Dynamic vascular imaging method and apparatus based on CT perfusion images

Abstract

The present invention relates to the field of image processing, and particularly relates to a dynamic vascular imaging method and apparatus based on CT perfusion images. The method comprises: acquiring a plurality of perfusion images obtained by means of performing CT scanning on each target layer in a region of interest, and determining an image set on the basis of the plurality of perfusion images; according to a scanning time sequence, defining every two adjacent images in the image set to be an image group, defining the scanning time of a second image in each image group to be later than that of a first image, and subtracting the first image from the second image in each image group, so as to obtain a corresponding local vascular image; acquiring all the local vascular images corresponding to all the target layers; obtaining three-dimensional vascular point cloud data on the basis of all the local vascular images; and inputting the three-dimensional vascular point cloud data into a preset deep learning model, so as to obtain a target vascular image. In the present invention, perfusion images are used to directly track blood vessels and reconstruct the blood vessels, thereby quickly obtaining a complete vascular image.

Description

Dynamic vascular imaging method and device based on CT perfusion images Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a vascular imaging method and device based on CT perfusion images. Background Art

[0002] Computed tomography perfusion (CTP) refers to the tracking of contrast agents by multiple consecutive same-layer scans of selected layers while contrast agents are injected intravenously, thereby dynamically obtaining the blood perfusion conditions of interest and evaluating physiological and pathological states and changes such as ischemia and necrosis in the region of interest.

[0003] Currently, existing technologies use perfusion images obtained through CTP to calculate a time-density (TDC) curve. This curve reflects changes in contrast agent concentration in the organ, indirectly reflecting changes in organ perfusion. Based on this curve, various mathematical models are used to calculate parameter maps corresponding to blood flow (BF), blood volume (BV), mean transit time (MTT), and transit time to the peak (TTP). Perfusion images provide not only perfusion information but also vascular information. Currently, research on using perfusion images to directly track and reconstruct blood vessels to rapidly obtain complete vascular images is still underway.

[0004] Summary of the Invention

[0005] In view of this, an object of the present invention is to provide a dynamic vascular imaging method and apparatus based on CT perfusion images.

[0006] In a first aspect, an embodiment of the present invention provides a method for vascular imaging, the method comprising:

[0007] Acquire multiple perfusion images obtained by performing CT scanning on a target layer of a region of interest, and determine an image set based on the multiple perfusion images;

[0008] defining each two adjacent images in the image set as an image group according to the scanning time sequence, defining the scanning time of the second image in each image group to be later than that of the first image, and subtracting the first image from the second image in each image group to obtain a corresponding local blood vessel image;

[0009] All the local blood vessel images are processed according to a preset deep learning model to obtain a target blood vessel image.

[0010] In combination with the first aspect, before the step of defining the image group, the method includes:

[0011] Determining a perfusion image in the image set as a reference perfusion image, and determining a first coordinate of a target feature point on the reference perfusion image;

[0012] determining a second coordinate of the target feature point on other perfusion images in the image set except the reference image;

[0013] Based on the first coordinates and the second coordinates, calculating a target transformation matrix for registering each of the other perfusion images to the reference perfusion image using a feature point matching method;

[0014] performing registration on the reference perfusion image and each of the other perfusion images according to the target transformation matrix to obtain other registered perfusion images;

[0015] All the registered other perfusion images and the reference perfusion image are taken as the image set.

[0016] In combination with the first aspect, the step of processing all the local vascular images according to a preset deep learning model to obtain a target vascular image includes:

[0017] Acquire local blood vessel images corresponding to all target layers of the region of interest;

[0018] Obtaining 3D point cloud data of blood vessels based on the local blood vessel images corresponding to all the target layers;

[0019] The three-dimensional point cloud data of the blood vessels is input into the preset deep learning model to obtain the target blood vessel image.

[0020] In combination with the first aspect, the step of inputting the three-dimensional point cloud data of the blood vessel into a preset deep learning model to obtain a target blood vessel image includes:

[0021] Acquiring original images obtained by performing CT scanning on all the target layers, wherein the original images are obtained by performing CT scanning on the target layers when the contrast agent does not flow into the target layers;

[0022] Obtaining original three-dimensional point cloud data according to the original image;

[0023] The original three-dimensional point cloud data and the blood vessel three-dimensional point cloud data are input into the preset deep learning model to obtain a target blood vessel image.

