Method and readable storage medium for acquiring perfusion parameter map and lesion region

Abstract

The application relates to a method for obtaining a perfusion parameter map and a lesion area and a computer readable storage medium, and the method comprises the following steps: obtaining three-dimensional dynamic CT perfusion images based on brain CT perfusion images at different time points; performing filtering processing on each voxel in the three-dimensional dynamic CT perfusion images to obtain filtered three-dimensional dynamic CT perfusion images, wherein the filtering processing comprises the following steps: at different time points, for each current voxel, filtering is performed by using spatial similarity and time intensity similarity of adjacent voxels of the current voxel; and obtaining a perfusion parameter map and a lesion area according to the filtered three-dimensional dynamic CT perfusion images. The application filters by using spatial similarity and time intensity similarity, considers not only spatial similarity information but also time information of continuous scanning CT perfusion images, improves the signal-to-noise ratio of the images and the calculation accuracy, and the obtained perfusion parameter map and lesion area in the subsequent process are more reliable.

Description

Technical Field

[0001] This application relates to the field of medical image processing, and in particular to a method and a readable storage medium for acquiring perfusion parameter maps and lesion areas. Background Technology

[0002] With the development of medical imaging and computer technology, brain CT perfusion imaging (CTP) has become an important imaging method for examining acute ischemic stroke. By quantitatively analyzing CT perfusion images, relevant cerebral hemodynamic perfusion parameters of the patient can be obtained, such as cerebral blood volume (CBV), cerebral blood flow (CBF), and mean transient time (MTT). Based on the perfusion parameter map, normal brain tissue, diseased tissue (infarct core), and reperfused ischemic tissue (ischemic penumbra) can be identified.

[0003] Rapid and accurate processing of CT perfusion images to obtain perfusion parameter maps and identify the location and extent of the infarct core and ischemic penumbra is crucial for the treatment of stroke patients. However, CT perfusion images are affected by noise, resulting in a low signal-to-noise ratio. The main sources of noise in CT perfusion images include quantum noise, noise introduced by inherent limitations of the CT hardware system, and noise introduced during image reconstruction. Noise in CT perfusion images is unavoidable, and this noise severely affects the accuracy of parameter map calculation, thus impacting the identification of the infarct core and ischemic penumbra. Therefore, filtering and denoising CT perfusion images before calculating parameter maps is essential. Currently, most post-processing methods for CT perfusion images use Gaussian filtering or bilateral filtering for noise reduction. However, these filtering methods only consider spatial and value domain information, ignoring the temporal information of continuously scanned CT perfusion images, thus limiting their noise reduction effectiveness. Summary of the Invention

[0004] Therefore, it is necessary to provide a method for obtaining perfusion parameter maps and lesion areas to address the aforementioned technical problems.

[0005] The method for obtaining perfusion parameter maps and lesion areas in this application includes:

[0006] Obtain three-dimensional dynamic CT perfusion images based on brain CT perfusion images at different times;

[0007] Each voxel in the three-dimensional dynamic CT perfusion image is filtered to obtain a filtered three-dimensional dynamic CT perfusion image. The filtering process includes: at different times, for each current voxel, filtering is performed using the spatial similarity and temporal intensity similarity of the voxels adjacent to the current voxel.

[0008] Based on the filtered three-dimensional dynamic CT perfusion images, perfusion parameter maps and lesion areas are obtained.

[0009] Optionally, the temporal intensity similarity is obtained by the following formula:

[0010] In the formula, p is the current voxel, n is the neighboring voxels of the current voxel, σ1 is the standard deviation of time-intensity similarity, exp is the exponential function, and MSD(p,n) is the root mean square error of the time-intensity curves of voxel p and neighboring voxels n.

[0011] Optionally, the mean square error is obtained by the following formula:

[0012] In the formula, T is the total number of different times in each layer of the three-dimensional dynamic CT perfusion image, I(p(x,y,z,t)) is the intensity value of voxel p at spatial location (x,y,z) at time t, and I(n(ξ,η,ζ,t)) is the intensity value of voxel n at spatial location (ξ,η,ζ) at time t.

