulting in more accurate and reliable occlusion simulation results.
[0024] 4. Iterative Generation Process: Once all the conditions are set, the next step is to use the conditional diffusion generative model to perform multiple iterations. In each iteration, the model gradually adjusts the pixel values in the image according to the given conditions until the desired occlusion effect is achieved.
[0025] 5. Obtain a simulated occlusion image: After multiple rounds of iterations, the final output is an image of an occluded blood vessel that meets the set conditions (i.e., a specific location and degree).
[0026] A method for simulating occluded blood vessels based on a conditional diffusion generative model precisely sets the target vessel segment, occlusion imaging coefficient, and removes background interference. Combined with multiple iterations of the conditional diffusion model, this method generates images of occluded blood vessels that reflect realistic pathological conditions. The method's innovation lies in the precise simulation of different stages and degrees of vascular occlusion through the introduction of conditional information and multiple rounds of noise removal, providing a more flexible and accurate occluded blood vessel generation solution.
[0027] As a preferred embodiment, the method further comprises a fusion step, comprising: The generative model and the segmentation model are combined to construct a multi-task learning framework, using the same neural network model to simultaneously perform occlusion simulation and vessel segmentation. The images generated by the generative module are trained in the segmentation module to calculate the training loss, and the calculated network gradients are used to synchronously update the generative module and the segmentation module. The performance of multiple models, including segmentation accuracy and the accuracy of generated occlusion, is compared to select the best solution.
[0028] Forward propagation: Input a non-occluded image and generate a simulated occluded image Input the simulated occluded image into the segmentation model to obtain the predicted mask The simulated occlusion image and the predicted mask are skeletonized and the topological continuity loss is calculated.
[0029] Backward Propagation: Multi-scale adversarial loss updates the generator and discriminator.
[0030] The topological continuity loss updates the shared encoder of the generator and segmenter.
[0031] The anatomical consistency feedback loss updates both the generator and the segmenter.
[0032] This paper combines a generative model with a vessel segmentation model to construct a multi-task learning framework, enabling simultaneous occlusion simulation and vessel segmentation using the same neural network model. This approach leverages the interaction and synergy between the two to enhance the performance of the generative and segmentation models, while simultaneously optimizing both generative performance and segmentation accuracy through synchronized gradient updates.
[0033] As a preferred embodiment, the fusion step further includes a clinical evaluation step: The model was evaluated and the model parameters were adjusted based on the centerline accuracy evaluation, sensitivity, specificity evaluation, image quality evaluation, and clinical accuracy evaluation.
[0034] As a preferred embodiment, the specific steps of the segmentation model are: Blood vessel centerline extraction: Use the centerline extraction algorithm to extract the centerline of the blood vessel from the CT image; Path and shape optimization: Based on the centerline of the blood vessel, the path and shape of the blood vessel are further optimized to ensure accurate segmentation of the blood vessel edge; Blood vessel segmentation optimization: Using a deep learning segmentation network, the segmentation results of blood vessel edges are gradually optimized to ensure high-quality image rendering; Image rendering: Generate 3D reconstructed images of blood vessels through maximum intensity projection (MIP) and minimum intensity projection (MiniMIP).
[0035] As a preferred embodiment, the method further comprises: Image display: Generate CT vascular reconstruction images, including original images, 3D VR reconstruction images, MIP reconstruction images, and MiniMIP reconstruction images; Occlusion segment identification: Identify and mark the location of the occlusion segment, and display the threshold, accuracy, and contrast of the occlusion segment.
[0036] Display optimization: Generate and display the reconstructed image of the occluded segment, and adjust the display threshold and accuracy.
[0037] Lesion annotation: Identify and annotate other lesions associated with occlusion, including aneurysms, dissections, and stenosis.
