A method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images

Abstract

This application relates to a method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images, comprising: reading CT perfusion images and performing spatial layering and preprocessing on the CT perfusion images to obtain perfusion images containing brain parenchyma and cerebrospinal fluid after removing the skull; using the perfusion images containing brain parenchyma and cerebrospinal fluid after removing the skull to determine the spatial range of the CT perfusion images containing cerebrospinal fluid regions; extracting feature images from the perfusion images containing brain parenchyma and cerebrospinal fluid within the spatial range of the CT perfusion images containing cerebrospinal fluid regions, and performing a first segmentation and screening, and a second segmentation on the feature images to obtain cerebrospinal fluid segmentation regions; extracting the cerebrospinal fluid segmentation regions for further processing to obtain the final cerebrospinal fluid region segmentation result; and removing the cerebrospinal fluid regions from the perfusion images after removing the skull to obtain the brain parenchyma regions.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to a method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images. Background Technology

[0002] Acute or chronic occlusion of cerebral arteries can lead to infarction of parts of the brain tissue due to insufficient blood supply, resulting in ischemic stroke. Acute ischemic stroke (cerebral infarction) has extremely high rates of disability and mortality. Brain CT perfusion imaging (CTP) is an imaging technique for evaluating the blood perfusion status of the brain parenchyma. It can accurately reflect the cerebral blood flow perfusion status and is currently an important imaging method for examining acute ischemic stroke. By processing brain CT perfusion images, perfusion parameter maps can be obtained, and the infarct core and ischemic penumbra can be identified based on these maps.

[0003] Identifying the location and volume of the infarct core and ischemic penumbra is crucial for the treatment of acute stroke patients. However, in acute ischemic stroke patients, the infarct core and ischemic penumbra are often connected to the cerebrospinal fluid (CSF) region. The appearance of CSF in CT perfusion images is similar to that of the ischemic region, and the perfusion parameters of the CSF region are also similar to those of the ischemic region on perfusion parameter maps. Therefore, the CSF region is prone to adhesion with the infarct core or ischemic penumbra region in three-dimensional space, affecting the calculation of the infarct core and ischemic penumbra volume. Therefore, it is necessary to segment the CT perfusion images into brain parenchyma and CSF regions to confine the infarct core and ischemic penumbra within the brain parenchyma, eliminating the influence of the CSF region, and thus accurately identifying the actual volume of the infarct core and ischemic penumbra within the CT perfusion images. Summary of the Invention

[0004] Therefore, it is necessary to provide a method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images to address the above-mentioned technical problems.

[0005] This application presents a method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images, including:

[0006] Read CT perfusion images, and perform spatial layering and preprocessing on the CT perfusion images to obtain perfusion images containing brain parenchyma and cerebrospinal fluid after removing the skull.

[0007] Using the perfusion images containing brain parenchyma and cerebrospinal fluid after skull removal, the spatial extent of the CT perfusion images containing cerebrospinal fluid regions is determined, including:

[0008] Determine the spatial location of the optimal CT perfusion image containing the cerebrospinal fluid region;

[0009] Determine the direction of the skull top and skull base in the CT perfusion images;

[0010] Based on the spatial location of the optimal CT perfusion image containing the cerebrospinal fluid region, a first spatial distance and a second spatial distance are found in the direction of the top of the skull and the base of the skull, respectively. Perfusion images whose spatial location is within this distance range are identified as CT perfusion images containing the cerebrospinal fluid region.

[0011] Within the CT perfusion image spatial range containing the cerebrospinal fluid region, feature images are extracted from the perfusion images containing brain parenchyma and cerebrospinal fluid, and the feature images are then segmented and filtered in the first stage and segmented in the second stage to obtain the cerebrospinal fluid segmentation region.

[0012] The cerebrospinal fluid segmentation region was extracted and further processed to obtain the final cerebrospinal fluid region segmentation result; the cerebrospinal fluid region was removed from the perfusion image after skull removal to obtain the brain parenchyma region.

[0013] Optionally, the spatial layering of the CT perfusion images includes: layering the perfusion images according to their scanning spatial location; the preprocessing includes spatial registration, temporal correction, and filtering and noise reduction of each layer of the CT perfusion images after layering, and segmentation and removal of the skull to obtain perfusion images containing brain parenchyma and cerebrospinal fluid.