[0024] In combination with the first aspect, the step of processing all the local blood vessel images according to a preset deep learning model includes:

[0025] The preset deep learning model includes a deformable convolutional layer;

[0026] All the local blood vessel images are processed using the deformable convolution layer.

[0027] In combination with the first aspect, the preset deep learning model includes:

[0028] The preset deep learning model includes an attention gate, which is connected to the deformable convolutional layer.

[0029] In combination with the first aspect, the training steps of the preset deep learning model include:

[0030] Acquiring training samples, wherein the training samples include acquiring a plurality of perfusion images obtained by performing CT scanning on a target layer of a region of interest of a test object;

[0031] defining every two adjacent images in the image set as an image group according to a scanning time sequence, defining the scanning time of the second image in each image group to be later than that of the first image, and obtaining an image set corresponding to the target layer of the test object based on the multiple perfusion images;

[0032] subtracting the first image from the second image in the image set corresponding to the target layer of the test object to obtain a corresponding local blood vessel image;

[0033] Inputting all local vascular images of the test object into an initial deep learning model to obtain a predicted vascular image;

[0034] Acquiring a CTA image corresponding to a region of interest of the test object, and obtaining a corresponding target blood vessel image based on the CTA image;

[0035] calculating a residual between the predicted blood vessel image and the target blood vessel image;

[0036] Accumulating the residuals to obtain a loss function;

[0037] The model is optimized according to the loss function until a preset condition is met to obtain the preset deep learning model.

[0038] In combination with the first aspect, the step of processing all the local vascular images according to a preset deep learning model to obtain a target vascular image includes:

[0039] determining, based on the plurality of perfusion images, a change characteristic of a contrast agent in a blood vessel of the region of interest, the change characteristic comprising one or more of a contrast agent arrival time, a contrast agent average transit time, a contrast agent peak time, a blood flow rate, a blood volume, a half-peak width, or a time length at 2 / 3 of the peak width;

[0040] All the local blood vessel images and the change features are input into the deep learning model to obtain the target blood vessel image.

[0041] In a second aspect, the present application provides a vascular imaging device, comprising:

[0042] Image acquisition module, used to obtain multiple perfusion images obtained by CT scanning of the target layer of the area of ​​interest ， determining an image set based on the plurality of perfusion images;

[0043] an image calculation module, defining each two adjacent images in the image set as an image group in scan time order, defining the second image in each image group to have a scan time later than the first image, and subtracting the first image from the second image in each image group corresponding to the target layer of the test object to obtain a corresponding local vascular image;

[0044] A deep learning module is used to process all the local blood vessel images according to a preset deep learning model to obtain a target blood vessel image.

[0045] The embodiments of the present invention bring the following beneficial effects: The present invention provides a vascular scanning method and device, the method comprising: acquiring multiple perfusion images obtained by performing CT scanning on a target layer of an area of ​​interest, and determining an image set based on the multiple perfusion images; subtracting a first image from a second image in the image set to obtain a corresponding local vascular image, wherein the scanning time of the second image is later than that of the first image; and processing all the local vascular images according to a preset deep learning model to obtain a target vascular image.

[0046] The present application achieves the purpose of tracking blood vessels by calculating multiple perfusion images, and processes the tracked blood vessels through a neural network, and then reconstructs the complete blood vessels to improve the accuracy of vascular imaging.

[0047] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The purposes and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0050] FIG1 is a flow chart of a vascular imaging method according to an embodiment of the present invention;

[0051] FIG2 is a schematic structural diagram of a vascular imaging device provided by an embodiment of the present invention;

[0052] FIG3 is a schematic diagram of a processor structure of an electronic device applied to a blood vessel imaging method according to an embodiment of the present invention;

[0053] Reference numerals: 10 - image acquisition module, 20 - image calculation module, 30 - deep learning module; 41 - processor, 42 - bus, 43 - communication interface, 44 - memory. DETAILED DESCRIPTION

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0055] To facilitate understanding of this embodiment, the following is a brief introduction to the technical terms used in this application.

