Perfusion image analysis method, device, equipment and storage medium

Abstract

The present application relates to a perfusion image analysis method, apparatus, device, and storage medium. The method includes: acquiring a brain perfusion image and determining multiple arterial segments in the brain in the brain perfusion image; using a preset segmentation network to segment the blood supply area in the brain perfusion image to determine the multiple divided blood supply areas, wherein the segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image; matching each blood supply area with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment; constructing a brain perfusion model based on the matching relationship to analyze the brain perfusion image. The use of this method can reduce the error of the obtained arterial input function AIF, thereby reducing the error of the results of the perfusion image analysis.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a perfusion image analysis method, apparatus, device and storage medium. Background Art

[0002] CT (Computed Tomography) perfusion imaging is different from dynamic scanning. It continuously scans the region of interest when a contrast agent is injected intravenously, thereby obtaining a time-density curve of the region of interest (a graph showing the change of CT value of the region of interest over time, which can be recorded as a TDC curve). The time-density curve can then be used to calculate various perfusion parameter maps. These perfusion parameter maps can be used to better analyze the region of interest in the CT perfusion image.

[0003] To obtain a time-density curve for a region of interest (ROI), it's generally necessary to first locate the artery to obtain a reference time-density curve, specifically the arterial input function (AIF). This function is then used to calculate the ROI's time-density curve. In related arteries, obtaining the AIF by locating the artery typically involves generating Voronoi regions for each artery using a distance transform from a segmented image of the artery. Each Voronoi region then uses the corresponding arterial time-density curve as the corresponding AIF.

[0004] However, the arterial input function AIF obtained by the above technology has a large error problem, which may cause the results of subsequent perfusion image analysis to also have a large error problem. Summary of the Invention

[0005] Based on this, it is necessary to provide a perfusion image analysis method, device, equipment and storage medium that can reduce the error of the obtained arterial input function AIF and thus reduce the error of the result of perfusion image analysis to address the above technical problems.

[0006] In a first aspect, the present application provides a perfusion image analysis method, the method comprising:

[0007] acquiring a brain perfusion image and determining a plurality of arterial blood vessel segments of the brain in the brain perfusion image;

[0008] A preset segmentation network is used to segment the blood supply areas in the brain perfusion image to determine multiple divided blood supply areas; the segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image;

[0009] Matching each blood supply area with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment;

[0010] A brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion images.

[0011] In one embodiment, matching each blood supply area with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment includes:

[0012] determining first characteristic information corresponding to each blood supply area according to the divided multiple blood supply areas;

[0013] Determining second characteristic information corresponding to each arterial blood vessel segment;

[0014] According to each first characteristic information and each second characteristic information, each blood supply area and each arterial blood vessel segment are matched to determine a matching relationship between each blood supply area and each arterial blood vessel segment.

[0015] In one embodiment, matching each blood supply area and each arterial segment based on each first characteristic information and each second characteristic information to determine the matching relationship between each blood supply area and each arterial segment includes:

[0016] Using a preset cost function to calculate the cost function value between each first feature information and each second feature information, to obtain multiple cost function values ​​corresponding to each blood supply area;

[0017] determining a target arterial segment corresponding to each blood supply area according to a plurality of cost function values ​​corresponding to each blood supply area;

[0018] A matching relationship is established between each blood supply area and the corresponding target arterial segment.

[0019] In one embodiment, determining the target arterial segment corresponding to each blood supply area according to the multiple cost function values ​​corresponding to each blood supply area includes:

[0020] For each blood supply area, determining a minimum cost function value from a plurality of cost function values ​​corresponding to the blood supply area;

[0021] The arterial segment corresponding to the minimum function value is determined as the target arterial segment corresponding to the blood supply area; wherein the target arterial segment corresponding to each blood supply area is different.

[0022] In one embodiment, the determining of the first characteristic information corresponding to each blood supply area according to the divided multiple blood supply areas includes:

[0023] Calculating an average time density curve corresponding to each blood supply area, and determining at least one curve characteristic value corresponding to each blood supply area based on each average time density curve;

[0024] Determine the relative centroid position of each blood supply area based on the centroid position of each blood supply area and the centroid position of the brain;

[0025] At least one curve characteristic value and a relative centroid position of each blood supply area are determined as first characteristic information of each blood supply area.

[0026] In one embodiment, determining a plurality of arterial segments in the brain in the brain perfusion image includes:

[0027] Divide the brain into the first hemisphere and the second hemisphere;

[0028] determining a target side of the brain from the first side of the brain and the second side of the brain; the target side of the brain being the side opposite to the side where the lesion occurs;

[0029] Multiple arterial vessel segments in the target side of the brain in the brain perfusion image are determined.

[0030] In one embodiment, determining the target side of the brain from the first side of the brain and the second side of the brain includes:

[0031] calculating a first peak time of the first side of the brain and calculating a second peak time of the second side of the brain;

[0032] A target side of the brain is determined from the first side of the brain and the second side of the brain according to the first peak time and the second peak time.

[0033] In one embodiment, determining a plurality of arterial vessel segments in the target-side brain in the brain perfusion image includes:

[0034] Extracting and processing blood vessels in the target side of the brain in the brain perfusion image to determine a blood vessel image; the blood vessel image includes the blood vessels to be classified;

[0035] Blood vessel segment clustering is performed on each voxel on the blood vessel to be classified to determine multiple clustered arterial blood vessel segments.

[0036] In one embodiment, the blood vessel segment clustering process is performed on each voxel on the blood vessel to be classified to determine a plurality of clustered arterial blood vessel segments, including:

[0037] Extract features from each voxel on the blood vessel to be classified to obtain extracted features corresponding to each voxel; the extracted features include spatial position information of the voxel;

[0038] According to the extracted features of each voxel, each voxel is clustered into blood vessel segments to determine multiple clustered arterial blood vessel segments.

[0039] In one embodiment, the feature extraction of each voxel on the blood vessel to be classified to obtain the extracted features corresponding to each voxel includes:

[0040] Obtaining the time density curve of each voxel on the blood vessel to be classified;

[0041] Calculating at least one curve characteristic value of each voxel according to the time density curve of each voxel;

[0042] At least one curve feature value and spatial position information of each voxel are determined as the extracted features of each voxel.

[0043] In a second aspect, the present application further provides a perfusion image analysis device, comprising:

[0044] an acquisition module, configured to acquire a brain perfusion image and determine a plurality of arterial vessel segments of the brain in the brain perfusion image;

[0045] a segmentation module for segmenting the blood supply areas in the brain perfusion image using a preset segmentation network to determine a plurality of divided blood supply areas; the segmentation network is trained based on a plurality of training perfusion images and a reference blood supply area corresponding to each training perfusion image;

[0046] A matching module is used to match each blood supply area with each arterial segment and determine the matching relationship between each blood supply area and each arterial segment;

[0047] The analysis module is used to construct a brain perfusion model based on the matching relationship to analyze the brain perfusion image.

[0048] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0049] acquiring a brain perfusion image and determining a plurality of arterial blood vessel segments of the brain in the brain perfusion image;

[0050] A preset segmentation network is used to segment the blood supply areas in the brain perfusion image to determine multiple divided blood supply areas; the segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image;

[0051] Matching each blood supply area with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment;

[0052] A brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion images.