[0013] Optionally, the spatial similarity is obtained by the following formula:

[0014] In the formula, d(p,n) is the Euclidean distance between voxel p and its neighboring voxel n, and σ2 is the standard deviation of spatial similarity.

[0015] Optionally, the filtering process is performed using the following formula:

[0016]

[0017] Where L(p) is the normalization coefficient, X, Y, and Z are the preset radii of the filter kernel, p is the current voxel, n is the neighboring voxels of the current voxel, Ts(p,n) is the temporal intensity similarity between voxel p and voxel n, Ds(p,n) is the spatial similarity between voxel p and voxel n, (ξ,η,ζ) are the three spatial coordinates of voxel n, (x,y,z) are the three spatial coordinates of voxel p, i, j, and k are variables used for accumulation, I(n(ξ+i,η+j,ζ+k,t)) is the intensity value of spatial position (ξ+i,η+j,ζ+k) at time t, and F(p(x,y,z,t)) is the intensity value of voxel p after filtering at time t.

[0018] Optionally, based on the filtered three-dimensional dynamic CT perfusion image, a perfusion parameter map is obtained, specifically including:

[0019] Based on the filtered three-dimensional dynamic CT perfusion images, tissue time density curves, arterial input functions, and venous output functions are obtained.

[0020] The arterial input function is modified using the venous output function to obtain the modified arterial input function;

[0021] The residual function is obtained based on the tissue time density curve and the modified arterial input function;

[0022] The perfusion parameter map is obtained based on the residual function, which includes the cerebral blood flow (CBF) parameter map and the residual function peak time (Tmax) parameter map.

[0023] Optionally, the three-dimensional dynamic CT perfusion image at each time moment is formed by sequentially performing image registration, skull removal, and lateral ventricle removal on the brain CT perfusion image at that time moment.

[0024] Optionally, obtaining the lesion area includes obtaining the infarct core area, specifically including:

[0025] Obtaining the three-dimensional connected domain of the low-perfusion zone: The Tmax parameter map of each layer is binarized, closed, and filled with holes in sequence using the first threshold, and the three-dimensional connected domain of the low-perfusion zone is obtained after combination.

[0026] To obtain a normal CBF reference benchmark: Divide the brain CT perfusion images of each slice into the left and right hemispheres. Based on the proportion of voxels whose residual function peak time is greater than the first threshold in the left and right hemispheres, classify the left and right hemispheres into abnormal and normal sides. Calculate the average cerebral blood flow of voxels in the normal side that are less than the second threshold. This average value is used as the normal CBF reference benchmark.

[0027] Obtaining the infarct core region: Obtain a relative CBF parameter map, which is a ratio map of cerebral blood flow of each voxel to the normal CBF reference benchmark. Use a third threshold to perform binarization, closure operation and hole filling on the relative CBF parameter map in sequence. After combination, take the intersection with the three-dimensional connected domain of the hypoperfusion area to obtain the three-dimensional intersection region. Perform closure operation and hole filling on the three-dimensional intersection region of each layer in sequence to obtain the infarct core region.

[0028] Optionally, obtaining the lesion area includes obtaining the penumbra area, specifically including:

[0029] The penumbra region is obtained by combining the three-dimensional connected domain of the low-perfusion zone and the infarct core region.

[0030] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for obtaining perfusion parameter maps and lesion areas as described in this application.

[0031] The method for obtaining perfusion parameter maps and lesion areas in this application has at least the following effects:

[0032] This application can perform fully automatic image post-processing of perfusion parameter maps. When filtering three-dimensional dynamic CT perfusion images, it uses the spatial similarity and temporal intensity similarity of adjacent voxels to filter them. It not only considers spatial similarity information, but also the temporal information of continuously scanned CT perfusion images, which improves the signal-to-noise ratio and calculation accuracy of the images. The perfusion parameter maps and lesion areas obtained in subsequent processes are more reliable. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating a method for obtaining perfusion parameter maps and lesion areas in one embodiment of this application.