[0056] In recent years, CTP has been increasingly studied and applied in the diagnosis, prognosis, and treatment evaluation of brain diseases. Perfusion scanning can be used to quantitatively analyze the vascular distribution and blood perfusion status of tissues, providing a diagnostic basis for disorders with abnormal cerebral perfusion (such as stroke) and an effective reference for treatment decision-making. Therefore, it is necessary to explore methods for rapidly obtaining complete vascular images.

[0057] After introducing the technical terms involved in this application, the application scenarios and design concepts of the embodiments of this application are briefly introduced.

[0058] Existing analysis methods often acquire CT angiography (CTA) scans to assess the status of major blood vessels.

[0059] Based on this, this application provides a method and apparatus for calculating and directly tracking blood vessels from multiple perfusion images and reconstructing them, thereby rapidly obtaining a complete vascular image. This improves the signal-to-noise ratio and accuracy of vascular images. The vascular imaging method provided in this embodiment is applied to a processor 41 in an electronic device. The electronic device also includes a memory 44 storing a computer program. The processor 41 executes the computer program to implement the method provided in this embodiment.

[0060] Example 1

[0061] This embodiment provides a blood vessel imaging method, as shown in FIG1 , including:

[0062] S110, the processor obtains multiple perfusion images obtained by performing CT scanning on the target layer of the region of interest ， An image set is determined based on the plurality of perfusion images.

[0063] S120 , subtracting the first image from the second image in the image set to obtain a corresponding local blood vessel image, where the second image is scanned later than the first image.

[0064] S130: Process all the local blood vessel images according to a preset deep learning model to obtain a target blood vessel image.

[0065] In this embodiment, step S110 is performed by injecting a contrast agent into the test subject, and a period of time after the injection of the contrast agent, a CT (Computed Tomography) device is used to scan the target layer of the region of interest, and the region of interest may be the brain of the test subject. Since it takes a certain amount of time for the contrast agent to flow into the brain after intravenous injection, the perfusion image obtained may be an image before the contrast agent flows into the target layer, or it may be an image after the contrast agent flows into the target layer. In one embodiment, all the perfusion images corresponding to each target layer are obtained as an image set. In another embodiment, all the perfusion images corresponding to each target layer are screened, and perfusion images with a signal-to-noise ratio higher than a preset threshold or perfusion images with a difference greater than a preset threshold from the perfusion image acquired at the previous moment are selected. The difference in the image may be image features such as brightness and pixel value, and the multiple perfusion images obtained by screening are taken as an image set.

[0066] In the above example, step S120 obtains the corresponding local vascular image by subtracting the first image from the second image in the image set. Since the contrast agent gradually flows into the region of interest, over time, more and more contrast agent flows into the target layer, and the blood vessels become increasingly visible. By subtracting the first image from the second image, which was scanned later than the first image, the most recently visible local vascular image can be obtained.

[0067] The present invention also provides an embodiment, wherein, before the step of subtracting the first image from the second image in the image set to obtain the corresponding local vascular image, an image registration step is further included. Specifically, a perfusion image in the image set is determined as a reference perfusion image, and the first coordinates of a target feature point on the reference perfusion image are determined. The step of determining the reference perfusion image includes: selecting the first image in the image set or the image with the highest signal-to-noise ratio in the image set as the reference perfusion image. The second coordinates of the same target feature point on other perfusion images other than the reference image are determined. Based on the first coordinates and the second coordinates corresponding to each other perfusion image, a feature point matching method is used to calculate a target transformation matrix for registering each other perfusion image to the reference perfusion image. Based on the target transformation matrix corresponding to each other perfusion image, the reference perfusion image and the other perfusion images are registered to obtain registered other perfusion images. All registered other perfusion images and the reference perfusion image are used as a new image set for calculating the vascular image.

[0068] The present invention provides an embodiment in which a first image and a second image with adjacent scanning times are selected in a new image set, and the corresponding local vascular image is obtained by subtracting the first image from the second image. Then, as time passes, the local vascular image of the target layer that has been developed is obtained by subtracting the first image obtained by scanning the target layer at the previous moment from the second image obtained by performing a CT scan of the target layer at the latest moment. This is repeated in this manner, that is, all images in the image set are calculated to obtain the complete vascular image corresponding to the target layer.

[0069] This method dynamically tracks the blood vessels of interest and supplements the blood vessel information of the target layer in real time.