[0053] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0054] acquiring a brain perfusion image and determining a plurality of arterial blood vessel segments of the brain in the brain perfusion image;

[0055] A preset segmentation network is used to segment the blood supply areas in the brain perfusion image to determine multiple divided blood supply areas; the segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image;

[0056] Matching each blood supply area with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment;

[0057] A brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion images.

[0058] In a fifth aspect, the present application further provides a computer program product, the computer program product comprising a computer program, which, when executed by a processor, implements the following steps:

[0059] acquiring a brain perfusion image and determining a plurality of arterial blood vessel segments of the brain in the brain perfusion image;

[0060] A preset segmentation network is used to segment the blood supply areas in the brain perfusion image to determine multiple divided blood supply areas; the segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image;

[0061] Matching each blood supply area with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment;

[0062] A brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion images.

[0063] The above-mentioned perfusion image analysis method, device, equipment and storage medium obtain a brain perfusion image and determine multiple arterial segments in the brain. Then, a preset segmentation network is used to segment the blood supply area in the brain perfusion image to determine the multiple divided blood supply areas, and each blood supply area is matched with each arterial segment respectively to determine the matching relationship between each blood supply area and each arterial segment. Then, a brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion image; wherein, the segmentation network is trained based on multiple training perfusion images and their corresponding reference blood supply areas. In this method, since the segmentation model can be trained through the labeled reference blood supply areas, the trained segmentation model can obtain more accurate blood supply area segmentation results when segmenting brain perfusion images. Therefore, when matching each blood supply area with each arterial segment, a more accurate matching result can be obtained. In this way, an accurate arterial input function AIF can be established based on the accurate matching result, and a more accurate brain perfusion model can be constructed through the accurate arterial input function AIF, thereby reducing the error of the results of the brain perfusion image analysis and improving the accuracy of the results of the brain perfusion image analysis using the brain perfusion model. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 is a diagram of the internal structure of a computer device in one embodiment;

[0065] Figure 2 1 is a schematic flow chart of a perfusion image analysis method according to an embodiment;

[0066] Figure 3 is a schematic flow chart of a perfusion image analysis method according to another embodiment;

[0067] Figure 4 is a schematic flow chart of a perfusion image analysis method according to another embodiment;

[0068] Figure 5 is a schematic flow chart of a perfusion image analysis method according to another embodiment;

[0069] Figure 6 A time-to-peak characteristic diagram provided in another embodiment;

[0070] Figure 7 is a schematic flow chart of a perfusion image analysis method according to another embodiment;

[0071] Figure 8 is an example diagram of spatial position information of voxels on a blood vessel to be classified in another embodiment;

[0072] Figure 9 is a schematic diagram of a specific process of a perfusion image analysis method in another embodiment;

[0073] Figure 10 It is an enlarged view of a time density curve in a specific flow diagram of a perfusion image analysis method in another embodiment;

[0074] Figure 11 FIG. 4 is a structural block diagram of a perfusion image analysis device in one embodiment. DETAILED DESCRIPTION

[0075] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0076] Currently, the process of obtaining time-density curves for regions of interest (ROIs) generally requires first locating arterial vessels to obtain a reference time-density curve, i.e., obtaining the arterial input function (AIF). The AIF is then used to calculate the ROI's time-density curve. When locating arteries to obtain the AIF, the mainstream approach for cerebral perfusion analysis uses a single arterial point, meaning that the perfusion model for the entire brain tissue is approximately calculated using a single vessel as a reference. This introduces errors into the model. One such method clusters the TDC curves of each spatial point during vascular localization and selects the cluster with the smallest cluster center as the final arterial point. However, this method does not consider the spatial location of the vessels and cannot distinguish between vessels corresponding to different blood supply zones. Alternatively, another approach constrains the extracted AIF curves by constraining them using a model based on prior knowledge of eigenvalues ​​such as vessel segmentation, peak value, peak width, and curve area. This approach can ensure the completeness or regularity of the extracted AIF curves. However, the drawback is that a single arterial point is used as an approximate blood supply for the entire brain tissue, which can introduce errors. In addition, there is another way to segment the artery by using the difference between the arterial curve and other tissue curves, and then generate the blood supply area of ​​each artery segment based on the Voronoi diagram, so that each blood supply area can use the blood vessel curve of the corresponding segment as AIF, thereby achieving the purpose of local AIF; this method uses multiple arterial points to provide reference for nearby tissues, which is an improvement over the method of a single arterial point, but the disadvantage is that its blood supply area is only a false blood supply area generated by "being close to a certain blood vessel" in the Voronoi diagram, and it cannot match the blood supply area in the true clinical sense, and there are also errors. At the same time, none of the above technologies consider the problem of the abnormal side of the brain, and the points selected for the cerebral perfusion model need to be selected on the "non-infarct side", that is, the abnormal cerebral blood vessels on the side of the brain must be excluded, otherwise it will also introduce errors into the cerebral perfusion model. Based on this, the embodiments of the present application provide a perfusion image analysis method, device, equipment and storage medium that can solve the above technical problems.

[0077] The perfusion image analysis method provided in the embodiment of the present application can be applied to a computer device, which can be a terminal or a server. Taking the terminal as an example, its internal structure diagram can be as follows: Figure 1 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a perfusion image analysis method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0078] Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0079] In one embodiment, Figure 2 As shown, a perfusion image analysis method is provided, which is applied to Figure 1 Taking the computer device in the example as an example, the method may include the following steps:

[0080] S202: Acquire a brain perfusion image and determine a plurality of arterial vessel segments in the brain in the brain perfusion image.

[0081] In this step, the brain perfusion image can be an image obtained by performing continuous dynamic scanning of the selected region of interest while a contrast agent is injected intravenously into the brain. The brain perfusion image can be a CT image, an MR (Magnetic Resonance) image, such as a CTP image / CTP data. In addition, the brain perfusion image can be a four-dimensional image (i.e., a 4D image), which can be composed of multiple three-dimensional brain images with time information (e.g., composed of multiple three-dimensional brain sequences). Of course, the brain perfusion image can also be a three-dimensional image or a two-dimensional image.

[0082] After obtaining a brain perfusion image, which includes the brain and arteries therein, the brain arteries in the brain perfusion image can be extracted to obtain multiple arterial segments. It should be noted that the arterial extraction can be performed for the entire brain or for a portion of the brain, such as for a single side of the brain.

[0083] When extracting arteries, threshold extraction can be performed based on the morphological features of the CT value and TDC curve, or arteries can be extracted using a deep learning model or neural network model trained with labeled data. Alternatively, other methods can be used for extraction, which are not specifically limited here, as long as multiple arterial segments in the brain can be extracted from the brain perfusion image.

[0084] Furthermore, after obtaining the brain perfusion image, the image can also be preprocessed to obtain a preprocessed brain perfusion image. This preprocessing may include, but is not limited to, registration, alignment, skull removal, and cerebrospinal fluid removal. Subsequent blood supply segmentation and cerebral artery extraction can use either the original brain perfusion image or the preprocessed brain perfusion image.

[0085] S204, using a preset segmentation network to segment the blood supply area in the brain perfusion image to determine multiple divided blood supply areas; the above segmentation network is trained based on multiple training perfusion images and the reference blood supply area corresponding to each training perfusion image.

[0086] In this step, a segmentation network can be pre-trained to segment the blood supply area in the brain perfusion image. The type and specific architecture of the segmentation network can be set according to actual conditions, such as a Vnet network.

