Method and computer readable storage medium for ct perfusion image parameter map correction
By automatically selecting AIF and VOF using machine learning and clustering algorithms, the problem of low efficiency and poor accuracy in CT perfusion image parametric map correction in existing technologies is solved, achieving fast and accurate perfusion parametric map correction and improving the quality of perfusion parametric maps.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU ARTERYFLOW TECH CO LTD
- Filing Date
- 2022-12-16
- Publication Date
- 2026-05-29
AI Technical Summary
In the prior art, the correction method for CT perfusion image parametric maps relies on manually selecting the arterial input function (AIF) and venous output function (VOF), which is inefficient and has poor repeatability. It cannot effectively correct the partial volume effect (PVE) and affects the accuracy of the perfusion parametric map.
An automated method based on machine learning and clustering algorithms was adopted. By baseline correction, integral screening, and roughness screening of density-time curves, the AIF and VOF were automatically selected from CT perfusion images using the maximum connected component algorithm. Partial volume effect correction was performed to obtain the corrected perfusion parameter map.
It achieves fully automated, rapid, and accurate selection of AIF and VOF, improves the quality of perfusion parameter maps, reduces the variability of manual intervention, enhances robustness to patient activity disturbances, and significantly improves the correction effect of perfusion parameter maps.
Smart Images
Figure CN115880159B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical image processing technology, and in particular to a method for correcting parametric maps of CT perfusion images and a computer-readable storage medium. Background Technology
[0002] Acute stroke is a cerebrovascular disease with high rates of disability and mortality. In recent years, the incidence of acute stroke has been gradually increasing, while treatment efficiency remains unsatisfactory, necessitating further improvement. Brain CT perfusion imaging is a routine method for examining blood perfusion in stroke patients. During the examination, a contrast agent is injected intravenously, followed by rapid, continuous head CT scans within a specific time window. As the contrast agent flows through, it appears as a high-signal shadow on the image, resulting in a four-dimensional CT perfusion image encompassing temporal and spatial dimensions. Perfusion parameters calculated from the perfusion images, such as cerebral blood volume (CBV), cerebral blood flow (CBF), mean transit time (MTT), and time to peak (Tmax) of the contrast agent residual function, are important indicators for assessing the patient's condition.
[0003] Deconvolution calculation methods are commonly used when processing CT perfusion images. This method is based on the contrast agent dilution theory, assuming that the time-dimensional change in contrast agent concentration observed in the tissue is c. t (t) depends on the tissue characteristics r(t) and the time dimension of contrast agent concentration change within the blood vessels supplying the tissue of interest c. a (t), the relationship between the three is represented in the form of convolution. Where c t (t) represents the tissue density-time curve; c a r(t) is the density-time curve of the supply artery, also known as the arterial input function (AIF); r(t) is the residual contrast agent curve in the tissue capillaries, which can be further calculated to obtain various perfusion parameters. Due to the partial volume effect (PVE) commonly present in tomographic imaging, the parameter maps obtained directly using AIF are usually blurry and cannot clearly show the distribution of each parameter.
[0004] To correct PVE and optimize CBV and CBF parameter plots, it is necessary to obtain the venous output function (VOF). The parameter plot correction formula is shown below: Therefore, selecting the Arterial Input Function (AIF) and Venous Output Function (VOF) is a necessary preprocessing step when performing computational analysis of cerebral perfusion images. The accuracy of AIF and VOF directly affects the correction of various parametric maps. AIF is the contrast agent time-density curve on the artery supplying the brain tissue, spatially located in the middle cerebral artery; VOF is the contrast agent time-density curve on the veins supplying the brain tissue, spatially located in the superior sagittal sinus. Incorrect selection of AIF / VOF will affect the correction of image parametric maps, leading to incorrect calculation of perfusion parameters and parametric maps, thus impacting patient condition assessment. Rapid and accurate selection of AIF / VOF is essential for the correction of CT perfusion image parametric maps. Most literature suggests that AIF should be selected on the M1 or M2 segment of the middle cerebral artery (MCA), and VOF should be selected on the superior sagittal sinus.