[0070] In step S130 , all local blood vessel images corresponding to the target layer are input into a preset deep learning model to obtain a target blood vessel image, where the target blood vessel is the complete blood vessel information of the target layer.

[0071] The present invention provides an embodiment, in which a preset deep learning model includes a feature extraction module and a deformable convolution layer, wherein the feature extraction module can be an ordinary convolution layer, and the deformable convolution layer is used to perform a convolution operation on the image features extracted by the feature extraction module to output high-dimensional features. The deformable convolution can have multiple deformable convolutions connected in sequence. By processing the image features through the deformable convolution, the output high-dimensional features contain richer local topological information, thereby improving the fidelity of the reconstructed vascular area.

[0072] The present invention provides an embodiment, in which the preset deep learning model also includes an attention gate, and each attention gate module includes an ordinary convolution layer and a nonlinear activation layer connected in sequence. The attention gate is connected to the deformable convolution layer. If there are multiple deformable convolutions connected in sequence, then the attention gate is connected to the last level of deformable convolution. For the attention gate connected to the deformable convolution layer, the attention gate is used to determine the weight coefficient based on the image features received by the deformable convolution layer and the high-dimensional features output by the deformable convolution layer; the image features and high-dimensional features received by the deformable convolution layer are weightedly calculated according to the weight coefficient to obtain the target image features. The interconnected deformable convolution layers and attention gates can be used to control the proportion of features learned by the deformable convolution layer and the ordinary convolution layer, so that the learned target image features contain richer local topological information.

[0073] The deep learning model also includes a reconstruction unit, which is used to reconstruct the image according to the target image features to obtain the target blood vessel image.

[0074] The present invention provides an embodiment for obtaining multiple perfusion images corresponding to each target layer of a region of interest, and determining a local vascular image corresponding to each target layer based on the multiple perfusion images. By performing CT perfusion scanning on multiple target layers of the region of interest, local vascular images corresponding to all target layers of the region of interest are obtained, wherein the target layers of the region of interest can be determined in a scanning protocol. For example, when a CT perfusion scan is performed on the brain of the subject, the target layer positions and the number of target layers scanned in the brain are determined in the scanning protocol. Based on the local vascular images corresponding to all target layers, vascular three-dimensional point cloud data is obtained, and the vascular three-dimensional point cloud data is input into a preset deep learning model to obtain a target vascular image. The preset deep learning model can be a point completion network (PCN), which reconstructs the input incomplete point cloud into a complete point cloud to obtain a target vascular image.

[0075] The present invention provides an embodiment for obtaining the original image obtained by CT scanning of all the target layers. ，The original image is obtained by performing a CT scan on the target layer when the contrast agent has not flowed into the target layer. Specifically, the original image can be obtained by performing a CT scan on all target layers of interest before the contrast agent is injected. The original image corresponding to all target layers can also be obtained by performing a CT scan on the target layer after the contrast agent is injected but before the contrast agent flows into the target layer. The original three-dimensional point cloud data is obtained based on the original image, and the original three-dimensional point cloud data and the three-dimensional point cloud data of the blood vessel are input into a preset deep learning model to obtain the target blood vessel image. In the process of completing the three-dimensional point cloud data of the blood vessel, more blood vessel morphological information is taken into account. For example, when the blood vessel is bifurcated or broken, the three-dimensional point cloud data of the blood vessel is completed based on the blood vessel information provided by other images, thereby improving the accuracy of the reconstructed blood vessel area.

[0076] The present invention provides an embodiment, comprising the steps of processing all local vascular images according to a preset deep learning model to obtain a target vascular image, further comprising determining, based on multiple perfusion images, the changing characteristics of a contrast agent within the blood vessels of a region of interest, the changing characteristics including one or more of contrast agent arrival time, average contrast agent transit time, contrast agent peak time, blood flow, blood volume, and the duration at half-peak width or 2 / 3 peak width; and inputting the local vascular images and the changing characteristics into the deep learning model to obtain the target vascular image. When obtaining the target vascular image, more vascular information is introduced, providing more reference information for image completion and reconstruction, thereby improving the accuracy of the reconstructed vascular region.

[0077] The present invention can be a preset deep learning neural network in different forms, such as a fully convolutional network with a U-shaped structure of an encoder and a decoder, and of course it can also be a fully connected network; the fully convolutional network can specifically be a ResNeXt network model, a ResNet network model, a graph convolutional network (GCN), etc.