[0087] When training a segmentation network, multiple training perfusion images may be first acquired, each of which may include a sample brain. Multiple medically defined blood supply regions may be pre-annotated in each training perfusion image to achieve annotation of the true blood supply regions in each training perfusion image. The annotated blood supply regions may be recorded as reference blood supply regions, and the specific annotation may be performed using blood supply region masks, blood supply region positions, or labels. Each training perfusion image may then be input into an initial segmentation network, which segments the blood supply regions in each training perfusion image to obtain multiple predicted blood supply regions corresponding to each training perfusion image. The loss between each predicted blood supply region and the corresponding reference blood supply region in each training perfusion image may then be calculated, and the parameters of the initial segmentation network may be adjusted based on the loss until the loss reaches a threshold or stabilizes, thereby obtaining a trained segmentation network.

[0088] In addition, after the segmentation network is trained, the training perfusion images can also be used to perform prediction tests on the trained segmentation network. The specific testing process is not specifically limited here.

[0089] After the segmentation network is trained, the brain perfusion image obtained above can be input into the trained segmentation network to segment the blood supply regions, thereby obtaining the segmented blood supply regions. The segmented blood supply regions obtained here can be one or more segmentation masks, wherein the segmentation masks can include the segmented blood supply regions and the background.

[0090] For example, in this embodiment, the blood supply area can be divided into four, such as the anterior artery blood supply area, the middle artery blood supply area, the posterior artery blood supply area, and the cerebellar brainstem blood supply area. Of course, in actual situations, the division can be finer, that is, divided into more blood supply areas, and after segmentation using the segmentation network, the corresponding areas can be merged into the above four blood supply areas.

[0091] In this step, since the segmentation network is trained based on clearly defined medically labeled blood supply areas, the blood supply areas obtained by the trained segmentation network after segmenting the brain perfusion image correspond to the actual blood supply areas in medicine, that is, the real blood supply areas. In this way, the subsequent matching of blood supply areas with arterial segments can be more accurate.

[0092] S206: Match each blood supply area with each arterial blood vessel segment to determine a matching relationship between each blood supply area and each arterial blood vessel segment.

[0093] Among them, the arterial segments mainly supply blood to the blood supply areas, so it is necessary to match the blood supply areas and the arterial segments. After obtaining multiple arterial segments in the brain perfusion image and segmenting to obtain multiple blood supply areas, each arterial segment can be matched with each blood supply area.

[0094] Specifically, when matching arterial segments and blood supply areas, the similarities between the features of each blood supply area and the features of each arterial segment can be compared, and arterial segments and blood supply areas with relatively similar features can be considered as a matching set of arterial segments and blood supply areas. Alternatively, the distance (such as the Euclidean distance) between each blood supply area and each arterial segment can be calculated, and then a matching arterial segment can be found for each blood supply area based on the calculated distance value. Other matching methods can also be used, which are not specifically limited here.

[0095] Through the above matching method, a matching arterial segment can be determined for each blood supply region. Each blood supply region is then matched to the corresponding arterial segment, resulting in a matching relationship between each blood supply region and each arterial segment. It should be noted that the matching relationship obtained above is a one-to-one matching relationship, and the number of arterial segments obtained is generally the same as the number of segmented blood supply regions. That is, one blood supply region is generally matched to one arterial segment, and different blood supply regions are generally matched to different arterial segments.

[0096] For example, there are four blood supply areas divided as above, namely: blood supply area 1, blood supply area 2, blood supply area 3, and blood supply area 4; there are also four extracted arterial segments, namely: anterior artery, middle artery, posterior artery and cerebellar artery. Then the matching relationship obtained after matching here can be, for example, the anterior artery corresponds to blood supply area 2, the middle artery corresponds to blood supply area 3, the posterior artery corresponds to blood supply area 1, and the cerebellar artery corresponds to blood supply area 4.

[0097] In addition, it should be noted that the arterial segment that matches the blood supply area can be an arterial segment of the entire brain or an arterial segment of a certain brain.

[0098] S208: Construct a brain perfusion model based on the matching relationship to analyze the brain perfusion image.

[0099] In this step, after obtaining the matching relationship between each blood supply area and each arterial segment as described above, the arterial segment that supplies blood to each blood supply area can be determined, and then a cerebral perfusion model can be constructed based on this. Specifically, the time-density curve of each arterial segment can be used as the AIF of the corresponding blood supply area. After that, a parameter graph can be calculated based on the AIF of each blood supply area and the time-density curve of each arterial segment, and the calculated parameter graph can be used to analyze the cerebral perfusion image to obtain the analysis result.

[0100] In the above-mentioned perfusion image analysis method, a brain perfusion image is obtained and multiple arterial segments in the brain are determined. Then, a preset segmentation network is used to segment the blood supply areas in the brain perfusion image to determine the multiple divided blood supply areas, and each blood supply area is matched with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment. Then, a brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion image; wherein, the segmentation network is trained based on multiple training perfusion images and their corresponding reference blood supply areas. In this method, since the segmentation model can be trained through the labeled reference blood supply areas, the trained segmentation model can obtain more accurate blood supply area segmentation results when segmenting brain perfusion images. Therefore, when matching each blood supply area with each arterial segment, a more accurate matching result can be obtained. In this way, an accurate arterial input function AIF can be established based on the accurate matching result, and a more accurate brain perfusion model can be constructed through the accurate arterial input function AIF, thereby reducing the error of the results of the brain perfusion image analysis and improving the accuracy of the results of the brain perfusion image analysis using the brain perfusion model.

[0101] The above embodiments mentioned that each blood supply area can be matched with each arterial segment in a variety of ways. The following embodiments illustrate the process of matching based on features.

[0102] In another embodiment, another perfusion image analysis method is provided. Based on the above embodiment, Figure 3 As shown, the above S206 may include the following steps:

[0103] S302: Determine first characteristic information corresponding to each blood supply area according to the divided multiple blood supply areas.

[0104] After the brain perfusion image is segmented to obtain the divided blood supply areas, characteristic information of each blood supply area can be calculated. As an optional embodiment, the following steps A1-A3 can be used for calculation:

[0105] Step A1: Calculate the average time density curve corresponding to each blood supply area, and determine at least one curve characteristic value corresponding to each blood supply area based on each average time density curve.

[0106] As mentioned above, the brain perfusion image can be composed of a combination of multiple three-dimensional brain images with time information, each of which includes various blood supply areas. The time-density curve of each blood supply area can be obtained through each three-dimensional brain image, that is, multiple time-density curves can be obtained for each blood supply area; then, the multiple time-density curves of each blood supply area can be averaged to obtain the average time-density curve corresponding to each blood supply area.

[0107] Afterwards, the peak value, half-peak width, time to peak and other curve characteristic values ​​corresponding to the average time-density curve of each blood supply area can be calculated through the average time-density curve of each blood supply area. The calculation process of the peak value, half-peak width, time to peak and other curve characteristic values ​​will be specifically described in the following embodiments.

[0108] Step A2: Determine the relative centroid position of each blood supply area based on the centroid position of each blood supply area and the centroid position of the brain.

[0109] In this step, the center of mass of the brain can be calculated in advance using the brain perfusion image. After obtaining each blood supply region in the brain perfusion image, the center of mass of each blood supply region can also be calculated. The center of mass of the brain can then be subtracted from the center of mass of each blood supply region to obtain the relative center of mass of each blood supply region. Calculating the relative center of mass for each blood supply region here can make the obtained spatial position of each blood supply region more accurate. Furthermore, the center of mass of the blood supply region, as well as the center of mass and relative center of mass of the brain, can all be expressed using coordinates.