[0005] Currently, methods for selecting the Arterial Flow Index (AIF) generally include manual selection, selection using a weighted model of the curve's characteristic functions, and selection of the AIF after manually selecting the region of interest (ROI) and then using clustering methods. Manual AIF selection is inefficient and heavily reliant on the operator's experience, resulting in low repeatability. Furthermore, due to significant differences in arterial morphology among patients and issues during data acquisition, such as noise from unconscious patient movements during imaging, AIF algorithms based on morphological feature models may have lower accuracy in handling anomalous data. The AIF algorithm using clustering after manually selecting the ROI is a semi-automatic algorithm, also suffering from low efficiency and reliance on experience. Selection of the VOF (Voice of Interest) is similar to AIF, but there is limited literature on this topic, and it also suffers from the same drawbacks. Summary of the Invention
[0006] Therefore, it is necessary to provide a method for correcting CT perfusion image parametric maps to address the aforementioned technical problems.
[0007] The method for CT perfusion image parametric correction of this application is characterized by comprising:
[0008] Based on the original CT perfusion images after preprocessing, candidate layers of arteries and veins are obtained. Voxels in the candidate layers of arteries and veins are then subjected to baseline correction, integral screening, and roughness screening of density-time curves in sequence to obtain the first type of voxels.
[0009] Artery candidate point groups are obtained by screening from the first type of voxels, and the artery candidate point groups are analyzed by the maximum connected component algorithm to obtain the artery result point groups, and then the artery input function is obtained.
[0010] The vein candidate point group is obtained by screening from the first type of voxels, and the vein candidate point group is analyzed by the maximum connected component algorithm to obtain the vein result point group, and then the vein output function is obtained.
[0011] A CT perfusion image parameter map is obtained. Using the arterial input function and the venous output function, a partial volume effect correction is performed on the CT perfusion image parameter map to obtain a corrected CT perfusion image parameter map.
[0012] Optionally, the preprocessing includes motion correction, image filtering, and skull removal;
[0013] Based on the preprocessed raw CT perfusion images, candidate arterial and venous layers are obtained, specifically including:
[0014] Read the brain tissue mask after removing the skull from each layer to obtain the area of the brain tissue mask. Find the first distance from the layer with the largest area towards the skull base to filter and obtain the candidate arteriovenous layer.
[0015] Optionally, for the remaining voxels that have completed the integral screening, the integral value of their density-time curve is greater than that of the other voxels that were screened out.
[0016] For the remaining voxels that have completed the roughness screening, their roughness is less than that of the other voxels that were screened out.
[0017] Optionally, the roughness screening includes: normalizing the curve area of the remaining voxels after integral screening to obtain the roughness of the normalized curve, specifically obtained by the following formula:
[0018] In the formula, C' n ' orm (t) is the normalized curve C norm The second derivative of (t), and R(v) is the integral of the second derivative over the time dimension, used to represent the roughness of the normalized curve.
[0019] Optionally, a candidate artery point group is obtained by screening from the first type of voxels, specifically including:
[0020] The first cluster analysis of the first type of voxels is performed on the first cluster target number, and the second type of voxels with the smallest average first moment are selected.
[0021] The expected number of voxels with higher distribution density in the second type of voxels is selected and the voxels in the expected number of voxels are used as the third type of voxels.
[0022] The density-time curves of the third type of voxels are obtained, and a second clustering analysis is performed on the second clustering target number to screen out the category with the smallest average peak time. The voxels in this category form the artery candidate point group.
[0023] Optionally, the artery candidate point group is analyzed using the maximum connected component algorithm to obtain the artery result point group, and then the artery input function is obtained, specifically including:
[0024] The mask image of the artery candidate point group is obtained, and the connected region with the largest area is obtained by the maximum connected component algorithm as the artery result point group. The average density time curve of the artery result point group is the artery input function.
[0025] The candidate vein point group is analyzed using the maximum connected component algorithm to obtain the vein result point group, and then the vein output function is obtained, specifically including:
[0026] The mask image of the vein candidate point group is obtained, and the connected region with the largest area is obtained by the maximum connected region algorithm as the vein result point group. The average density time curve of the vein result point group is the vein output function.
[0027] Optionally, a group of candidate vein points is obtained by screening from the first type of voxels, specifically including:
[0028] A fourth type of voxel was obtained from the first type of voxels, where the centroid of brain tissue was biased towards the hindbrain.