[0078] The present invention provides a method for training a preset deep learning model, comprising the following steps: obtaining training samples, the training samples comprising: obtaining multiple perfusion images obtained by CT scanning a target layer of a region of interest of a test subject; obtaining an image set corresponding to the target layer of the test subject based on the multiple perfusion images; subtracting a first image from a second image in the image set corresponding to the target layer of the test subject to obtain a corresponding local vascular image; inputting all local vascular images of the test subject into an initial deep learning model to obtain a predicted vascular image; obtaining a CTA image corresponding to the region of interest of the test subject, and obtaining a corresponding target vascular image based on the CTA image; calculating the residual between the predicted vascular image and the target vascular image; accumulating the residual to obtain a loss function; and optimizing the model according to the loss function until a preset condition is met to obtain a preset deep learning model. CTA (CT angiography) refers to obtaining an image obtained by scanning the region of interest using CT after intravenous injection of a contrast agent. After computer processing of the image, the vascular system can be displayed in three dimensions. In addition, the input for training the preset deep learning model can be: all local vascular images calculated from the image set corresponding to the target layer of the test subject; or all local vascular images corresponding to all target layers of the test subject. Correspondingly, the target vascular image can be a tomographic image corresponding to the target layer obtained by CTA or a three-dimensional vascular image corresponding to the region of interest. The present invention provides a method for training a preset deep learning model based on CTA images of a region of interest. Because CTA images can intuitively and completely display the overall condition of the blood vessels, the CTA image of interest is used as the target result of the deep learning model to be trained, thereby improving the accuracy of the output results of the trained preset deep learning model and the completeness of the vascular image results.

[0079] In a second aspect, the present application provides a vascular imaging device for implementing the vascular imaging method provided by the present invention. As shown in FIG2 , the device includes: an image acquisition module 10 , an image calculation module 20 , and a deep learning module 30 .

[0080] The image acquisition module 10 is used to obtain multiple perfusion images obtained by CT scanning the target layer of the region of interest. ， An image set is determined from the plurality of perfusion images.

[0081] The image calculation module 20 is used to subtract the first image from the second image in the image set to obtain a corresponding local blood vessel image, where the second image is scanned later than the first image.

[0082] The deep learning module 30 is used to process all local blood vessel images according to a preset deep learning model to obtain a target blood vessel image.

[0083] In a third aspect, the present application provides an electronic device for implementing the vascular imaging method provided by the present invention. As shown in FIG3 , the electronic device includes a memory 44 and a processor 41. The memory 44 stores a computer program, and the processor 41 executes the computer program to implement the method described above. As shown in FIG4 , the electronic device also includes a bus 42 and a communication interface 43, wherein the processor 41, the communication interface 43, and the memory 44 are connected via the bus 42. The memory 44 may include a high-speed random access memory (RAM) or a non-volatile memory, such as at least one disk storage. The communication connection between the system network element and at least one other network element is achieved through at least one communication interface 43 (which may be wired or wireless), and the Internet, a wide area network, a local area network, a metropolitan area network, etc. may be used. Bus 42 can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, or an AMBA (Advanced Microcontroller Bus Architecture) bus. AMBA defines three types of buses, including an APB (Advanced Peripheral Bus) bus, an AHB (Advanced High-performance Bus) bus, and an AXI (Advanced eXtensible Interface) bus. Bus 42 can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, FIG4 shows only one bidirectional arrow, but this does not mean that there is only one bus or one type of bus.

[0084] The processor 41 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in the processor 41 or by software instructions. The above-mentioned processor 41 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly implemented as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory 44 , and the processor 41 reads the information in the memory 44 and implements the method shown in FIG. 1 in conjunction with its hardware.

[0085] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored. The processor 41 executes the computer program to implement the method as described above. For specific implementation, please refer to the method embodiment, which will not be repeated here.