[0014] Optionally, the spatial location of the CT perfusion image containing the cerebrospinal fluid region is determined by the area of ​​the largest connected region of the perfusion image containing brain parenchyma and cerebrospinal fluid after removing the skull.

[0015] Optionally, the CT perfusion image direction is obtained by the ratio of the perfusion image area containing brain parenchyma and cerebrospinal fluid after skull removal in each layer to the area of ​​its processed convex hull image.

[0016] The processing of perfusion images containing brain parenchyma and cerebrospinal fluid includes erosion, extraction of the largest connected component, etc.

[0017] Optionally, the method for determining the direction of the cranial vault and cranial base based on the area ratio is as follows:

[0018] The average of the area ratios of all perfusion images located below the optimal CT perfusion image containing the cerebrospinal fluid region is obtained to obtain the first area ratio.

[0019] The average area ratio of all perfusion images located above the optimal CT perfusion image containing the cerebrospinal fluid region is obtained to obtain the second area ratio.

[0020] If the first area ratio is greater than the second area ratio, then the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is downward in the direction of the top of the skull, and the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is upward in the direction of the base of the skull.

[0021] If the first area ratio is less than the second area ratio, then the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is downward in the direction of the skull base, and the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is upward in the direction of the skull top.

[0022] Optionally, the feature image of the CT perfusion image is one of the following:

[0023] Maximum CT value image, minimum CT value image, average CT value image, and the difference image between the maximum and minimum CT value images.

[0024] Optionally, the method for performing a first segmentation and a second segmentation on the feature image to obtain the lateral ventricle segmentation result includes:

[0025] The feature image of each spatial location is segmented for the first time using an adaptive threshold of the perfusion image at each spatial location, resulting in the first segmentation result of the cerebrospinal fluid region at each spatial location.

[0026] Based on the positional relationship between the first segmentation result of each spatial location and the centroid of the perfusion image containing brain parenchyma and cerebrospinal fluid at that spatial location, the first segmentation result of each spatial location is filtered and segmented a second time to obtain the second segmentation result of the cerebrospinal fluid region at that spatial location.

[0027] Optionally, the adaptive threshold for the perfusion image at each spatial location can be obtained as follows:

[0028] For the feature image of the spatial location, obtain the average value Imean and standard deviation Istd of the gray values ​​of each pixel;

[0029] The feature image at this spatial location is normalized in the first interval to obtain the gradient-enhanced image in the first interval, where the first interval is [0, Imean-Istd].

[0030] The gray values ​​of each pixel in the gradient-enhanced image at this spatial location are statistically analyzed to obtain the gray-level histogram of the gradient-enhanced image. The gray value at the trough position between two peaks in the gray-level histogram is found and defined as the adaptive segmentation threshold of the gradient-enhanced image at this spatial location.

[0031] Optionally, the method for filtering the first segmentation results and performing the second segmentation for each spatial location is as follows:

[0032] Extract connected components from the first segmentation result at each spatial location to obtain all segmented connected components at the current spatial location;

[0033] Obtain the centroid location of the perfusion image containing brain parenchyma and cerebrospinal fluid at each spatial location;

[0034] Based on the centroid position at this spatial location, obtain the first and second ranges within the plane;

[0035] For all connected components at this spatial location, obtain the centroid and boundary locations of the connected components;

[0036] Filter all split connected components: if the centroid of a connected component is outside the first range in the plane, delete the connected component.

[0037] For all connected components whose centroid positions satisfy the conditions, determine their boundary positions. If the boundary position exceeds the second range in the plane, then re-acquire the adaptive threshold for the connected component and perform a second segmentation.

[0038] Repeat the above steps until the expected result is achieved, and obtain the second segmentation result of the cerebrospinal fluid region at that spatial location.

[0039] Optionally, the method for further processing the extracted cerebrospinal fluid segmentation region to obtain the final cerebrospinal fluid region segmentation result is as follows:

[0040] The second segmentation results of the cerebrospinal fluid region at all spatial locations are combined into a three-dimensional model;

[0041] Take the N largest connected components from the 3D model and delete the rest.

[0042] The maximum N connected components of the 3D model are remapped to each spatial location to obtain the cerebrospinal fluid region segmentation result at that spatial location.