[0029] A fifth type of voxel is obtained by screening from the fourth type of voxels, whose signal density integral and peak time satisfy preset conditions;
[0030] A third clustering analysis of the target number of the third clustering is performed on the fifth type of voxels to screen out the category with the largest average first moment. The voxels in this category form a vein candidate point group.
[0031] Optionally, the preset conditions include: S(v) > S(AIF), t p >t p (AIF), where: S(v) is the signal density integral of the density-time curve of each voxel in the fifth type of voxel, t p The peak time of the density-time curve of each voxel in the fifth type of voxel is given by t, where S(AIF) is the signal density integral of the arterial input function. p (AIF) is the peak time of the arterial input function.
[0032] Optionally, the CT perfusion image parameter map includes a cerebral blood volume parameter map and a cerebral blood flow parameter map;
[0033] Using the arterial input function and the venous output function, partial volume effect correction is performed on the CT perfusion image parameter map. Specifically, this includes: performing partial volume effect correction on the CT perfusion image parameter map using a partial volume effect coefficient, which is obtained by the following formula:
[0034] P = S(AIF) / S(VOF), where P is the partial volume effect coefficient, S(AIF) is the signal density integral of the arterial input function, and S(VOF) is the signal density integral of the venous output function.
[0035] 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 CT perfusion image parametric map correction method described in this application.
[0036] The method for CT perfusion image parametric correction in this application has at least the following effects:
[0037] The method for CT perfusion image parametric correction provided in this application is based on machine learning and clustering algorithms. The entire method can achieve fully automated calculation without human interaction or manual operation. It has a fast calculation speed and excellent results, and can improve the accuracy and speed of deconvolution methods in calculating brain tissue perfusion parameters.
[0038] The arterial input function and venous output function obtained in this application are superior, avoiding differences caused by manual selection. Furthermore, this application does not refer to morphological features in CT perfusion images when selecting the AIF / VOF curve, thus avoiding the influence of morphological differences in CT perfusion images and exhibiting good robustness to minor disturbances caused by patient activity. Excellent results have been obtained on various case data. Each embodiment utilizes the accurately obtained arterial input function and venous output function to perform PVE correction on the perfusion parameter map, with significant correction effects, greatly improving the quality of the perfusion parameter map. Attached Figure Description
[0039] Figure 1 This is a schematic flowchart of a CT perfusion image parameter map correction method in one embodiment of this application;
[0040] Figure 2 This is a flowchart of a CT perfusion image parameter map correction method in one embodiment of this application;
[0041] Figure 3 for Figure 1 A schematic diagram of the AIF and VOF curves after baseline correction;
[0042] Figure 4 This is a schematic diagram of an arterial result point group obtained in one embodiment of this application (the location indicated by the arrow in the figure);
[0043] Figure 5 This is a schematic diagram of a group of vein result points obtained in one embodiment of this application (the locations indicated by the arrows in the diagram);
[0044] Figure 6 This is a diagram of cerebral blood volume parameters before correction in one embodiment of this application;
[0045] Figure 7 This is a diagram of corrected cerebral blood volume parameters in one embodiment of this application;
[0046] Figure 8 This is a diagram of cerebral blood flow parameters before correction in one embodiment of this application;
[0047] Figure 9 This is a diagram of corrected cerebral blood flow parameters in one embodiment of this application;
[0048] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0049] 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.
[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.
[0051] In this application, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number or order of the indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0052] In this application, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a system, product, step, or apparatus that includes a series of units is not necessarily limited to those units that are explicitly listed, but may include other units that are not explicitly listed or that are inherent to such products or apparatus.
[0053] See Figure 1 and Figure 2 One embodiment of this application provides a method for CT perfusion image parameter map correction, including steps S100 to S400.
[0054] Step S100: Based on the original CT perfusion image after preprocessing, obtain the arteriovenous candidate layer, and sequentially perform baseline correction, integral screening, and roughness screening on the voxels in the arteriovenous candidate layer to obtain the first type of voxels.
[0055] Step S200: Select artery candidate point group from the first type of voxels, analyze the artery candidate point group using the maximum connected component algorithm, obtain artery result point group, and then obtain artery input function;
[0056] Step S300: Select the vein candidate point group from the first type of voxels, analyze the vein candidate point group using the maximum connected component algorithm, obtain the vein result point group, and then obtain the vein output function.