[0110] Step A3: determining at least one curve characteristic value and relative centroid position of each blood supply area as first characteristic information of each blood supply area.

[0111] After obtaining the curve characteristic value and relative centroid position of each blood supply area as described above, the curve characteristic value and relative centroid position of each blood supply area can be used as the first characteristic information of the corresponding blood supply area.

[0112] S304: Determine second characteristic information corresponding to each arterial blood vessel segment.

[0113] In this step, after obtaining each arterial segment through the brain perfusion image, the second characteristic information of each arterial segment can also be obtained according to the above-mentioned characteristic acquisition method of each blood supply area. For example, it can be obtained by the following steps: calculating the average time density curve corresponding to each arterial segment, and determining at least one curve characteristic value corresponding to each arterial segment based on each average time density curve; determining the relative center of mass position corresponding to each arterial segment based on the center of mass position of each arterial segment and the center of mass position of the brain; and determining at least one curve characteristic value and the relative center of mass position of each arterial segment as the second characteristic information of each arterial segment.

[0114] S306 , matching each blood supply area with each arterial blood vessel segment according to each first characteristic information and each second characteristic information, and determining a matching relationship between each blood supply area and each arterial blood vessel segment.

[0115] In this step, after obtaining the first characteristic information corresponding to each blood supply area and the second characteristic information corresponding to each arterial blood vessel segment, the similarity between each first characteristic information and each second characteristic information can be calculated, and the matching relationship between each blood supply area and each arterial blood vessel segment can be established through the calculated similarities.

[0116] In this embodiment, each blood supply region and each arterial segment is matched based on the determined characteristic information of each blood supply region and the characteristic information of each arterial segment, thereby establishing a matching relationship between each blood supply region and each arterial segment. This feature information matching method can refine the matching relationship establishment process, thereby improving the accuracy of the ultimately determined matching relationship between the blood supply region and the arterial segment. Furthermore, by calculating the curve characteristic values ​​and relative spatial position of each blood supply region or arterial segment as characteristic information, this refined and enriched feature can further improve the accuracy of subsequent matching between blood supply regions and arterial segments.

[0117] The following embodiments illustrate the above process of matching blood supply areas and arterial segments using feature information.

[0118] In another embodiment, another perfusion image analysis method is provided. Based on the above embodiment, Figure 4 As shown, the above S306 may include the following steps:

[0119] S402: Calculate the cost function value between each first feature information and each second feature information using a preset cost function to obtain a plurality of cost function values ​​corresponding to each blood supply area.

[0120] The formula of the preset cost function can be shown in the following formula (1):

[0121]

[0122] Among them, C represents the value of the cost function; H partition 、W partition 、A partition Respectively represent the peak value, half-peak width, peak time and other characteristic values ​​corresponding to the average time density curve of a single blood supply area; X partition 、Y partition , Z partition Respectively represent the relative centroid position of a single blood supply area; H vessel 、W vessel 、A vessel Respectively represent the peak value, half-peak width, peak time and other characteristic values ​​corresponding to the average time density curve of a single arterial segment; X vessel 、Y vessel , Z vesselThey respectively represent the relative centroid position of a single arterial segment; k1-k6 are empirical values, which can be set according to actual conditions and are known quantities.

[0123] By respectively bringing the first characteristic information of each blood supply area (including peak value, half-peak width, time to peak value, relative centroid position) and the second characteristic information of each arterial segment (including peak value, half-peak width, time to peak value, relative centroid position) into the above formula (1) for calculation, the cost function value between each blood supply area and each arterial segment can be obtained. That is, multiple cost function values ​​can be calculated for each blood supply area. For example, if there are 4 arterial segments, 4 cost function values ​​will be calculated for each blood supply area.

[0124] S404 : Determine a target arterial segment corresponding to each blood supply area according to a plurality of cost function values ​​corresponding to each blood supply area.

[0125] In this step, after obtaining multiple cost function values ​​for each blood supply area, as an optional embodiment, the minimum cost function value can be determined for each blood supply area from the multiple cost function values ​​corresponding to the blood supply area; the arterial segment corresponding to the minimum function value is determined as the target arterial segment corresponding to the blood supply area; wherein, the target arterial segment corresponding to each blood supply area is different.

[0126] In other words, the minimum cost function value can be selected from the multiple cost function values ​​for each blood supply region. The arterial segment corresponding to this minimum cost function value is the arterial segment that matches the blood supply region, and thus this matching arterial segment can be used as the target arterial segment for the blood supply region. In this way, a matching target arterial segment can be found for each blood supply region, where each blood supply region corresponds to a target arterial segment, and each blood supply region corresponds to a different target arterial segment.

[0127] S406: Establish a matching relationship between each blood supply area and the corresponding target arterial vessel segment.

[0128] In this step, after determining the matching target arterial segment for each blood supply area, each blood supply area and its matching target arterial segment can be bound together to obtain a matching relationship between each blood supply area and the corresponding target arterial segment.

[0129] In this embodiment, a preset cost function is used to calculate the cost function value between each blood supply region and each arterial segment. Based on this cost function, the target arterial segment corresponding to each blood supply region is determined, and a matching relationship is established. This cost function allows for rapid and accurate determination of the target arterial segment corresponding to each blood supply region, thereby improving the efficiency and accuracy of establishing matching relationships. Furthermore, by selecting the minimum cost function value from multiple cost function values ​​for each blood supply region and determining the target arterial segment based on this value, the selected target arterial segment and the corresponding blood supply region more closely match the actual spatial positional relationship. Consequently, the subsequently established matching relationship between the blood supply region and the arterial segment is more realistic and authentic.

[0130] The above embodiments mentioned that all arteries in a brain perfusion image can be extracted, or arteries in part of the brain can be extracted. When extracting arteries in part of the brain, it is first necessary to determine which part of the brain arteries are to be extracted. The following embodiments illustrate this process.

[0131] In another embodiment, another perfusion image analysis method is provided. Based on the above embodiment, Figure 5 As shown, the step of determining multiple arterial vessel segments in the brain in the brain perfusion image in S202 may include the following steps:

[0132] S502: Divide the brain into a first side of the brain and a second side of the brain.

[0133] In this step, the brain perfusion image obtained above generally includes the entire brain. Here, the brain can be divided into left brain tissue and right brain tissue, which are respectively recorded as the first side of the brain and the second side of the brain.

[0134] The specific method of dividing the brain may be achieved by using a hemispheric segmentation model, or by segmenting or dividing the brain according to the center of mass position / center of mass coordinates of the brain tissue, or by other methods, which are not specifically limited here.

[0135] S504: determining a target side of the brain from the first side of the brain and the second side of the brain; the target side of the brain is the side of the brain opposite to the side where the brain lesion occurs.

[0136] In this step, after obtaining the divided first and second side brain parts, the side of the brain without lesions can be found and the side of the brain without lesions can be used as the target side of the brain, that is, the normal side of the brain or the side without abnormalities can be used as the target side of the brain.

[0137] Since the peak time may be abnormal when a lesion exists, when determining the target side of the brain based on whether a lesion exists, the first peak time of the first side of the brain and the second peak time of the second side of the brain can be optionally calculated; based on the first peak time and the second peak time, the target side of the brain is determined from the first side of the brain and the second side of the brain.