[0043] The cerebrospinal fluid region segmentation results at each spatial location are mapped onto the perfusion image containing brain parenchyma and cerebrospinal fluid at that spatial location; the cerebrospinal fluid region is removed from the perfusion image after removing the skull to obtain the brain parenchyma region.

[0044] The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images in this application has at least the following effects:

[0045] After reading brain CT perfusion images, this application can automatically complete the preprocessing of brain CT perfusion images, including spatial registration, temporal correction, filtering and noise reduction, and skull removal.

[0046] The cranial top and cranial base directions are obtained through automatic processing, which improves the efficiency of cerebrospinal fluid region and brain parenchyma region segmentation.

[0047] Based on the perfusion images containing brain parenchyma and cerebrospinal fluid after the skull is removed, this application can automatically determine the spatial range of CT perfusion images containing cerebrospinal fluid regions, so as to effectively eliminate the influence of the sulci and gyri of the cranial vertex region on the segmentation results and improve the accuracy of segmentation of cerebrospinal fluid regions and brain parenchyma regions.

[0048] The first segmentation of the cerebrospinal fluid region in this application can automatically adjust the adaptive threshold according to different images, which improves the applicability of the method. The first segmentation results of the cerebrospinal fluid region in this application are screened to effectively eliminate the influence of low perfusion areas and noise that are not connected to the cerebrospinal fluid region on the segmentation results. The second segmentation of the cerebrospinal fluid region in this application can effectively eliminate the influence of low perfusion areas and noise connected to the cerebrospinal fluid region on the segmentation results. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating a fully automated method for segmenting brain parenchyma and cerebrospinal fluid regions from brain CT perfusion images in one embodiment of this application.

[0050] Figure 2 This is a flowchart of a method for screening the first segmentation result and performing a second segmentation of the cerebrospinal fluid region in one embodiment of this application;

[0051] Figure 3a This is a feature image from one embodiment of this application;

[0052] Figure 3b This is the first segmentation result of the cerebrospinal fluid region at a certain spatial location in one embodiment of this application;

[0053] Figure 3c This is the second segmentation result of the cerebrospinal fluid region at a certain spatial location in one embodiment of this application;

[0054] Figure 3d This is the final result of cerebrospinal fluid region segmentation at a certain spatial location in one embodiment of this application. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] See Figure 1 A method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images, including steps S100 to S500.

[0057] Step S100: Read the CT perfusion image and perform spatial layering and preprocessing on the CT perfusion image to obtain a perfusion image containing brain parenchyma and cerebrospinal fluid after removing the skull.

[0058] CT perfusion images include CT perfusion images at different times and spatial locations.

[0059] Brain CT perfusion image sequences were read and layered according to the spatial location of the perfusion images. Each layer (i.e., each spatial location) contained perfusion images at all different times.

[0060] For each spatial location, the CT perfusion image is time-corrected based on the scan time information contained in the image.

[0061] Spatial registration is performed on the CT perfusion images at each spatial location. Images at different scan times at each spatial location are registered to the first time image at this spatial location, so that the in-plane spatial positions of each voxel point at this spatial location coincide in all time images.

[0062] All registered CT perfusion images are filtered to reduce noise, improve the signal-to-noise ratio of the perfusion images, and enhance the segmentation accuracy of this method.

[0063] The CT perfusion images at each spatial location are segmented to remove the skull. The skull with high CT values ​​is removed according to the threshold method or other segmentation methods, resulting in perfusion images containing brain parenchyma and cerebrospinal fluid after skull removal.

[0064] Step S200 utilizes the perfusion image containing brain parenchyma and cerebrospinal fluid after skull removal to determine the spatial extent of the CT perfusion image containing the cerebrospinal fluid region. This includes steps S210–S230, which involve determining the optimal spatial location of the CT perfusion image containing the cerebrospinal fluid region, determining the cranial top and base directions of the CT perfusion image, and determining the spatial extent of the CT perfusion image containing the cerebrospinal fluid region. This step excludes CT perfusion images located near the cranial top, ensuring that sulci with lower CT values ​​in these images do not affect the segmentation of the cerebrospinal fluid region.