[0057] Step S400: Obtain the CT perfusion image parameter map. Using the arterial input function and the venous output function, perform partial volume effect correction on the CT perfusion image parameter map to obtain the corrected CT perfusion image parameter map.
[0058] In the following embodiments, each of steps S100 to S400 includes optional sub-steps or detailed steps.
[0059] Step S100 includes steps S110 to S120. Wherein:
[0060] Step S110: Based on the preprocessed original CT perfusion images, obtain the arteriovenous candidate layers. In step S110:
[0061] Preprocessing includes motion correction, image filtering, and skull removal. Specifically, the preprocessed four-dimensional CT perfusion images are read as the original CT perfusion images, which include three-dimensional CT perfusion images composed of cross-sections at different scan times. The four dimensions are time T, spatial slice number L, image height H, and image width W.
[0062] Obtaining candidate arterial and venous layers includes: reading the brain tissue mask after removing the skull from each layer, obtaining the area of the brain tissue mask, finding the first distance from the layer with the largest area towards the skull base, and filtering to obtain candidate arterial and venous layers. These candidate arterial and venous layers simultaneously serve as candidate arterial layers (middle cerebral artery (MCA) layer) and candidate venous layers.
[0063] The brain tissue mask obtained after skull removal from preprocessing is read and arranged in the direction from skull base to skull top. The area (L) of the brain tissue mask at each layer is calculated, and the maximum area is found by iterating through each layer. maxTowards the skull base, based on the slice spacing ΔL of the CT perfusion images, select n layers of brain tissue data such that the selected data covers approximately the first distance (e.g., 40 mm) of brain tissue in the slice dimension L, i.e., n = [40 / ΔL]. The selected n layers of brain tissue data are the arteriovenous candidate layers, which serve as input data for subsequent steps.
[0064] Step S120: Perform baseline correction, integral screening, and roughness screening on the voxels in the arteriovenous candidate layer according to the density-time curve to obtain the first type of voxels.
[0065] Baseline correction of density-time curves refers to obtaining the signal density-time curve C(t) of each voxel in each layer of brain tissue along its time dimension. The average signal density over the first m time points when the signal is stable is taken as the density baseline. m can be determined by assessing the degree of curve variation or empirically. Subtracting the corresponding baseline value from each point on the density-time curve C(t) of all voxels yields the following result: Figure 3 The baseline-corrected curve is shown, and all curve point values less than zero are set to zero.
[0066] On the one hand, for the remaining voxels that have completed the integral screening, their density-time curve integral values are greater than those of the other voxels that were screened out. Specifically, the signal density integral is obtained by integrating the density-time curves C(t) of all voxels. Sort the signal density integral S(v) of all voxels, filter out voxels with smaller S(v), and the filtering ratio parameter is p. S p S As a hyperparameter, the optimal p can be determined by adjusting the parameter and observing the precision of the results. S Parameters. The reference MCA has a relatively small proportion in brain tissue, p S A larger value can be selected, generally above 0.9, meaning the filtration ratio parameter is p. S The voxels.
[0067] On the other hand, for the remaining voxels that have completed roughness screening, their roughness is less than that of the other voxels that were screened out. Roughness screening includes: normalizing the curve area of the remaining voxels that have completed integral screening to obtain the roughness of the normalized curve, specifically obtained by the following formula: In the formula, C” norm (t) is the normalized curve C norm The second derivative of (t), and R(v) is the integral of the square of the second derivative over the time dimension, used to represent the roughness of the normalized curve.
[0068] During roughness screening, the curve roughness R(v) of the remaining voxels after integration screening is sorted, voxels with relatively large R(v) are filtered out, and voxels with relatively small R(v) are retained. The filtering ratio is p. R pR As a hyperparameter, the optimal p can be determined by adjusting the parameter and observing the precision of the results. R Parameters. To filter out as many noise curves as possible without removing too many AIF candidate curves, p R Choose the smaller value, generally below 0.2, that is, the filtration ratio is p. R The voxels.
[0069] Roughness screening requires area normalization, which prevents direct integral screening. This embodiment performs integral screening first, followed by roughness screening, thus saving steps in the automated processing flow.
[0070] Step S200 includes steps S210 to S220.
[0071] Step S210 involves selecting candidate artery points from the first group of voxels. This step uses the k-means++ clustering algorithm, a component of unsupervised machine learning, for cluster analysis. Two cluster analyses are performed on the first group of voxels to extract AIF curves and obtain the candidate artery points. Specifically, this includes steps S211 to S213.