[0138] As mentioned above, the brain perfusion image can be a four-dimensional (ie, 4D) image, that is, including the three spatial axes xyz and time axis t. Here, the brain perfusion image can be denoted as I x,y,z,t , where x, y, and z are the coordinates of the X, Y, and Z axes respectively, and t is the time coordinate relative to the first scan time in the time series. The peak time characteristic diagram can be expressed using the following formula (2):

[0139]

[0140] Where, T represents the acquisition time of brain perfusion images; argmax represents the maximum peak time in brain perfusion images; A x,y,z Feature graph showing the time to peak.

[0141] The peak time characteristic map corresponding to the brain perfusion image can be calculated by the above formula (2), see Figure 6 The time-to-peak characteristic map shown in FIG. 1 may include the time-to-peak of each point in the brain perfusion image. Based on the aforementioned division of the two brain regions, the average time-to-peak of each side of the brain may be calculated. The average time-to-peak of each side of the brain may be calculated by summing and averaging the time-to-peak of each point corresponding to that side of the brain in the time-to-peak characteristic map to obtain the average time-to-peak of that side of the brain.

[0142] After obtaining the average peak time of each side of the brain, for example, the first peak time and the second peak time can be recorded. Since the peak time may be delayed when a lesion occurs in the brain, that is, the peak time will become longer, the larger peak time can be selected from the two peak times, and the side of the brain corresponding to the larger peak time can be used as the side of the brain where the lesion occurs, and the side of the brain corresponding to the smaller peak time can be used as the target side of the brain.

[0143] S506: Determine multiple arterial vessel segments in the target-side brain in the brain perfusion image.

[0144] In this step, after the target side of the brain is determined, multiple arterial vessels of the target side of the brain in the brain perfusion image can be extracted to obtain multiple arterial vessel segments of the target side of the brain.

[0145] When obtaining multiple arterial segments of the target side of the brain, the arteries of the entire brain can be extracted, and then the arterial segments of the target side of the brain can be selected from the extracted arteries; or the arterial segments of the target side of the brain can be directly extracted.

[0146] In this embodiment, by dividing the brain into two sides and identifying the side without the lesion, multiple arterial segments on the side without the lesion are then determined. This method reduces the computational complexity of arterial extraction and prevents arterial points from being selected on the wrong side of the brain. This allows the subsequent brain perfusion model to be established on the side without the lesion. This eliminates the impact of abnormal brain vessels on the accuracy of the brain perfusion model, further improving the accuracy of the subsequent brain perfusion model. Furthermore, the peak time of the two sides is used to determine the side without the lesion. This simple and intuitive process can effectively improve the efficiency of identifying the target side of the brain, further improving the efficiency of brain perfusion model establishment and brain perfusion image analysis.

[0147] The following examples illustrate the extraction process of the arterial vessel segments mentioned above.

[0148] In another embodiment, another perfusion image analysis method is provided. Based on the above embodiment, Figure 7 As shown, the above S506 may include the following steps:

[0149] S602 , extracting and processing blood vessels in the target side of the brain in the brain perfusion image to determine a blood vessel image; the blood vessel image includes the blood vessels to be classified.

[0150] In this step, after the target side of the brain is determined, the corresponding image of the target side of the brain can be determined from the brain perfusion image, and then the arteries in the image of the target side of the brain can be extracted to determine the vascular image.

[0151] It should be noted that the vascular image here includes the extracted arterial vessels, but the arterial vessels are only the extracted blood vessels in the brain and have not been classified, so they can be recorded as vessels to be classified.

[0152] S604 , performing vessel segment clustering processing on each voxel on the vessel to be classified, and determining a plurality of clustered arterial vessel segments.

[0153] The extracted blood vessels to be classified include a plurality of voxels. The blood vessels to be classified may be clustered based on the voxels on the blood vessels to be classified to achieve classification of the blood vessels to be classified.

[0154] As an optional embodiment, the clustering process may be performed using the following steps B1 and B2:

[0155] Step B1: extract features from each voxel on the blood vessel to be classified to obtain extracted features corresponding to each voxel; the extracted features include spatial position information of the voxels.

[0156] Step B2: performing blood vessel segment clustering processing on each voxel based on the extracted features of each voxel, and determining a plurality of clustered arterial blood vessel segments.

[0157] When extracting features from each voxel on the blood vessel to be classified, optionally, a time density curve of each voxel on the blood vessel to be classified may be obtained; at least one curve characteristic value of each voxel may be calculated based on the time density curve of each voxel; and at least one curve characteristic value and spatial position information of each voxel may be determined as the extracted features of each voxel.

[0158] After obtaining each voxel on the vessel to be classified, the time density curve of each voxel on the vessel to be classified can also be obtained through the brain perfusion image. Then, the curve characteristic value of each voxel can be calculated based on the time density curve of each voxel. The curve characteristic value here can include characteristic values ​​such as peak value, half-peak width, and time to peak. The calculation of peak value, half-peak width, and time to peak can be specifically calculated using the following formula (3):

[0159]

[0160] Among them, I x,y,z,t represents the brain perfusion image; x, y, z represent the spatial coordinates of a voxel on the blood vessel to be classified on the x, y, z axes; t represents the time coordinate of each sequence in the brain perfusion image, which is the time coordinate relative to the first scan time in the time sequence; T represents the acquisition time of the brain perfusion image; H x,y,z represents the peak value of the (x, y, z) voxel; W x,y,z A represents the half-maximum width of the (x, y, z) voxel; x,y,z represents the time to peak of the (x, y, z) voxel; max represents the maximum value in the brain perfusion image; argmax represents the maximum time to peak in the brain perfusion image.

[0161] The above formula (3) can be used to calculate the corresponding peak value, half-width, time to peak and other curve characteristic values ​​for each voxel. Then, the spatial coordinates of each voxel can be obtained from the brain perfusion image, and the center of mass position of the brain tissue can also be obtained. The spatial coordinates of each voxel can then be normalized by the center of mass position to obtain the normalized spatial position information. The specific normalization can be processed using the following formula (4):

[0162] X x,y,z =xx mass

[0163] Y x,y,z =yy mass (4)

[0164] Z x,y,z =zz mass

[0165] Among them, x mass ,y mass , z mass represents the center of mass of the brain tissue, x, y, z represent the spatial coordinates of the voxel, X x,y,z 、Y x,y,z , Z x,y,z Represents the spatial position information of the voxel after normalization.

[0166] For example, see Figure 8 The example diagram of the spatial position information of voxels is shown, where the white box represents the spatial coordinates of the voxels in the blood vessels to be classified, and the solid white dot represents the center of mass of the brain tissue.

[0167] The normalized spatial position information of each voxel can be calculated by the above formula (4), and then the curve characteristic value and spatial position information of each voxel can be used as the extracted features of the voxel.

[0168] Then, based on the extracted features of each voxel, a preset clustering algorithm (such as the K-means clustering algorithm) can be used to directly cluster the extracted features of each voxel. Alternatively, the extracted features of each voxel can be weighted to obtain weighted features, such as [K1*H, K2*W, K3*A, K4*X, K5*Y, K6*Z], where K1-K6 are empirical values ​​that can be determined experimentally, H represents the peak value, W represents the half-width, A represents the time to peak value, and X, Y, and Z represent the spatial location information of the voxel. Ultimately, each voxel can be clustered into multiple arterial segments based on the extracted features or weighted features of each voxel.