[0034] Figure 2 This is a flowchart of a method for obtaining perfusion parameter maps and lesion areas in one embodiment of this application;

[0035] Figure 3 This is a flowchart illustrating the process of obtaining the lesion area in one embodiment of this application;

[0036] Figure 4 To utilize Figure 1 A schematic diagram of the infarct core obtained by the method shown;

[0037] Figure 5 To utilize Figure 1 A schematic diagram of the low-perfusion region obtained by the method shown.

[0038] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0039] The location and size of the infarct core and ischemic penumbra are important indicators affecting the treatment of acute stroke patients. Therefore, rapid and accurate post-processing of brain CT perfusion images is necessary to accurately identify the location and volume of the infarct core and penumbra. However, traditional filtering methods, which do not utilize the temporal changes in contrast agent during perfusion, have poor filtering effects on brain CT perfusion images, affecting the accuracy of parametric mapping and the identification of the infarct core and penumbra.

[0040] 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.

[0041] To solve the above technical problems, see [link to relevant documentation]. Figure 1 and Figure 2One embodiment of this application provides a method for obtaining perfusion parameter maps and lesion areas, including:

[0042] Step S100: Obtain three-dimensional dynamic CT perfusion images based on brain CT perfusion images at different times.

[0043] In step S100, the three-dimensional dynamic CT perfusion image at each time moment is formed by sequentially performing image registration, skull removal, and lateral ventricle removal on the brain CT perfusion image at that time moment. This process is completed sequentially. Figure 2 The image shown is received, brain CT perfusion images are automatically extracted, perfusion images are read, layered and sorted by time, and CT perfusion images are preprocessed.

[0044] For example, images (DICOM images) are received via connection to a PACS system. These images, in addition to brain CT perfusion images, may also include plain CT images, CT contrast images, and other images such as CT post-processing results. After receiving the images, the brain CT perfusion images are identified and extracted. Brain CT perfusion images are obtained by performing continuous CT scans of the patient's brain at different levels after intravenous injection of an iodinated contrast agent. They are characterized by multiple images at different times for each level, and the number of images per level is the same.

[0045] Therefore, the method for automatically identifying and extracting brain CT perfusion images is as follows: Traverse each image, obtain all image labels, obtain the image sequence number, and perform preliminary classification based on the sequence number, grouping images with the same sequence number into the same image sequence. For each image sequence, determine its number of image channels, and delete images with a channel count not equal to 1 (such as color images from post-processing CT equipment), retaining only grayscale image sequences. Traverse each image sequence, count the scanning position information of all images in the sequence, and obtain the number of images corresponding to each scanning position. If the number of images corresponding to all scanning positions in the sequence is the same, then the sequence is a brain CT perfusion sequence.

[0046] Brain CT perfusion image sequences were read, and the images were layered according to the scan location information contained in the images. Each layer contained perfusion images at all different times. For the images in each layer, the images were sorted by time according to the scan time information contained in the images.

[0047] Image preprocessing operations on the time-ordered images include: image registration, brain tissue segmentation, and lateral ventricle segmentation. Image registration involves registering images from different scan times at each slice to the image from the first scan time at that slice, ensuring that voxels on all scan times spatially coincide. After image registration, the axis of symmetry for each slice of the CT perfusion image can be obtained based on the symmetry of the skull image. Brain tissue segmentation involves removing high-CT-value skull fragments using thresholding or other segmentation methods to obtain brain tissue images. Lateral ventricle segmentation involves removing low-CT-value lateral ventricles containing cerebrospinal fluid using thresholding or other segmentation methods.

[0048] That is, by sequentially registering images, removing the skull, and removing the lateral ventricles, three-dimensional dynamic CT perfusion images are formed at each moment.

[0049] Step S200: Filter each voxel in the three-dimensional dynamic CT perfusion image to obtain the filtered three-dimensional dynamic CT perfusion image. The filtering process includes: at different times, for each current voxel, filtering is performed using the spatial similarity and temporal intensity similarity of the voxels adjacent to the current voxel.

[0050] Step S200 employs a 4D time-space coupled filtering method (4D-Filter) to perform filtering and noise reduction processing on the three-dimensional dynamic CT perfusion image, namely... Figure 2 The CT perfusion images were filtered and denoised.