[0072] Step S211: Perform the first cluster analysis on the first type of voxels based on the first cluster target number, and screen out the second type of voxels with the smallest average first moment.
[0073] Specifically, the normalized curve C of the first type of voxels that has completed roughness screening. norm (t) is used for the first cluster analysis. The first cluster analysis uses k-means++ as the initialization method for the cluster centers, and the target number for the first cluster is set to k1. k1 is a hyperparameter, and the optimal k1 parameter can be determined by adjusting the precision of the observation results. For example, k1 = 5.
[0074] The first cluster analysis divided the data into categories based on the target number of the first cluster. For each voxel within each category, the first moment was calculated: μ = ∑t·C norm (), and calculate the average first moment of each class after classification. Retain the average first moment Minimum category The voxels in the voxels were obtained, thus obtaining the second type of voxels.
[0075] Step S212: Select the layers with higher expected distribution density among the second type of voxels, and use the voxels in the layers with the expected number of voxels as the third type of voxels.
[0076] Specifically, the remaining voxels (mean first moment) are statistically analyzed. The distribution of the smallest voxel across each arterial and venous candidate layer is determined. The arterial and venous candidate layer with the highest expected number of remaining voxels (e.g., 2 layers) is retained, and the voxels distributed on it are classified as Class III voxels. The expected number can be adaptively adjusted based on the thickness of the original CT perfusion image.
[0077] Step S213: Obtain the density-time curve of the third type of voxels, perform a second cluster analysis of the second cluster target number, and screen out the category with the smallest average peak time. The voxels in this category form the artery candidate point group.
[0078] After restoring the normalized density-time curve area of the third type of voxels, the density-time curve C(t) is obtained again, and a second cluster analysis is performed. The second cluster analysis uses k-means++ as the initialization method for the cluster centers, and the target number for the second cluster is set to k2. k2 is a hyperparameter, and the optimal k2 parameter can be determined by adjusting the accuracy of the observation results, for example, setting k2 = 5.
[0079] The second cluster analysis divided the target number into several categories. For each category, the peak time (time to reach peak value) t was calculated for each voxel. p And calculate the average peak time for each category after classification. Retain average peak time Minimum category The voxels in the data serve as candidate sites for arterial infarction (AIF), thus forming a group of candidate arterial infarction sites.
[0080] From the perspective of screening methods, average first-order moment screening is more effective than average peak time screening. Layered screening narrows the spatial range, which is beneficial for clustering by average peak time screening to result in more concentrated midpoint sets. Following the steps provided in this embodiment sequentially can improve the speed of automated processing.
[0081] Step S220: Analyze the candidate artery point group using the maximum connected component algorithm to obtain, as shown below... Figure 4 The arterial result point group shown is used to obtain the arterial input function. Specifically, this includes: obtaining a mask image of the arterial candidate point group; obtaining the connected region with the largest area as the arterial result point group through the maximum connected region algorithm; and the average density time curve of the arterial result point group is the arterial input function.
[0082] A mask image of the artery candidate point group is constructed. Using a maximum connected component algorithm, such as the 8-connected maximum connected component algorithm, c connected components of the artery candidate point group are calculated, and the area Area(c) of each connected component is calculated. The artery candidate point group in the connected component with the largest area is selected as the AIF result point group. The average density-time curve of the AIF result point group is the artery input function (AIF curve).
[0083] Step S300 includes steps S310 to S320.
[0084] Step S310 involves selecting a group of candidate vein points from the first type of voxels, specifically including steps S311 to S313.
[0085] Step S311: Select fourth type voxels from the first type voxels, where the centroid of the brain tissue is biased towards the posterior brain direction (below the centroid of the brain tissue);
[0086] Step S312: Select fifth voxels from the fourth voxels that meet the preset conditions for signal density integral and peak time;
[0087] Specifically, the preset conditions used include: S(v) > S(AIF), t p >t p (AIF), where S(v) is the signal density integral of the density-time curve of each voxel in the fifth class of voxels, t p S(AIF) represents the peak time of the density-time curves of each voxel in the fifth class of voxels, and S(AIF) represents the signal density integral of the arterial input function. p (AIF) represents the peak time of the arterial input function. This step uses the signal density integral and peak time of the arterial input function based on human physiological parameters to perform preliminary screening of candidate voxels for veins, in order to quickly obtain the expected cluster of vein result points.