[0169] For example, assuming that the K-means clustering algorithm is clustered into four categories, the resulting clustered arterial segments may be, for example, the anterior artery, middle artery, posterior artery, and arteriole. Of course, clustering can also be performed into more or fewer categories, and the specific number of categories can be set based on actual conditions. Ultimately, clustered arterial segments are obtained.

[0170] In this embodiment, the blood vessels in the target lateral brain are extracted to obtain the vessels to be classified and the voxels thereon. These voxels are then clustered to obtain multiple arterial segments. Clustering the voxels within the vessels to be classified refines the clustering granularity, thereby improving clustering accuracy. Furthermore, clustering is performed based on the spatial positional characteristics of the voxels within the vessels to be classified. This consideration of the spatial position of the vascular voxels ensures that the clustered vessels are more consistent with the spatial position of the actual vessels, further improving the accuracy of the clustered arterial segments. Furthermore, the extracted features of each voxel can be derived from its curve eigenvalues ​​and spatial position information for clustering. Using multiple types of features enriches the feature information of the voxels, effectively clustering arterial segments and further improving clustering accuracy.

[0171] The following is a detailed example to illustrate the technical solution of this application. Figure 9 The detailed process example diagram shown and see Figure 10 The time density curve is shown in the enlarged view. Figure 9 Based on the time density curve in the example graph, the method may include the following steps:

[0172] S1, obtain brain perfusion images, i.e., CTP data;

[0173] S2, preprocessing the brain perfusion image to obtain a preprocessed brain perfusion image, that is, obtaining preprocessed CTP data;

[0174] S3, using a preset segmentation network to segment the blood supply area in the brain perfusion image to determine a plurality of divided blood supply areas; the segmentation network is trained based on a plurality of training perfusion images and a reference blood supply area corresponding to each training perfusion image;

[0175] S4, divides the brain into the first and second hemispheres;

[0176] S5, calculating a first peak time of the first side of the brain and calculating a second peak time of the second side of the brain;

[0177] S6, determining a target side of the brain from the first side of the brain and the second side of the brain based on the first peak time and the second peak time; the target side of the brain is the side of the brain opposite to the side where the lesion occurs;

[0178] S7, extracting and processing blood vessels in the target side of the brain in the brain perfusion image to determine a blood vessel image; the blood vessel image includes the blood vessels to be classified;

[0179] S8, extracting features from each voxel on the blood vessel to be classified to obtain extracted features corresponding to each voxel; the extracted features include spatial position information and curve feature values ​​of the voxels;

[0180] S9, performing vascular segment clustering processing on each voxel according to the extracted features of each voxel, and determining multiple clustered arterial vascular segments;

[0181] S10, determining first characteristic information corresponding to each blood supply area according to the divided multiple blood supply areas, and determining second characteristic information corresponding to each arterial vessel segment;

[0182] S11, using a preset cost function to calculate a cost function value between each first feature information and each second feature information, to obtain multiple cost function values ​​corresponding to each blood supply area;

[0183] S12, for each blood supply area, determining a minimum cost function value from multiple cost function values ​​corresponding to the blood supply area;

[0184] S13, determining the arterial segment corresponding to the minimum function value as the target arterial segment corresponding to the blood supply area; wherein the target arterial segment corresponding to each blood supply area is different;

[0185] S14, constructing a brain perfusion model based on the matching relationship between each blood supply area and each arterial segment to analyze the brain perfusion image.

[0186] It should be noted that Figure 6 、 8 , 9, and 10 are just examples, and the text therein does not affect the essence of the technical solution.

[0187] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0188] Based on the same inventive concept, embodiments of the present application further provide a perfusion image analysis device for implementing the aforementioned perfusion image analysis method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more of the following perfusion image analysis device embodiments can be found in the above-described limitations of the perfusion image analysis method and are not further elaborated here.

[0189] In one embodiment, Figure 11 As shown, a perfusion image analysis device is provided, comprising: an acquisition module, a segmentation module, a matching module and an analysis module, wherein:

[0190] an acquisition module, configured to acquire a brain perfusion image and determine a plurality of arterial vessel segments of the brain in the brain perfusion image;

[0191] a segmentation module for segmenting the blood supply areas in the brain perfusion image using a preset segmentation network to determine a plurality of divided blood supply areas; the segmentation network is trained based on a plurality of training perfusion images and a reference blood supply area corresponding to each training perfusion image;

[0192] A matching module is used to match each blood supply area with each arterial segment and determine the matching relationship between each blood supply area and each arterial segment;

[0193] The analysis module is used to construct a brain perfusion model based on the matching relationship to analyze the brain perfusion image.

[0194] In another embodiment, another perfusion image analysis device is provided. Based on the above embodiment, the matching module may include:

[0195] A first characteristic information determining unit is configured to determine first characteristic information corresponding to each blood supply area according to the divided multiple blood supply areas;

[0196] a second characteristic information determining unit, configured to determine second characteristic information corresponding to each arterial blood vessel segment;

[0197] The matching unit is used to match each blood supply area with each arterial blood vessel segment according to each first characteristic information and each second characteristic information, and determine the matching relationship between each blood supply area and each arterial blood vessel segment.

[0198] Optionally, the matching unit may include:

[0199] a calculation subunit, configured to calculate a cost function value between each first feature information and each second feature information using a preset cost function, and obtain a plurality of cost function values ​​corresponding to each blood supply area;

[0200] a target vessel determination subunit, configured to determine a target arterial vessel segment corresponding to each blood supply area according to a plurality of cost function values ​​corresponding to each blood supply area;

[0201] The matching relationship establishment subunit is used to establish a matching relationship between each blood supply area and the corresponding target arterial blood vessel segment.

[0202] Optionally, the above-mentioned target blood vessel determination subunit is specifically used to determine, for each blood supply area, a minimum cost function value from multiple cost function values ​​corresponding to the blood supply area; and determine the arterial blood vessel segment corresponding to the minimum function value as the target arterial blood vessel segment corresponding to the blood supply area; wherein the target arterial blood vessel segments corresponding to each blood supply area are different.

[0203] Optionally, the above-mentioned first characteristic information determination unit is specifically used to calculate the average time density curve corresponding to each blood supply area, and determine at least one curve characteristic value corresponding to each blood supply area based on each average time density curve; determine the relative center of mass position corresponding to each blood supply area based on the center of mass position of each blood supply area and the center of mass position of the brain; and determine at least one curve characteristic value and the relative center of mass position of each blood supply area as the first characteristic information of each blood supply area.

[0204] In another embodiment, another perfusion image analysis device is provided. Based on the above embodiment, the upper acquisition module may include:

[0205] a dividing unit for dividing the brain into a first side brain and a second side brain;

[0206] a target side determination unit, configured to determine a target side of the brain from the first side of the brain and the second side of the brain; the target side of the brain being the side opposite to the side where the brain lesion occurs;

[0207] The blood vessel determination unit is used to determine a plurality of arterial blood vessel segments in the target side brain in the brain perfusion image.

[0208] Optionally, the target side determination unit is specifically configured to calculate a first peak time of the first side of the brain and a second peak time of the second side of the brain; and determine the target side of the brain from the first side of the brain and the second side of the brain based on the first peak time and the second peak time.