[0051] Temporal intensity similarity is obtained by the following formula: In the formula, p is the current voxel, n is the neighboring voxels of the current voxel, σ1 is the standard deviation of time-intensity similarity, exp is the exponential function, and MSD(p,n) is the root mean square error of the time-intensity curves of voxel p and its neighboring voxel n. σ1 is used to measure at what level the root mean square error MSD is required for the time-intensity curves of two voxels p and n to still be considered similar.

[0052] The mean square error is obtained by the following formula: In the formula, T is the total number of different times in each layer of the three-dimensional dynamic CT perfusion image, that is, T is the size of the time dimension. Each layer of the CT perfusion image has T times of image. I(p(x,y,z,t)) is the intensity value (i.e. CT intensity value) of voxel p at spatial position (x,y,z) at time t. I(n(ξ,η,ζ,t)) is the intensity value (i.e. CT intensity value) of voxel n at spatial position (ξ,η,ζ) at time t.

[0053] Spatial similarity is obtained by the following formula: In the formula, d(p,n) is the Euclidean distance between voxel p and its adjacent voxel n, and σ² is the standard deviation of spatial similarity. σ² is used to measure at what spatial distance d two voxels p and n are still considered similar in terms of time-intensity curves.

[0054] The 4D time-space coupled filtering method combines the time-intensity similarity function with the spatial similarity function. The filtering process is performed using the following formula to obtain the filtered intensity value:

[0055]

[0056] Where L(p) is the normalization coefficient, X, Y, and Z are the preset radii of the filter kernel, p is the current voxel, n is the neighboring voxels of the current voxel, Ts(p,n) is the temporal intensity similarity between voxel p and voxel n, Ds(p,n) is the spatial similarity between voxel p and voxel n, (ξ,η,ζ) are the three spatial coordinates of voxel n, (x,y,z) are the three spatial coordinates of voxel p, i, j, and k are variables used for accumulation, I(n(ξ+i,η+j,ζ+k,t)) is the intensity value of spatial position (ξ+i,η+j,ζ+k) at time t, and F(p(x,y,z,t)) is the intensity value of voxel p after filtering at time t.

[0057] Specifically, During filtering, each spatial point of each scan time of the CT perfusion image is traversed. After filtering is completed, the filtered data is obtained, which is the filtered three-dimensional dynamic CT perfusion image, used to perform subsequent image processing steps.

[0058] The preset radius of the filter kernel refers to the maximum pixel distance between the current voxel and its adjacent voxels in three-dimensional space. It is controlled and adjusted using three spatial dimensions: X, Y, and Z. For example, X can be 2–5, Y can be 2–5, and Z can be 1–2, with each value representing the maximum pixel distance in its respective spatial dimension. Voxels within this maximum pixel distance can be used as adjacent voxels for filtering.

[0059] Step S300: Based on the filtered three-dimensional dynamic CT perfusion image, obtain the perfusion parameter map and the lesion area.

[0060] Step S300 includes steps S310 to S330. Specifically, step S310 involves obtaining a perfusion parameter map. Step S320 involves obtaining the infarct core region (the lesion area). Step S330 involves obtaining the penumbra region.

[0061] Step S310, obtaining the perfusion parameter map, specifically including:

[0062] Step S311: Based on the filtered three-dimensional dynamic CT perfusion image, obtain the tissue time density curve, arterial input function, and venous output function;

[0063] In step S311, the arterial input function AIF(t), the venous output function VOF(t), and the tissue time density curve are all signal-time curves of contrast agent concentration.

[0064] This method uses k-means clustering or other methods based on tissue-density time curve characteristics to automatically select points on the middle cerebral artery (MCA) as the global arterial input function (AIF) and points on the superior sagittal sinus (SSS) as the global venous output function (VOF).

[0065] Step S312: Correct the arterial input function using the venous output function to obtain the corrected arterial input function;

[0066] Because CT perfusion images have a relatively large slice thickness (generally around 5 mm), there is a significant partial volume effect (PVE), resulting in the contrast agent concentration at arterial points obtained from CT perfusion images being lower than the actual contrast agent concentration at the arterial points. Therefore, this method uses the venous output function (VOF) to correct the arterial input function (AIF).