[0038] The details are as follows: 1. Select preoperative and postoperative medical images (CT plain scan images, CT angiography images) of patients with acute ischemic stroke who had large vessel occlusion and were successfully recanalized after endovascular treatment. Develop an automatic matching algorithm to determine the gold standard of the preoperative occluded segment based on the postoperative images. a) Based on the previous normal cranio-cervical CT vessel segmentation algorithm (reference XXX), the preoperative and postoperative CT vessel segmentation models of patients with acute ischemic stroke large vessel occlusion were extracted respectively. b) Preoperative and postoperative vascular registration, adjustment, and verification c) Obtain training data, which is the gold standard after registration d) Build a neural network model and train it using training data e) Test the trained neural network model 2. Deep learning generative model performs occlusion simulation on normal CT data a) Extract the normal CT vascular segmentation model based on the previous normal cranial and cervical CT vascular segmentation algorithm (reference XXX) b) Based on the above occlusion model, the generative model is trained on normal cranial and cervical CT to generate occluded blood vessels to expand the sample size c) Test the trained neural network model 3. Deep learning segmentation model for vascular segmentation of acute ischemic stroke CT data a) Obtaining the vessel centerline b) Obtaining blood vessel shape and path based on the centerline c) Optimize blood vessel edge segmentation d) Image rendering e) Build model and train 4. AI algorithm for CT occluded vascular reconstruction in acute ischemic stroke based on a fusion generation and segmentation multi-task deep learning model a) Fusion generation and segmentation of multi-task deep learning models, comparing the performance of various models b) Large sample verification c) Clinical evaluation (centerline accuracy evaluation; sensitivity and specificity evaluation; image quality evaluation; clinical accuracy evaluation) 5. User result display interface, which can adjust the display threshold, accuracy, contrast, etc. according to user needs a) Generate complete cranial and cervical CT vascular reconstruction images, including original images, VR reconstruction, MIP reconstruction, and MiniMIP reconstruction b) Accurately identify and mark occluded segments c) Generate reconstructed images of occluded segments and adjust display threshold, accuracy, contrast, etc. or cancel display according to doctor's requirements d) Label and identify other related lesions such as aneurysm, arterial dissection, arterial stenosis, etc.
[0039] The above describes the method for CT angiography of an occluded segment in acute ischemic stroke according to an embodiment of the present invention. The following describes the apparatus for CT angiography of an occluded segment in acute ischemic stroke according to an embodiment of the present invention. The acquisition unit acquires preoperative and postoperative CT images of patients with acute ischemic stroke and large vessel occlusion who have been successfully recanalized by endovascular treatment; A generative model construction unit, wherein the generative model is used to generate simulated occluded blood vessels for non-occluded blood vessels, and the training samples of the generative model are pre- and post-operative CT images; a data amplification unit, which generates a simulated occluded blood vessel for the non-occluded blood vessel based on the generation model to obtain simulated occlusion data; The segmentation unit, the segmentation model is used to perform vascular segmentation on the acute ischemic stroke CT image and complete the reconstruction of the occluded CT blood vessels of the acute ischemic stroke; the training samples of the segmentation model include the simulated occlusion data and the pre- and post-operative CT images.
[0040] An embodiment of the present invention further provides an electronic device that may vary significantly due to different configurations or performance, and may include one or more central processing units (CPUs) (e.g., one or more processors) and memory, and one or more storage media (e.g., one or more mass storage devices) for storing applications or data. The memory and storage media may be either transient or persistent storage. The program stored on the storage medium may include one or more modules, each of which may include a series of instruction operations on the electronic device. Furthermore, the processor may be configured to communicate with the storage medium and execute the series of instruction operations on the storage medium on the electronic device.
[0041] The electronic device may further include one or more power supplies, one or more wired or wireless network interfaces, one or more input and output interfaces, and / or one or more operating systems, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that the electronic device structure in this embodiment does not constitute a limitation to electronic devices, and may include more or fewer components, or combine certain components, or arrange the components differently.
[0042] An embodiment of the present invention provides a structure of an electronic device, which may vary significantly due to different configurations or performance, and may include one or more processors (central processing units, CPUs) (for example, one or more processors) and memories, and one or more storage media for storing applications or data (for example, one or more mass storage devices). The memories and storage media may be either transient or persistent storage. The program stored in the storage medium may include one or more modules, each of which may include a series of instruction operations on the electronic device. Furthermore, the processor may be configured to communicate with the storage medium to execute the series of instruction operations in the storage medium on the electronic device.
[0043] The electronic device may further include one or more power supplies, one or more wired or wireless network interfaces, one or more input and output interfaces, and / or one or more operating systems, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that the electronic device structure does not constitute a limitation on the electronic device, and may include more or fewer components than those described above, or combine certain components, or arrange the components differently.
[0044] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to execute the steps of the aforementioned method.