[0088] Step S313: Perform a third cluster analysis on the fifth type of voxels based on the target number of the third cluster, and select the category with the largest average first moment. The voxels in this category form a vein candidate point group.
[0089] Specifically, the normalized density-time curve C for the fifth type of voxel. norm (t) Perform a third cluster analysis, for example, using k-means++ cluster analysis. Use k-means++ as the initialization method for the cluster centers, and set the target number for the third cluster to k3. k3 is a hyperparameter, and the optimal k3 parameter can be determined by adjusting the accuracy of the observation results. For example, let k3 = 5.
[0090] The third cluster analysis divided the target number into three categories. For each voxel within each category, the first moment was calculated, i.e., μ = ∑t·C. norm (), and calculate the average first moment of each class after classification. Retain the average first moment Largest category The voxels in the data are used as VOF candidate sites, i.e., vein candidate sites.
[0091] Step S320: Analyze the candidate vein point group using the maximum connected component algorithm to obtain, as shown below... Figure 5The vein result point group shown is used to obtain the vein output function. Specifically, this includes: obtaining the mask image of the vein candidate point group, obtaining the connected region with the largest area as the vein result point group through the maximum connected region algorithm, and the average density time curve of the vein result point group is the vein output function.
[0092] A mask image of the vein candidate point group is constructed. Using a maximum connected component algorithm, such as the 8-connected maximum connected component algorithm, c connected components of the vein candidate point group are calculated, and the area Area(c) of each connected component is calculated. The VOF candidate points in the connected component with the largest area are selected as the VOF result point group. The average density-time curve of the VOF result point group is the vein output function VOF.
[0093] Step S400: Obtain the CT perfusion image parameter map. Using the arterial input function and venous output function, perform partial volume effect correction on the CT perfusion image parameter map to obtain the corrected CT perfusion image parameter map. The CT perfusion image parameter map includes, for example: Figure 6 The cerebral blood volume parameter diagram (CBV) shown is as follows: Figure 7 The diagram shows the cerebral blood flow parameters (CBF).
[0094] Partial volume effect correction is performed on the CT perfusion image parametric map using arterial input and venous output functions. Specifically, this includes using a partial volume effect coefficient to correct for the partial volume effect in the CT perfusion image parametric map. The partial volume effect coefficient is obtained using the following formula:
[0095] P = S(AIF) / S(VOF), where P is the partial volume effect coefficient, S(AIF) is the signal density integral of the arterial input function, and S(VOF) is the signal density integral of the venous output function. Then, PVE correction is performed on the cerebral blood volume parameter map CBV and the cerebral blood flow parameter map CBF using the formulas CBV′ = CBV*P and CBF′ = CBF*P. CBV′ is the corrected cerebral blood volume parameter map, as shown below. Figure 7 As shown, CBF′ is a graph of the corrected cerebral blood flow parameters, as... Figure 9 As shown. Signal density integral of the arterial input function (AIF curve). Signal density integral of venous output function (VOF curve)
[0096] In one embodiment of this application, the method for CT perfusion image parametric map correction mainly includes the following steps one through ten. Steps one through four correspond to step S100. Steps five and six correspond to step S200. Steps seven through nine correspond to step S300. Step ten corresponds to step S400.
[0097] The first step is to read the pre-processed 4D CT perfusion images, calculate the area of each brain tissue region, and select data covering a 40mm thick layer of brain tissue from the layer with the largest brain tissue area towards the skull base as the arterial and venous candidate layers. The arterial and venous candidate layers serve as both the middle cerebral artery (MCA) candidate layers and the venous candidate layers.
[0098] The second step is to use the average signal density at the moment when the signal is stable as the baseline to perform baseline correction on the density-time curve of each voxel in each layer; this completes the baseline correction in step S100.
[0099] The third step is to integrate the density-time curves of all voxels to obtain the signal density integral, and then filter out the voxels with smaller signal density integrals; this completes the integration screening in step S100.
[0100] The fourth step is to calculate the roughness of the remaining voxel density time normalization curve and filter out the voxels with larger roughness; thus, the first type of voxels in step S100 are obtained.