[0067] AIF corrected (t) = PVE * AIF(t)

[0068] AIF corrected This is the arterial input function after PVE correction (adjustment), where PVE is the correction coefficient:

[0069]

[0070] Step S313: Obtain the residual function based on the tissue time density curve and the corrected arterial input function;

[0071] because Therefore, the residual function R(t) can be obtained from the tissue-density time curve C(t) and the corrected arterial input function AIF(t).

[0072] Step S314: Obtain the perfusion parameter map based on the residual function. The perfusion parameter map includes the cerebral blood flow (CBF) parameter map and the residual function peak time (Tmax) parameter map.

[0073] The residual function R(t) is calculated using methods such as singular value decomposition (SVD) or Fourier transform (FT), and then the perfusion parameter map is obtained: cerebral blood volume. Cerebral blood flow (CBF) = max(R(t)), mean transit time Residual function peak time Tmax = arg max t (R(t)). Completing step S314 completes the process. Figure 2 The diagram shows the calculated infusion parameters.

[0074] See Figure 3 Step S320, obtaining the lesion area includes obtaining the infarct core area, which specifically includes:

[0075] S321, Obtain the three-dimensional connected domain of the low-perfusion zone: Binarize, close, and fill holes in the Tmax parameter map of each layer sequentially using the first threshold, and then combine them to obtain the three-dimensional connected domain of the low-perfusion zone.

[0076] For each layer, the Tmax parameter map is binarized using Tmax>6s as the first threshold to obtain region A; closing operations and hole filling are performed on region A of each layer to obtain region B of that layer; all layer regions B are combined into three-dimensional volume data, and three-dimensional connected components are extracted from it. In some embodiments, the N largest connected components are retained to obtain the three-dimensional connected components of the low-perfusion zone. That is, the first threshold value is 6s. (Completing step S321 completes the process.) Figure 3 (Identification of low-perfusion areas)

[0077] S322, Obtain the normal CBF reference benchmark: Divide the brain CT perfusion images of each slice into left and right hemispheres. Based on the proportion of voxels with residual function peak times greater than the first threshold in both hemispheres, classify the left and right hemispheres as abnormal and normal sides. Calculate the average cerebral blood flow of voxels with blood flow less than the second threshold in the normal side. This average value is used as the normal CBF reference benchmark. (Completing step S322 completes the process.) Figure 3 (This involves identifying the normal and abnormal sides, as well as determining the normal CBF value).

[0078] The division into left and right hemispheres can be accomplished based on the symmetry axis obtained earlier, or by using other feasible methods from existing technologies. In this step, the left and right hemispheres of each layer are traversed, and the number of voxels with Tmax > 6s is counted. The side with more voxels with Tmax > 6s is determined as the abnormal side, and the other side as the normal side. That is, based on the proportion of voxels with residual function peak time Tmax greater than the first threshold on both sides of the brain, the sides are classified as abnormal and normal. All voxels on the normal side are traversed, and the average cerebral blood flow (CBF) of all voxels with Tmax < 4s is calculated. This average CBF value is taken as the normal CBF, i.e., as the normal CBF reference benchmark (normal cerebral blood flow reference benchmark).

[0079] S323, Obtain the infarct core region: Obtain the relative CBF parameter map, which is a ratio map of cerebral blood flow to normal CBF reference for each voxel. Use the third threshold to perform binarization, closure operation and hole filling on the relative CBF parameter map in sequence. After combination, take the intersection with the three-dimensional connected domain of the low perfusion area to obtain the three-dimensional intersection region. Perform closure operation and hole filling on the three-dimensional intersection region of each layer in sequence to obtain the infarct core region.

[0080] In step S321 above, the low perfusion zone is defined as the region where Tmax > the first threshold (e.g., 6s). In this step, the infarct core is defined as the region where the relative CBF (i.e., the relative CBF parameter map) < the third threshold (e.g., 30%).