[0045] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0046] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0047] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for CT vascular reconstruction of an occluded segment in acute ischemic stroke, characterized in that: The following steps are involved: To obtain preoperative and postoperative CT images of patients with acute ischemic stroke and large vessel occlusion who were successfully recanalized by endovascular treatment; Constructing a generative model, wherein the generative model is used to simulate occluded blood vessels for non-occluded blood vessels, and the training samples of the generative model are pre-operative and post-operative CT images; generating a simulated occluded blood vessel for a non-occluded blood vessel based on the generation model to obtain simulated occlusion data; A segmentation model is constructed, which is used to perform blood vessel segmentation on acute ischemic stroke CT images and complete the reconstruction of CT blood vessels in the occluded segment of acute ischemic stroke; the training samples of the segmentation model include the simulated occlusion data and pre- and post-operative CT images.
2. The method for CT vascular reconstruction of an occluded segment of acute ischemic stroke according to claim 1, characterized in that: The method further comprises a fusing step comprising: Combining the generative model with the segmentation model, a multi-task learning framework was constructed to perform occlusion simulation and vessel segmentation simultaneously using the same neural network model. The image generated by the generation module is passed through the segmentation module to calculate the training loss, and the calculated network gradient is used to synchronously update the generation module and the segmentation module; Compare the performance of multiple models, including segmentation accuracy and occlusion generation accuracy, and select the best solution.
3. The method for CT vascular reconstruction of an occluded segment of acute ischemic stroke according to claim 1, characterized in that: The fusion step also includes a clinical evaluation step: The model was evaluated and the model parameters were adjusted based on the centerline accuracy evaluation, sensitivity, specificity evaluation, image quality evaluation, and clinical accuracy evaluation.
4. The method for CT vascular reconstruction of an occluded segment of acute ischemic stroke according to claim 1, characterized in that: The specific steps of the segmentation model are: Blood vessel centerline extraction: Use the centerline extraction algorithm to extract the centerline of the blood vessel from the CT image; Path and shape optimization: Based on the centerline of the blood vessel, the path and shape of the blood vessel are further optimized to ensure accurate segmentation of the blood vessel edge; Blood vessel segmentation optimization: Using a deep learning segmentation network, the segmentation results of blood vessel edges are gradually optimized to ensure high-quality image rendering; Image rendering: Generate 3D reconstructed images of blood vessels through maximum intensity projection (MIP) and minimum intensity projection (MiniMIP).
5. The method for CT vascular reconstruction of an occluded segment of acute ischemic stroke according to claim 1, characterized in that: The method further comprises: Image display: Generate CT vascular reconstruction images, including original images, 3D VR reconstruction images, MIP reconstruction images, and MiniMIP reconstruction images; Occlusion segment identification: Identify and mark the location of the occlusion segment, and display the threshold, accuracy, and contrast of the occlusion segment.
6. The method for CT vascular reconstruction of an occluded segment of acute ischemic stroke according to claim 1, characterized in that: The method further comprises: Display optimization: Generate and display the reconstructed image of the occluded segment, and adjust the display threshold and accuracy.
7. The method for CT vascular reconstruction of an occluded segment of acute ischemic stroke according to claim 1, characterized in that: The method further comprises: Lesion annotation: Identify and annotate other lesions associated with occlusion, including aneurysms, dissections, and stenosis.
8. A CT vascular reconstruction system for occluded segments of acute ischemic stroke, characterized in that: The system comprises: The acquisition unit acquires preoperative and postoperative CT images of patients with acute ischemic stroke and large vessel occlusion who have been successfully recanalized by endovascular treatment; A generative model construction unit, wherein the generative model is used to generate simulated occluded blood vessels for non-occluded blood vessels, and the training samples of the generative model are pre- and post-operative CT images; a data amplification unit, which generates a simulated occluded blood vessel for the non-occluded blood vessel based on the generation model to obtain simulated occlusion data; The segmentation unit, the segmentation model is used to perform vascular segmentation on the acute ischemic stroke CT image and complete the reconstruction of the occluded CT blood vessels of the acute ischemic stroke; the training samples of the segmentation model include the simulated occlusion data and the pre- and post-operative CT images.
9. An electronic device comprising a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory to enable the electronic device to execute each step of the method for CT angiography reconstruction of an occluded segment of acute ischemic stroke according to any one of claims 1 to 7.
10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the method for CT vascular reconstruction of an occluded segment of acute ischemic stroke as claimed in any one of claims 1 to 7 are implemented.