[0209] Optionally, the blood vessel determination unit may include:

[0210] A blood vessel extraction subunit is used to extract and process blood vessels in the target side of the brain in the brain perfusion image to determine a blood vessel image; the blood vessel image includes blood vessels to be classified;

[0211] The clustering subunit is used to perform vascular segment clustering processing on each voxel on the blood vessel to be classified, and determine multiple clustered arterial blood vessel segments.

[0212] Optionally, the clustering subunit is specifically used to extract features from each voxel on the blood vessel to be classified to obtain the extracted features corresponding to each voxel; the extracted features include the spatial position information of the voxel; based on the extracted features of each voxel, each voxel is clustered into blood vessel segments to determine multiple arterial blood vessel segments after clustering.

[0213] Optionally, the clustering subunit is specifically used to obtain a time density curve of each voxel on the blood vessel to be classified; calculate at least one curve feature value of each voxel based on the time density curve of each voxel; and determine at least one curve feature value and spatial position information of each voxel as the extracted features of each voxel.

[0214] Each module in the perfusion image analysis device described above may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0215] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0216] A brain perfusion image is obtained and multiple arterial segments of the brain in the brain perfusion image are determined; a preset segmentation network is used to segment the blood supply area in the brain perfusion image to determine the multiple divided blood supply areas; the above-mentioned segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image; each blood supply area is matched with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment; and a brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion image.

[0217] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0218] The first characteristic information corresponding to each blood supply area is determined based on the multiple divided blood supply areas; the second characteristic information corresponding to each arterial blood vessel segment is determined; and based on each first characteristic information and each second characteristic information, each blood supply area and each arterial blood vessel segment is matched to determine the matching relationship between each blood supply area and each arterial blood vessel segment.

[0219] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0220] A preset cost function is used to calculate the cost function value between each first feature information and each second feature information to obtain multiple cost function values ​​corresponding to each blood supply area; the target arterial segment corresponding to each blood supply area is determined based on the multiple cost function values ​​corresponding to each blood supply area; and a matching relationship is established between each blood supply area and the corresponding target arterial segment.

[0221] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0222] For each blood supply area, a minimum cost function value is determined from multiple cost function values ​​corresponding to the blood supply area; the arterial segment corresponding to the minimum function value is determined as the target arterial segment corresponding to the blood supply area; wherein the target arterial segment corresponding to each blood supply area is different.

[0223] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0224] Calculate the average time density curve corresponding to each blood supply area, and determine at least one curve characteristic value corresponding to each blood supply area based on each average time density curve; determine the relative center of mass position corresponding to each blood supply area based on the center of mass position of each blood supply area and the center of mass position of the brain; determine at least one curve characteristic value and the relative center of mass position of each blood supply area as the first characteristic information of each blood supply area.

[0225] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0226] The brain is divided into a first side of the brain and a second side of the brain; a target side of the brain is determined from the first side of the brain and the second side of the brain; the target side of the brain is the opposite side of the brain to the side where the lesion occurs; and multiple arterial blood vessel segments of the target side of the brain are determined in the brain perfusion image.

[0227] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0228] A first peak time of the first side of the brain and a second peak time of the second side of the brain are calculated; and a target side of the brain is determined from the first side of the brain and the second side of the brain according to the first peak time and the second peak time.

[0229] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0230] The blood vessels of the target side of the brain in the brain perfusion image are extracted and processed to determine a blood vessel image; the blood vessel image includes the blood vessels to be classified; and the blood vessel segments are clustered on each voxel on the blood vessels to be classified to determine multiple clustered arterial blood vessel segments.

[0231] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0232] Feature extraction is performed on each voxel on the blood vessel to be classified to obtain the corresponding extracted features of each voxel; the extracted features include the spatial position information of the voxel; based on the extracted features of each voxel, each voxel is clustered into blood vessel segments to determine multiple clustered arterial blood vessel segments.

[0233] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0234] Obtain a time density curve of each voxel on the blood vessel to be classified; calculate at least one curve characteristic value of each voxel based on the time density curve of each voxel; and determine the at least one curve characteristic value and spatial position information of each voxel as the extraction feature of each voxel.

[0235] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0236] A brain perfusion image is obtained and multiple arterial segments of the brain in the brain perfusion image are determined; a preset segmentation network is used to segment the blood supply area in the brain perfusion image to determine the multiple divided blood supply areas; the above-mentioned segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image; each blood supply area is matched with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment; and a brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion image.

[0237] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0238] The first characteristic information corresponding to each blood supply area is determined based on the multiple divided blood supply areas; the second characteristic information corresponding to each arterial blood vessel segment is determined; and based on each first characteristic information and each second characteristic information, each blood supply area and each arterial blood vessel segment is matched to determine the matching relationship between each blood supply area and each arterial blood vessel segment.

[0239] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0240] A preset cost function is used to calculate the cost function value between each first feature information and each second feature information to obtain multiple cost function values ​​corresponding to each blood supply area; the target arterial segment corresponding to each blood supply area is determined based on the multiple cost function values ​​corresponding to each blood supply area; and a matching relationship is established between each blood supply area and the corresponding target arterial segment.

[0241] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0242] For each blood supply area, a minimum cost function value is determined from multiple cost function values ​​corresponding to the blood supply area; the arterial segment corresponding to the minimum function value is determined as the target arterial segment corresponding to the blood supply area; wherein the target arterial segment corresponding to each blood supply area is different.

[0243] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0244] Calculate the average time density curve corresponding to each blood supply area, and determine at least one curve characteristic value corresponding to each blood supply area based on each average time density curve; determine the relative center of mass position corresponding to each blood supply area based on the center of mass position of each blood supply area and the center of mass position of the brain; determine at least one curve characteristic value and the relative center of mass position of each blood supply area as the first characteristic information of each blood supply area.

[0245] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0246] The brain is divided into a first side of the brain and a second side of the brain; a target side of the brain is determined from the first side of the brain and the second side of the brain; the target side of the brain is the opposite side of the brain to the side where the lesion occurs; and multiple arterial blood vessel segments of the target side of the brain are determined in the brain perfusion image.

[0247] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0248] A first peak time of the first side of the brain and a second peak time of the second side of the brain are calculated; and a target side of the brain is determined from the first side of the brain and the second side of the brain according to the first peak time and the second peak time.

[0249] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0250] The blood vessels of the target side of the brain in the brain perfusion image are extracted and processed to determine a blood vessel image; the blood vessel image includes the blood vessels to be classified; and the blood vessel segments are clustered on each voxel on the blood vessels to be classified to determine multiple clustered arterial blood vessel segments.

[0251] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0252] Feature extraction is performed on each voxel on the blood vessel to be classified to obtain the corresponding extracted features of each voxel; the extracted features include the spatial position information of the voxel; based on the extracted features of each voxel, each voxel is clustered into blood vessel segments to determine multiple clustered arterial blood vessel segments.

[0253] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0254] Obtain a time density curve of each voxel on the blood vessel to be classified; calculate at least one curve characteristic value of each voxel based on the time density curve of each voxel; and determine the at least one curve characteristic value and spatial position information of each voxel as the extraction feature of each voxel.

[0255] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0256] A brain perfusion image is obtained and multiple arterial segments of the brain in the brain perfusion image are determined; a preset segmentation network is used to segment the blood supply area in the brain perfusion image to determine the multiple divided blood supply areas; the above-mentioned segmentation network is trained based on multiple training perfusion images and reference blood supply areas corresponding to each training perfusion image; each blood supply area is matched with each arterial segment to determine the matching relationship between each blood supply area and each arterial segment; and a brain perfusion model is constructed based on the matching relationship to analyze the brain perfusion image.