[0081] For each layer, the CBF parameter map is binarized using a relative CBF < 30%, resulting in region P, where relative CBF is defined as the ratio of the CBF of each voxel to the normal CBF. The intersection of region P (relative CBF < 30%) with the low-perfusion region of that layer is calculated. Ensuring that the region with relative CBF < 30% is within the low-perfusion region of the layer, the intersection region Q is obtained. Closure operations and void filling are performed on region Q of each layer to obtain region R of that layer. All layer regions R are combined into 3D volume data, and 3D connected components are extracted to obtain the infarct core region. Obtaining the infarct core region also includes further filtering and retaining the N connected components with the largest volume as the infarct core region.

[0082] Step S330, obtaining the lesion area includes obtaining the penumbra region, which specifically includes: the three-dimensional connected domain of the hypoperfusion area and the infarct core region (e.g. Figure 4 As shown), the penumbra region is obtained. The penumbra is the low-perfusion region (e.g., Figure 5 The region shown is not part of the infarct core.

[0083] In this step, the volume V of the infarct core region is calculated based on information such as the spatial size of the voxel points and the interslice spacing of the CT perfusion images. core and low-perfusion zone volume V ischemia The volume of the half-dark band is V. penumbra =V ischemia -V core .

[0084] This application provides various embodiments of a method for obtaining perfusion parameter maps and lesion regions. The methods of each embodiment can automatically identify and acquire brain CT perfusion images after acquiring patient images via a PACS system or other means; they can perform image preprocessing on the CT perfusion images and use 4D filtering methods to filter and reduce noise in the CT perfusion images, improving the signal-to-noise ratio and calculation accuracy; they can acquire arterial input functions and venous output functions, and correct for partial volume effects of the arterial input function; they can calculate perfusion parameter maps and identify the infarct core and ischemic penumbra region and volume.

[0085] The methods described in these embodiments are fully automated, and the application of techniques such as automatic perfusion image recognition and 4D filtering enables this method to process brain CT perfusion images more quickly and accurately, and to identify the lesion areas required for clinical diagnosis. Therefore, this method can provide more imaging information, helping doctors to make better diagnoses of patients with acute stroke.

[0086] The methods in each embodiment can automatically complete the post-processing of brain CT perfusion images, including image preprocessing and parametric map calculation, and can obtain more accurate locations and volumes of the infarct core and penumbra based on the parametric maps. The method uses a 4D temporal-spatial coupled filtering method that incorporates temporal information, which can better reduce noise in CT perfusion images, improve the image signal-to-noise ratio, and thus obtain more accurate parametric map results; at the same time, it can accurately and quickly identify the location and size of the infarct core and penumbra from the parametric map results, providing more information for clinical treatment.

[0087] It should be understood that, 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.

[0088] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a method for obtaining perfusion parameter maps and lesion areas. The display screen can be a liquid crystal display (LCD) or an e-ink display. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0089] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0090] Step S100: Obtain three-dimensional dynamic CT perfusion images based on brain CT perfusion images at different times;

[0091] Step S200: Filter each voxel in the three-dimensional dynamic CT perfusion image to obtain the filtered three-dimensional dynamic CT perfusion image. The filtering process includes: at different times, for each current voxel, filtering is performed using the spatial similarity and temporal intensity similarity of the voxels adjacent to the current voxel.

[0092] Step S300: Based on the filtered three-dimensional dynamic CT perfusion image, obtain the perfusion parameter map and the lesion area.

[0093] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0094] 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.

[0095] 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 obtaining perfusion parameter maps and lesion areas, characterized in that, include: Three-dimensional dynamic CT perfusion images based on brain CT perfusion images at different times are obtained; wherein, the three-dimensional dynamic CT perfusion image at each time moment is formed by sequentially performing image registration, skull removal, and lateral ventricle removal on the brain CT perfusion image at that time moment. Each voxel in the three-dimensional dynamic CT perfusion image is filtered to obtain a filtered three-dimensional dynamic CT perfusion image. The filtering process includes: at different times, for each current voxel, filtering is performed using the spatial similarity and temporal intensity similarity of the voxel adjacent to the current voxel; wherein, the temporal intensity similarity is the similarity between the curves of the CT values ​​of the current voxel and the adjacent voxels changing over time. Based on the filtered three-dimensional dynamic CT perfusion images, perfusion parameter maps and lesion areas are obtained.