[0065] Step S210: Using the perfusion images containing brain parenchyma and cerebrospinal fluid after skull removal, determine the spatial location of the optimal CT perfusion image containing the cerebrospinal fluid region. Specific steps include:

[0066] (1) Binarize the perfusion image Ii containing brain parenchyma and cerebrospinal fluid at the i-th spatial location after removing the skull to obtain the binarized image Bi.

[0067] (2) Extract the connected components from the binarized image Bi to obtain the largest connected component Ci;

[0068] (3) Calculate the area Ai of the connected component Ci;

[0069] (4) Perform the above processing on the perfusion images containing brain parenchyma and cerebrospinal fluid at all spatial positions to obtain the area of the connected component at each spatial position;

[0070] (5) Compare the areas of the connected components at all spatial positions. If the area of the connected component at the j-th spatial position is the largest, then the j-th spatial position is the spatial position where the optimal CT perfusion image containing the cerebrospinal fluid region is located.

[0071] Step S220, determine the cranial vertex direction and cranial base direction of the CT perfusion image, and the specific steps include:

[0072] (1) Binarize the perfusion image Ii containing brain parenchyma and cerebrospinal fluid after removing the skull at the i-th spatial position to obtain a binary image Bi;

[0073] (2) Use an erosion kernel of 5*5 voxel points to perform erosion processing on the Bi image to obtain an eroded image Ei; take the largest connected component of the image Ei to obtain the largest connected component image Ci;

[0074] (3) For the largest connected component image Ci at the i-th spatial position, obtain its convex hull image Hi;

[0075] (4) At the i-th spatial position, calculate the area Ai of the largest connected component image Ci and the area A’i of the convex hull image Hi, and find the ratio ri = Ai / A’i of the two areas;

[0076] (5) Perform the above processing on the perfusion images containing brain parenchyma and cerebrospinal fluid at all spatial positions to obtain the area ratio of the largest connected component image to its corresponding convex hull image for each layer;

[0077] (6) Obtain the mean value of the area ratios of all perfusion images whose spatial coordinates are below the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region to obtain the first area ratio R1; obtain the mean value of the area ratios of all perfusion images whose spatial coordinates are above the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region to obtain the second area ratio R2;

[0078] (7) Judge the cranial vertex and cranial base directions: If R1 > R2, then the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is downward as the cranial vertex direction, and the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is upward as the cranial base direction; if R1 < R2, then the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is downward as the cranial base direction, and the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region is upward as the cranial vertex direction.

[0079] Step S230: A first distance L1 is found from the spatial location of the optimal CT perfusion image containing the cerebrospinal fluid (CSF) region towards the top of the skull; a first distance L2 is found from the spatial location of the optimal CT perfusion image containing the CSF region towards the base of the skull. Perfusion images within this distance range are considered CT perfusion images containing the CSF region. The values ​​of the first and second distances can be determined using human physiological parameters. In reality, the distribution of images containing the CSF region is uneven on both sides of the optimal image location; that is, the ideal first and second distances are not the same. The first distance L1 is typically 40–50 mm, and the second distance L2 is typically 20–40 mm. This step automatically obtains the top and base directions of the skull, and when obtaining images containing the CSF region, excludes non-target images that do not contain the CSF region as much as possible, ensuring the reliability of the samples and significantly improving the accuracy of subsequent processes.

[0080] Step S300 includes: Step S310, extracting feature images from the perfusion images containing brain parenchyma and cerebrospinal fluid within the CT perfusion image spatial range containing the cerebrospinal fluid region; Step S320, performing a first segmentation on the feature images; Step S330, filtering the first segmentation results and performing a second segmentation to obtain the cerebrospinal fluid segmentation region.

[0081] In step S310, the feature image is one of the following: the maximum CT value image Imax, the minimum CT value image Imin, the average CT value image Imean, or the difference image between the maximum and minimum CT value images I. diff =(I max -I min ).

[0082] Step S320, feature images (e.g., within the spatial range of the CT perfusion image containing the cerebrospinal fluid region) at each spatial location. Figure 3a The first segmentation of the cerebrospinal fluid region is performed on the maximum CT value image (Imax) shown. The specific steps include:

[0083] (1) For the feature image at each spatial location, obtain the average gray value I of each pixel. mean and standard deviation I std ;

[0084] (2) Normalize the feature image at the spatial location in the first interval to obtain the gradient-enhanced image in the first interval, where the first interval is [0, I]. mean -I std Specifically, the CT values ​​in the feature images of each layer are between 0 and (I... mean -I stdPoints within the first interval are normalized to 0-255 HU. For CT values ​​higher than (I... mean -I std Points with CT values ​​below 0 HU are assigned a value of 255 HU, while points with CT values ​​below 0 HU are assigned a value of 0 HU, thus obtaining gradient-enhanced images. This operation can expand the CT value range of the cerebrospinal fluid region, enhancing the contrast between the cerebrospinal fluid region and the surrounding brain parenchyma.