[0101] The fifth step involves using the kmeans++ machine learning algorithm to perform two cluster analyses on the density-time curves of the remaining voxels. The first analysis retains the cluster with the smallest average first moment, and the second analysis retains the cluster with the smallest average peak time as the AIF candidate cluster. This yields the artery candidate cluster in S200.
[0102] Step 6: Construct a mask image of AIF candidate points on the MCA candidate layer. Using the maximum connected component algorithm, select the AIF candidate points in the connected component with the largest area as the AIF result point group. The average density time curve of the AIF result point group is the arterial input function AIF. Thus, the arterial input function in S200 is obtained.
[0103] Step 7: Based on the AIF curve results, select VOF candidate voxels from the voxel density time curves obtained in Step 4. The following conditions must be met simultaneously: ① the signal density integral is greater than the signal density integral of the AIF curve, ② the peak time is greater than the peak time of the AIF curve, and ③ the voxel is located below the centroid of the brain tissue.
[0104] Step 8: Use the kmeans++ machine learning algorithm to perform cluster analysis on the density time curves of VOF candidate voxels, and retain the class with the largest average first moment as the VOF candidate point group; thus, the vein candidate point group in S300 is obtained.
[0105] Step 9: Construct a VOF candidate point mask image. Using the maximum connected component algorithm, select the VOF candidate points in the connected component with the largest area as the VOF result point group. The average density time curve of the VOF result point group is the vein output function VOF. Thus, the vein output function in S300 is obtained.
[0106] Step 10: Calculate the signal density integrals of AIF and VOF, calculate the correction coefficients, and use the correction coefficients to complete the PVE correction of the CBV and CBF parameter diagrams. This completes step S400.
[0107] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As 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 the computer program is executed by the processor, it implements a method for correcting parametric maps of CT perfusion images. 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.
[0108] 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:
[0109] Step S100: Based on the original CT perfusion image after preprocessing, obtain the arteriovenous candidate layer, and sequentially perform baseline correction, integral screening, and roughness screening on the voxels in the arteriovenous candidate layer to obtain the first type of voxels.
[0110] Step S200: Select artery candidate point group from the first type of voxels, analyze the artery candidate point group using the maximum connected component algorithm, obtain artery result point group, and then obtain artery input function;
[0111] Step S300: Select the vein candidate point group from the first type of voxels, analyze the vein candidate point group using the maximum connected component algorithm, obtain the vein result point group, and then obtain the vein output function.
[0112] Step S400: Obtain the CT perfusion image parameter map. Using the arterial input function and the venous output function, perform partial volume effect correction on the CT perfusion image parameter map to obtain the corrected CT perfusion image parameter map.
[0113] The CT perfusion image parameter correction methods provided in the embodiments of this application are based on machine learning and clustering algorithms. The entire method can achieve fully automated calculation without human interaction or manual operation, with fast calculation speed and superior results, improving the accuracy and speed of brain tissue perfusion parameter calculation using deconvolution methods. The kmeans++ algorithm used in each cluster analysis has a fast classification convergence speed, further shortening the process time and significantly improving the speed of CT perfusion imaging analysis, thus accelerating the diagnostic process.
[0114] The arterial input function and venous output function (AIF / VOF curves) obtained in each embodiment are superior, avoiding differences caused by manual selection. Furthermore, this method does not refer to morphological features in CT perfusion images when selecting the AIF / VOF curve, thus avoiding the influence of morphological differences in CT perfusion images. It also exhibits good robustness to minor disturbances caused by patient activity and has achieved good results across various case datasets.
[0115] Each embodiment utilizes the accurately obtained arterial input function and venous output function to perform PVE correction on the perfusion parameter map, with significant correction effect, greatly improving the quality of the perfusion parameter map.
[0116] 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 memory bus dynamic RAM (RDRAM), etc.
[0117] 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.
[0118] 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 correcting parametric maps of CT perfusion images, characterized in that, include: Based on the original CT perfusion images after preprocessing, candidate layers of arteries and veins are obtained. Voxels in the candidate layers of arteries and veins are then subjected to baseline correction, integral screening, and roughness screening of density-time curves in sequence to obtain the first type of voxels. Artery candidate point groups are obtained by screening from the first type of voxels, and the artery candidate point groups are analyzed by the maximum connected component algorithm to obtain the artery result point groups, and then the artery input function is obtained. The vein candidate point group is obtained by screening from the first type of voxels, and the vein candidate point group is analyzed by the maximum connected component algorithm to obtain the vein result point group, and then the vein output function is obtained. A CT perfusion image parameter map is obtained. Using the arterial input function and the venous output function, a partial volume effect correction is performed on the CT perfusion image parameter map to obtain a corrected CT perfusion image parameter map.