2. The method for obtaining perfusion parameter maps and lesion areas according to claim 1, characterized in that, The temporal intensity similarity is obtained by the following formula: In the formula, For the current voxel, The adjacent voxels of the current voxel The standard deviation of time-intensity similarity. It is an exponential function. voxels Adjacent voxels The mean square error of the time-intensity curve.

3. The method for obtaining perfusion parameter maps and lesion areas according to claim 2, characterized in that, The mean square error is obtained by the following formula: In the formula, T represents the total number of different moments in each slice of the three-dimensional dynamic CT perfusion image. For spatial location The intensity value of voxel p at time t. For spatial location The intensity value of voxel n at time t.

4. The method for obtaining perfusion parameter maps and lesion areas according to claim 1, characterized in that, The spatial similarity is obtained by the following formula: In the formula, voxels Adjacent voxels Euclidean distance, represents the standard deviation of spatial similarity.

5. The method for obtaining perfusion parameter maps and lesion areas according to claim 1, characterized in that, The filtering process is performed using the following formula: ； in, Here, X, Y, and Z are the normalization coefficients, and X, Y, and Z are the preset radii of the filter kernel. For the current voxel, The adjacent voxels of the current voxel, voxels With voxels Temporal intensity similarity voxels With voxels Spatial similarity, Let be the coordinates of the three spatial positions of voxel n. Let be the coordinates of the three spatial positions of voxel p. For the variable used for accumulation, Spatial location The intensity value at time t, This represents the intensity value at time t after voxel p filtering.

6. The method for obtaining perfusion parameter maps and lesion areas according to claim 1, characterized in that, Based on the filtered 3D dynamic CT perfusion images, perfusion parameter maps are obtained, specifically including: Based on the filtered three-dimensional dynamic CT perfusion images, tissue time density curves, arterial input functions, and venous output functions are obtained. The arterial input function is modified using the venous output function to obtain the modified arterial input function; The residual function is obtained based on the tissue time density curve and the modified arterial input function; The perfusion parameter map is obtained based on the residual function, and the perfusion parameter map includes cerebral blood flow. Parameter plot and peak time of residual function Parameter diagram.

7. The method for obtaining perfusion parameter maps and lesion areas according to claim 1, characterized in that, Obtaining the lesion area includes obtaining the infarct core area, specifically including: Obtaining the three-dimensional connected components of low-perfusion regions: using a first threshold for each layer The parameter map is binarized, closed, and filled with holes in sequence, and then combined to obtain a three-dimensional connected domain in the low-infusion zone. Get normal Reference benchmark: Brain CT perfusion images at each slice are divided into left and right hemispheres. Based on the proportion of voxels with residual function peak times greater than a first threshold in each hemisphere, the left and right hemispheres are classified as abnormal and normal sides. The average cerebral blood flow of voxels with blood flow less than a second threshold in the normal side is calculated, and this average value is used as the normal threshold. Reference standard; Obtain the infarct core region: Obtain the relative Parameter diagram, the relative The parameter graph shows the cerebral blood flow of each voxel as a function of the normal range. Referring to the benchmark ratio chart, the relative ratio is adjusted using a third threshold. The parameter map is binarized, closed, and filled with holes in sequence. After being combined, it is intersected with the three-dimensional connected domain of the low-perfusion zone to obtain the three-dimensional intersection region. The three-dimensional intersection region of each layer is closed and filled with holes in sequence to obtain the infarct core region.

8. The method for obtaining perfusion parameter maps and lesion areas according to claim 7, characterized in that, Obtaining the lesion area includes obtaining the penumbra area, specifically including: The penumbra region is obtained by combining the three-dimensional connected domain of the low-perfusion zone and the infarct core region.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method for obtaining perfusion parameter maps and lesion areas as described in any one of claims 1 to 8.

Images