[0257] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0258] The first characteristic information corresponding to each blood supply area is determined based on the multiple divided blood supply areas; the second characteristic information corresponding to each arterial blood vessel segment is determined; and based on each first characteristic information and each second characteristic information, each blood supply area and each arterial blood vessel segment is matched to determine the matching relationship between each blood supply area and each arterial blood vessel segment.

[0259] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0260] A preset cost function is used to calculate the cost function value between each first feature information and each second feature information to obtain multiple cost function values ​​corresponding to each blood supply area; the target arterial segment corresponding to each blood supply area is determined based on the multiple cost function values ​​corresponding to each blood supply area; and a matching relationship is established between each blood supply area and the corresponding target arterial segment.

[0261] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0262] For each blood supply area, a minimum cost function value is determined from multiple cost function values ​​corresponding to the blood supply area; the arterial segment corresponding to the minimum function value is determined as the target arterial segment corresponding to the blood supply area; wherein the target arterial segment corresponding to each blood supply area is different.

[0263] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0264] Calculate the average time density curve corresponding to each blood supply area, and determine at least one curve characteristic value corresponding to each blood supply area based on each average time density curve; determine the relative center of mass position corresponding to each blood supply area based on the center of mass position of each blood supply area and the center of mass position of the brain; determine at least one curve characteristic value and the relative center of mass position of each blood supply area as the first characteristic information of each blood supply area.

[0265] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0266] The brain is divided into a first side of the brain and a second side of the brain; a target side of the brain is determined from the first side of the brain and the second side of the brain; the target side of the brain is the opposite side of the brain to the side where the lesion occurs; and multiple arterial blood vessel segments of the target side of the brain are determined in the brain perfusion image.

[0267] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0268] A first peak time of the first side of the brain and a second peak time of the second side of the brain are calculated; and a target side of the brain is determined from the first side of the brain and the second side of the brain according to the first peak time and the second peak time.

[0269] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0270] The blood vessels of the target side of the brain in the brain perfusion image are extracted and processed to determine a blood vessel image; the blood vessel image includes the blood vessels to be classified; and the blood vessel segments are clustered on each voxel on the blood vessels to be classified to determine multiple clustered arterial blood vessel segments.

[0271] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0272] Feature extraction is performed on each voxel on the blood vessel to be classified to obtain the corresponding extracted features of each voxel; the extracted features include the spatial position information of the voxel; based on the extracted features of each voxel, each voxel is clustered into blood vessel segments to determine multiple clustered arterial blood vessel segments.

[0273] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0274] Obtain a time density curve of each voxel on the blood vessel to be classified; calculate at least one curve characteristic value of each voxel based on the time density curve of each voxel; and determine the at least one curve characteristic value and spatial position information of each voxel as the extraction feature of each voxel.

[0275] It should be noted that the data involved in this application (including but not limited to data used for analysis, storage, display, etc.) are all data fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0276] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0277] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0278] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A perfusion image analysis method, characterized in that: The method comprises: Acquiring a brain perfusion image and determining a plurality of arterial vessel segments of the brain in the brain perfusion image; Using a preset segmentation network to segment the blood supply area in the brain perfusion image to determine a plurality of divided blood supply areas; the segmentation network is trained based on a plurality of training perfusion images and a reference blood supply area corresponding to each training perfusion image; Matching each of the blood supply areas with each of the arterial blood vessel segments to determine a matching relationship between each of the blood supply areas and each of the arterial blood vessel segments; A brain perfusion model is constructed according to the matching relationship to analyze the brain perfusion image.

2. The method according to claim 1, characterized in that Matching each of the blood supply areas with each of the arterial blood vessel segments to determine a matching relationship between each of the blood supply areas and each of the arterial blood vessel segments includes: determining first characteristic information corresponding to each of the divided blood supply areas according to the plurality of divided blood supply areas; Determining second characteristic information corresponding to each of the arterial blood vessel segments; According to each of the first characteristic information and each of the second characteristic information, each of the blood supply areas and each of the arterial blood vessel segments are matched to determine a matching relationship between each of the blood supply areas and each of the arterial blood vessel segments.

3. The method according to claim 2, characterized in that Matching each of the blood supply areas and each of the arterial blood vessel segments according to each of the first characteristic information and each of the second characteristic information to determine a matching relationship between each of the blood supply areas and each of the arterial blood vessel segments includes: Using a preset cost function to calculate a cost function value between each first feature information and each second feature information, to obtain a plurality of cost function values ​​corresponding to each blood supply area; determining a target arterial segment corresponding to each blood supply area according to a plurality of cost function values ​​corresponding to each blood supply area; A matching relationship is established between each of the blood supply areas and the corresponding target arterial vessel segment.

4. The method according to any one of claims 2 to 3, characterized in that: The determining, based on the divided multiple blood supply areas, first characteristic information corresponding to each of the blood supply areas includes: Calculating an average time density curve corresponding to each of the blood supply areas, and determining at least one curve characteristic value corresponding to each of the blood supply areas based on each of the average time density curves; Determining the relative centroid position corresponding to each blood supply area according to the centroid position of each blood supply area and the centroid position of the brain; At least one curve characteristic value and a relative centroid position of each of the blood supply areas are determined as first characteristic information of each of the blood supply areas.

5. The method according to any one of claims 1 to 3, characterized in that The determining of a plurality of arterial vessel segments in the brain in the brain perfusion image comprises: dividing the brain into a first side of the brain and a second side of the brain; determining a target side of the brain from the first side of the brain and the second side of the brain; the target side of the brain is the side of the brain opposite to the side where the lesion occurs; A plurality of arterial vessel segments in the target-side brain in the brain perfusion image are determined.

6. The method according to claim 5, characterized in that The determining of a plurality of arterial vessel segments in the target-side brain in the brain perfusion image comprises: Extracting and processing blood vessels in the target side of the brain in the brain perfusion image to determine a blood vessel image; the blood vessel image includes blood vessels to be classified; Blood vessel segment clustering is performed on each voxel on the blood vessel to be classified to determine a plurality of clustered arterial blood vessel segments.

7. The method according to claim 6, characterized in that The performing blood vessel segment clustering processing on each voxel on the blood vessel to be classified to determine a plurality of clustered arterial blood vessel segments includes: Extracting features from each voxel on the blood vessel to be classified to obtain extracted features corresponding to each voxel; the extracted features include spatial position information of the voxel; According to the extracted features of each voxel, each voxel is clustered into blood vessel segments to determine a plurality of clustered arterial blood vessel segments.

8. A perfusion image analysis device, characterized in that: The device comprises: an acquisition module, configured to acquire a brain perfusion image and determine a plurality of arterial vessel segments of the brain in the brain perfusion image; a segmentation module, configured to segment the blood supply areas in the brain perfusion image using a preset segmentation network to determine a plurality of divided blood supply areas; the segmentation network is trained based on a plurality of training perfusion images and a reference blood supply area corresponding to each training perfusion image; a matching module, configured to match each of the blood supply areas with each of the arterial blood vessel segments, and determine a matching relationship between each of the blood supply areas and each of the arterial blood vessel segments; An analysis module is used to construct a brain perfusion model according to the matching relationship to analyze the brain perfusion image.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images