2. The method for CT perfusion image parametric map correction according to claim 1, characterized in that, The preprocessing includes motion correction, image filtering, and skull removal; Based on the preprocessed raw CT perfusion images, candidate arterial and venous layers are obtained, specifically including: Read the brain tissue mask after removing the skull from each layer to obtain the area of the brain tissue mask. Find the first distance from the layer with the largest area towards the skull base to filter and obtain the candidate arteriovenous layer.
3. The method for CT perfusion image parametric map correction according to claim 1, characterized in that, For the remaining voxels that have completed the integral screening, the integral value of their density-time curve is greater than that of the other voxels that were screened out. For the remaining voxels that have completed the roughness screening, their roughness is less than that of the other voxels that were screened out.
4. The method for CT perfusion image parametric map correction according to claim 1, characterized in that, The roughness screening includes: normalizing the curve area of the remaining voxels after integral screening to obtain the roughness of the normalized curve, specifically obtained by the following formula: In the formula, Normalized curve The second derivative, The integral of the second derivative over the time dimension is used to represent the roughness of the normalized curve.
5. The method for CT perfusion image parametric map correction according to claim 1, characterized in that, Artery candidate point clusters are obtained by screening the first type of voxels, specifically including: The first cluster analysis of the first type of voxels is performed on the first cluster target number, and the second type of voxels with the smallest average first moment are selected. The expected number of voxels with higher distribution density in the second type of voxels is selected and the voxels in the expected number of voxels are used as the third type of voxels. The density-time curves of the third type of voxels are obtained, and a second clustering analysis is performed on the second clustering target number to screen out the category with the smallest average peak time. The voxels in this category form the artery candidate point group.
6. The method for CT perfusion image parametric map correction according to claim 1, characterized in that, The artery candidate point group is analyzed using the maximum connected component algorithm to obtain the artery result point group, and then the artery input function is obtained, specifically including: The mask image of the artery candidate point group is obtained, and the connected region with the largest area is obtained by the maximum connected component algorithm as the artery result point group. The average density time curve of the artery result point group is the artery input function. The candidate vein point group is analyzed using the maximum connected component algorithm to obtain the vein result point group, and then the vein output function is obtained, specifically including: The mask image of the vein candidate point group is obtained, and the connected region with the largest area is obtained by the maximum connected region algorithm as the vein result point group. The average density time curve of the vein result point group is the vein output function.
7. The method for CT perfusion image parametric map correction according to claim 1, characterized in that, The candidate vein point group is obtained by screening the first type of voxels, specifically including: A fourth type of voxel was obtained from the first type of voxels, where the centroid of brain tissue was biased towards the hindbrain. A fifth type of voxel is obtained by screening from the fourth type of voxels, whose signal density integral and peak time satisfy preset conditions; A third clustering analysis of the target number of the third clustering is performed on the fifth type of voxels to screen out the category with the largest average first moment. The voxels in this category form a vein candidate point group.
8. The method for CT perfusion image parametric map correction according to claim 7, characterized in that, The preset conditions include: , In the formula: The signal density integral is the density-time curve of each voxel in the fifth type of voxels. This refers to the peak time of the density-time curve of each voxel in the fifth type of voxels. The signal density integral of the arterial input function, The peak time of the arterial input function.
9. The method for CT perfusion image parametric map correction according to claim 1, characterized in that, The CT perfusion imaging parameter map includes a cerebral blood volume parameter map and a cerebral blood flow parameter map; Using the arterial input function and the venous output function, partial volume effect correction is performed on the CT perfusion image parameter map. Specifically, this includes: performing partial volume effect correction on the CT perfusion image parameter map using a partial volume effect coefficient, which is obtained by the following formula: In the formula, This is the partial volume effect coefficient. The signal density integral of the arterial input function, The signal density integral of the vein output function.
10. 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 CT perfusion image parametric map correction as described in any one of claims 1 to 9.