[0085] (3) Statistically calculate the gray values ​​of each pixel in the gradient-enhanced image at this spatial location, obtain the gray-level histogram of the gradient-enhanced image, find the gray value at the valley position between two peaks in the gray-level histogram, and define the gray value as the adaptive segmentation threshold of the gradient-enhanced image at this spatial location.

[0086] (4) For this spatial location, the gradient enhancement image at this spatial location is segmented using the corresponding adaptive segmentation threshold. Voxel points with pixel values ​​greater than 0 and lower than the threshold are defined as foreground points and assigned a value of 255; other points are defined as background points and assigned a value of 0; thus, the first segmentation binary image of the cerebrospinal fluid region of the feature image at this spatial location is obtained.

[0087] (5) Perform steps (1) to (4) on the feature images of each spatial location within the CT perfusion image spatial range containing the cerebrospinal fluid region to obtain the first segmentation binary image of the cerebrospinal fluid region at each spatial location, such as... Figure 3b As shown.

[0088] S330, the first segmentation results for each spatial location are filtered and a second segmentation is performed to obtain the second segmentation results for the cerebrospinal fluid region, such as... Figure 2 As shown, the specific steps include:

[0089] (1) Extract the connected components from the first segmentation result at each spatial location to obtain all segmented connected components at the current spatial location;

[0090] (2) Obtain the centroid position M0(x0,y0) of the perfusion image containing brain parenchyma and cerebrospinal fluid at each spatial location, which is obtained by the following formula.

[0091]

[0092]

[0093] In the formula, i and j are the coordinates of all points in the perfusion image at this spatial location in the x and y directions; m and n are the number of points in the perfusion image at this spatial location in the x and y directions; I ij The gray value is the point with coordinates (i,j) in the first segmentation result image of the cerebrospinal fluid region at this spatial location.

[0094] (3) Based on the centroid position at the spatial location, obtain the first range and the second range in the plane;

[0095] The first range is defined by the following formula:

[0096] x0-α·(x0-x L )<x<x0+α·(x R -x0)

[0097] y0-α·(y0-y U )<y<y0+α·(y B -y0)

[0098] Where x and y are the centroid positions of the two-dimensional connected domain, and α is a given coefficient, for example, α takes the value between 0.4 and 0.6. L To remove the left boundary of the brain parenchyma and cerebrospinal fluid region behind the skull, x R To remove the right boundary of the brain parenchyma and cerebrospinal fluid region behind the skull, y U To remove the upper boundary of the brain parenchyma and cerebrospinal fluid region behind the skull, y B This refers to the lower boundary of the brain parenchyma and cerebrospinal fluid region after the skull has been removed.

[0099] The range of the second expected location is limited by the following formula:

[0100] x0-β·(x0-x L )<x<x0+β·(x R -x0)

[0101] y0-β·(y0-y U )<y<y0+β·(y B -y0)

[0102] In the formula, x and y are the boundary positions of the two-dimensional connected domain, and β is a given coefficient, for example, 0.6 to 0.8.

[0103] (4) Obtain the centroid position M of all connected components at this spatial location. i and boundary location;

[0104] (5) Filter all partitioned connected components. If the centroid of a certain connected component is M... i If the region exceeds the first range within the plane, the connected region is deleted. Since the centroid of the noise region or low perfusion region that is not connected to the cerebrospinal fluid region is not within the first range mentioned above, this step can exclude the noise region or low perfusion region that is not connected to the cerebrospinal fluid region.

[0105] (6) For all centroid positions M iFor connected components that meet the conditions, determine their boundary positions. If the boundary position exceeds the second range in the plane, then re-acquire the adaptive threshold for the connected component and perform a second segmentation. It can be understood that if it exceeds the second range, there are noise regions or low perfusion regions connected to the cerebrospinal fluid region. This step can separate them from the cerebrospinal fluid region.