[0086] 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 and devices can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0087] In addition, in the description of the embodiments of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to mechanical connections or electrical connections; they may refer to direct connections or indirect connections through an intermediate medium; and they may refer to internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0088] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a 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 codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0089] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0090] Finally, it should be noted that the above embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A dynamic vascular imaging method based on CT perfusion images, characterized in that: The method comprises: Acquire multiple perfusion images obtained by performing CT scanning on each target layer of the region of interest, and determine an image set based on the multiple perfusion images; defining each two adjacent images in the image set as an image group according to the scanning time sequence, defining the scanning time of the second image in each image group to be later than that of the first image, and subtracting the first image from the second image in each image group to obtain a corresponding local blood vessel image; Acquiring all the local blood vessel images corresponding to all the target layers; Obtaining blood vessel three-dimensional point cloud data based on all the local blood vessel images; The three-dimensional point cloud data of the blood vessels is input into the preset deep learning model to obtain the target blood vessel image.

2. The method according to claim 1, characterized in that Before the steps to define the image group, include: Determining a perfusion image in the image set as a reference perfusion image, and determining a first coordinate of a target feature point on the reference perfusion image; determining a second coordinate of the target feature point on other perfusion images in the image set except the reference image; Based on the first coordinates and the second coordinates, calculating a target transformation matrix for registering each of the other perfusion images to the reference perfusion image using a feature point matching method; performing registration on the reference perfusion image and each of the other perfusion images according to the target transformation matrix to obtain other registered perfusion images; All the registered other perfusion images and the reference perfusion image are taken as the image set.

3. The method according to claim 1, characterized in that The step of inputting the three-dimensional point cloud data of the blood vessel into a preset deep learning model to obtain a target blood vessel image includes: Acquiring original images obtained by performing CT scanning on all the target layers, wherein the original images are obtained by performing CT scanning on the target layers when the contrast agent does not flow into the target layers; Obtaining original three-dimensional point cloud data according to the original image; The original three-dimensional point cloud data and the blood vessel three-dimensional point cloud data are input into the preset deep learning model to obtain a target blood vessel image.

4. The method according to claim 1, wherein The step of processing all the local blood vessel images according to a preset deep learning model includes: The preset deep learning model includes a deformable convolutional layer; All the local blood vessel images are processed using the deformable convolution layer.

5. The method according to claim 4, characterized in that The preset deep learning model also includes: The preset deep learning model includes an attention gate, which is connected to the deformable convolutional layer.

6. The method according to claim 1, wherein The training steps of the preset deep learning model include: Acquiring training samples, wherein the training samples include acquiring a plurality of perfusion images obtained by performing CT scanning on a target layer of a region of interest of a test object; obtaining an image set corresponding to a target layer of the test object according to the multiple perfusion images; In the order of scanning time, each two adjacent images in the image set are defined as a Image groups, defining that a second image in each image group is scanned later than the first image, and subtracting the first image from the second image in each image group corresponding to the target layer of the test object to obtain a corresponding local blood vessel image; Inputting all local vascular images of the test object into an initial deep learning model to obtain a predicted vascular image; Acquiring a CTA image corresponding to a region of interest of the test object, and obtaining a corresponding target blood vessel image based on the CTA image; calculating a residual between the predicted blood vessel image and the target blood vessel image; Accumulating the residuals to obtain a loss function; The model is optimized according to the loss function until a preset condition is met to obtain the preset deep learning model.

7. The method according to claim 1, characterized in that The step of processing all the local blood vessel images according to a preset deep learning model to obtain a target blood vessel image further includes: determining, based on the plurality of perfusion images, a change characteristic of a contrast agent in a blood vessel of the region of interest, the change characteristic comprising one or more of a contrast agent arrival time, a contrast agent average transit time, a contrast agent peak time, a blood flow rate, a blood volume, a half-peak width, or a time length at 2 / 3 of the peak width; All the local blood vessel images and the change features are input into the deep learning model to obtain the target blood vessel image.

8. A vascular imaging device for implementing the vascular imaging method provided by any one of claims 1 to 8, characterized in that: The device comprises: Image acquisition module, used to obtain the target layer of the area of interest by CT scanning a plurality of perfusion images, and determining an image set based on the plurality of perfusion images; an image calculation module, configured to define, in order of scanning time, every two adjacent images in the image set as an image group, define a scanning time of the second image in each image group to be later than that of the first image, and subtract the first image from the second image in each image group corresponding to the target layer of the test object to obtain a corresponding local vascular image; A deep learning module is used to process all the local blood vessel images according to a preset deep learning model to obtain a target blood vessel image.

Images