[0106] (7) Repeat the above steps until the expected result is achieved, obtaining the second segmentation result of the cerebrospinal fluid region at that spatial location, such as... Figure 3c As shown.

[0107] If the boundary of the two-dimensional connected domain exceeds the second range, the two-dimensional connected domain is segmented again using an adaptive threshold method, and steps (4) to (6) are executed again after segmentation. This continues until both the centroid of the two-dimensional connected domain is in the first position and the boundary of the two-dimensional connected domain is in the second position (until the expected conditions are met). Connected domains that meet the above conditions are merged to obtain the second segmentation result of the cerebrospinal fluid region at that spatial location.

[0108] S400, extracts the segmented cerebrospinal fluid region for further processing to obtain the final cerebrospinal fluid region segmentation result, specifically including:

[0109] (1) Combine the second segmentation results of the cerebrospinal fluid region at all spatial locations into a three-dimensional model;

[0110] (2) Take the largest connected component of the three-dimensional model and delete the rest;

[0111] (3) Remap the largest connected component of the 3D model to each spatial location to obtain the cerebrospinal fluid region segmentation result at that spatial location, such as... Figure 3d As shown.

[0112] In this step, the second segmentation results of the cerebrospinal fluid (CSF) region at all spatial locations are mapped onto a 3D model, and the maximum connected component is extracted from this model to obtain a 3D model of the entire CSF region. This 3D model is then mapped to each spatial location to obtain the final CSF region segmentation result at each location. This step determines the final CSF region based on the spatial connectivity of the segmentation results, using the maximum connected component as the final CSF region. This step results in better spatial continuity of the segmented CSF region, better reflecting its anatomical characteristics.

[0113] Step S500: Remove the cerebrospinal fluid region from the perfusion image after skull removal to obtain the brain parenchyma region.

[0114] The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images proposed in the embodiments of this application can read brain CT perfusion images, perform preprocessing, and automatically complete the segmentation of cerebrospinal fluid regions and brain parenchyma from brain CT perfusion images. This method can automatically determine the spatial extent of CT perfusion images containing cerebrospinal fluid regions, so as to accurately exclude the influence of cerebrospinal fluid regions in subsequent identification of infarct cores and hypoperfusion areas, helping doctors to better diagnose patients with acute stroke and providing more information for clinical treatment.

[0115] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0116] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered to be within the scope of this specification. When technical features of different embodiments are embodied in the same drawing, it can be regarded as the drawing also disclosing examples of combinations of the various embodiments involved.

[0117] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images, characterized in that, include: Read CT perfusion images, and perform spatial layering and preprocessing on the CT perfusion images to obtain perfusion images containing brain parenchyma and cerebrospinal fluid after removing the skull. Using the perfusion images containing brain parenchyma and cerebrospinal fluid after skull removal, the spatial extent of the CT perfusion images containing cerebrospinal fluid regions is determined, including: Determine the spatial location of the optimal CT perfusion image containing the cerebrospinal fluid region; The direction of the skull top and skull base in CT perfusion images is determined by the ratio of the area of ​​the perfusion image containing brain parenchyma and cerebrospinal fluid after skull removal to the area of ​​its processed convex hull image. Specifically, this includes: obtaining the average of the area ratios of all perfusion images located below the optimal CT perfusion image containing the cerebrospinal fluid region to obtain a first area ratio; obtaining the average of the area ratios of all perfusion images located above the optimal CT perfusion image containing the cerebrospinal fluid region to obtain a second area ratio; if the first area ratio is greater than the second area ratio, then the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region downwards is the skull top direction, and the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region upwards is the skull base direction; if the first area ratio is less than the second area ratio, then the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region downwards is the skull base direction, and the spatial position of the optimal CT perfusion image containing the cerebrospinal fluid region upwards is the skull top direction. Based on the spatial location of the optimal CT perfusion image containing the cerebrospinal fluid region, a first spatial distance and a second spatial distance are found in the direction of the top of the skull and the base of the skull, respectively. Perfusion images whose spatial location is within this distance range are identified as CT perfusion images containing the cerebrospinal fluid region. Within the CT perfusion image spatial range containing the cerebrospinal fluid region, feature images are extracted from the perfusion images containing brain parenchyma and cerebrospinal fluid, and the feature images are then segmented and filtered in the first stage and segmented in the second stage to obtain the cerebrospinal fluid segmentation region. The cerebrospinal fluid (CSF) segmentation region was extracted and further processed to obtain the final CSF region segmentation result; the CSF region was removed from the perfusion image after skull removal to obtain the brain parenchyma region.

2. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 1, characterized in that, The spatial layering of the CT perfusion images includes: layering the perfusion images according to their scanning spatial location; the preprocessing includes spatial registration, temporal correction and filtering noise reduction of each layer of CT perfusion images after layering, and segmentation and removal of the skull to obtain perfusion images containing brain parenchyma and cerebrospinal fluid.

3. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 1, characterized in that, The spatial location of the optimal CT perfusion image containing the cerebrospinal fluid region is determined by the area of ​​the largest connected region of the perfusion image containing brain parenchyma and cerebrospinal fluid after removing the skull.

4. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 1, characterized in that, Processing of perfusion images containing brain parenchyma and cerebrospinal fluid includes: erosion and extraction of the largest connected components.

5. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 1, characterized in that, The characteristic image of the CT perfusion image is one of the following: Maximum CT value image, minimum CT value image, average CT value image, and the difference image between the maximum and minimum CT value images.

6. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 1, characterized in that, The feature image undergoes a first segmentation and filtering, followed by a second segmentation, including: The feature image of each spatial location is segmented for the first time using an adaptive threshold of the perfusion image at each spatial location, resulting in the first segmentation result of the cerebrospinal fluid region at each spatial location. Based on the positional relationship between the first segmentation result of each spatial location and the centroid of the perfusion image containing brain parenchyma and cerebrospinal fluid at that spatial location, the first segmentation result of each spatial location is filtered and segmented a second time to obtain the second segmentation result of the cerebrospinal fluid region at that spatial location.

7. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 6, characterized in that, The method for obtaining the adaptive threshold of the perfusion image at each spatial location is as follows: For the feature image of the spatial location, obtain the average gray value I of each pixel. mean and standard deviation I std ; The feature image at this spatial location is normalized in the first interval to obtain a gradient-enhanced image in the first interval, where the first interval is [0, I]. mean -I std ]; The gray values ​​of each pixel in the gradient-enhanced image at this spatial location are statistically analyzed to obtain the gray-level histogram of the gradient-enhanced image. The gray value at the trough position between two peaks in the gray-level histogram is found and defined as the adaptive segmentation threshold of the gradient-enhanced image at this spatial location.

8. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 6, characterized in that, The method for filtering the first segmentation results and performing a second segmentation for each spatial location is as follows: Extract connected components from the first segmentation result at each spatial location to obtain all segmented connected components at the current spatial location; Obtain the centroid location of the perfusion image containing brain parenchyma and cerebrospinal fluid at each spatial location; Based on the centroid position at this spatial location, obtain the first and second ranges within the plane; For all connected components at this spatial location, obtain the centroid and boundary locations of the connected components; Filter all split connected components: if the centroid of a connected component is outside the first range in the plane, delete the connected component. For all connected components whose centroid positions satisfy the conditions, determine their boundary positions. If the boundary position exceeds the second range in the plane, then re-acquire the adaptive threshold for the connected component and perform a second segmentation. Repeat the above steps until the expected result is achieved, and obtain the second segmentation result of the cerebrospinal fluid region at that spatial location.

9. The method for segmenting brain parenchyma and cerebrospinal fluid regions based on brain CT perfusion images according to claim 1, characterized in that, The method for further processing the extracted cerebrospinal fluid segmentation region to obtain the final cerebrospinal fluid region segmentation result is as follows: The second segmentation results of the cerebrospinal fluid region at all spatial locations are combined into a three-dimensional model; Take the N largest connected components from the 3D model and delete the rest. The maximum N connected components of the 3D model are remapped to each spatial location to obtain the cerebrospinal fluid region segmentation result at that spatial location. The cerebrospinal fluid region segmentation results at each spatial location are mapped onto the perfusion image containing brain parenchyma and cerebrospinal fluid at that spatial location; the cerebrospinal fluid region is removed from the perfusion image after removing the skull to obtain the brain parenchyma region.

Images