Methods for predicting thrombectomy intra- or post-procedure tissue hypoperfusion and related devices

Abstract

The embodiment of the present disclosure provides a method for predicting tissue hypoperfusion during or after thrombectomy and a related device, which relates to the technical field of medical image processing, and comprises the following steps: acquiring a digital subtraction angiography sequence of a patient during or after thrombectomy; dividing the digital subtraction angiography sequence into a plurality of target regions based on the structure of the brain blood supply system; for each target region, determining a corresponding hemodynamic characteristic parameter based on the law of contrast agent concentration changing with time; for different target regions, constructing a time difference matrix based on blood flow differences, and determining a spatial heterogeneity characteristic parameter based on the time difference matrix; constructing a multi-dimensional feature vector using the hemodynamic characteristic parameter, the time difference matrix and the spatial heterogeneity characteristic parameter, inputting the multi-dimensional feature vector into a tissue hypoperfusion prediction model, and obtaining a prediction result of the patient's occurrence of tissue hypoperfusion output by the tissue hypoperfusion prediction model. The present application can improve the accuracy and consistency of hypoperfusion prediction.

Description

Technical Field

[0001] This disclosure relates to the field of medical image processing technology, and in particular to a method and related apparatus for predicting tissue hypoperfusion during or after thrombectomy. Background Technology

[0002] In patients with acute ischemic stroke, cerebral blood vessels are blocked by thrombi, leading to cerebral ischemia and hypoxia. Mechanical thrombectomy uses interventional devices to directly remove or break up the thrombus blocking the blood vessel, rapidly restoring cerebral blood flow. Although the recanalization rate of large vessels after endovascular mechanical thrombectomy can reach over 85%-90%, a rate as high as 20%-50% still fails to achieve adequate blood flow reperfusion at the cerebral tissue level.

[0003] Traditional assessments of cerebral blood flow reperfusion after endovascular mechanical thrombectomy are mainly based on the eTICI scoring system. This scoring system relies on the surgeon's experience and visual interpretation, which is easily affected by subjectivity. Furthermore, the assessment of perfusion at the brain tissue level is usually completed 2-24 hours after mechanical thrombectomy, and cannot be obtained immediately during or after the procedure. This leads to a delay in the identification of low perfusion, making it impossible to make immediate adjustments to the treatment strategy during the procedure or to provide immediate intervention after the procedure. Summary of the Invention

[0004] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, this disclosure provides a method and related device for predicting tissue hypoperfusion during or after thrombectomy.

[0005] This disclosure provides a method for predicting tissue hypoperfusion during or after thrombectomy. The method includes: acquiring digital subtraction angiography (DSA) sequences during or after thrombectomy; wherein the DSA sequence includes multiple consecutive two-dimensional images; dividing the DSA sequence into multiple target regions based on the structure of the cerebral blood supply system; for each target region, determining corresponding hemodynamic characteristic parameters based on the time-varying pattern of contrast agent concentration; for different target regions, constructing a time difference matrix based on blood flow differences, and determining spatial heterogeneity characteristic parameters based on the time difference matrix; constructing a multidimensional feature vector using the hemodynamic characteristic parameters, the time difference matrix, and the spatial heterogeneity characteristic parameters, and inputting the multidimensional feature vector into a tissue hypoperfusion prediction model to obtain the predicted result of tissue hypoperfusion in the patient output by the tissue hypoperfusion prediction model; wherein the tissue hypoperfusion prediction model is trained using multidimensional feature vector samples from historical patients during or after thrombectomy and the corresponding tissue hypoperfusion occurrence results, and the predicted result of tissue hypoperfusion is an indicator used to characterize the development trend of tissue hypoperfusion.

[0006] In this implementation, digital subtraction angiography sequences are acquired and analyzed immediately during or after thrombectomy to achieve real-time assessment of cerebral hypoperfusion, providing a valuable window for timely intervention. Target regions are divided according to the structure of the cerebral blood supply system to enhance the correlation between characteristic parameters and tissue perfusion status. A multidimensional feature vector is constructed by combining single-region hemodynamic characteristic parameters, cross-regional blood flow time difference matrices, and spatial heterogeneity characteristic parameters. The model automatically performs integrated analysis of multidimensional feature parameters and determines the likelihood of future hypoperfusion, eliminating subjective errors from manual assessment and improving the accuracy and consistency of hypoperfusion prediction. This application enables early identification of tissue hypoperfusion, guiding intraoperative decision-making and individualized treatment.

[0007] In one possible implementation, the digital subtraction angiography sequence is divided into multiple target regions based on the structure of the brain's blood supply system, including: dividing the digital subtraction angiography sequence into multiple arterial regions according to the arterial blood supply system of the brain's blood supply system, the arterial regions including the origin of the internal carotid artery, the intracranial internal carotid artery segment, the M1 segment of the middle cerebral artery, the M2 segment of the middle cerebral artery, or cortical branch segments; and dividing the digital subtraction angiography sequence into venous regions according to the venous drainage system of the brain's blood supply system, the venous regions including the superior sagittal sinus segment.

[0008] In this approach, by refining the arterial regions such as the initial segment of the internal carotid artery, the intracranial segment, the middle cerebral artery, and cortical branches, the differences in blood perfusion in different arterial segments can be captured in a targeted manner, providing anatomical anchors for the regional quantitative analysis of parameters and helping to identify low-perfusion sites in the arterial blood supply layer; it also incorporates the superior sagittal sinus segment venous region, taking into account the synergistic assessment of arterial blood supply and venous drainage.

[0009] In one possible implementation, the corresponding hemodynamic characteristic parameters are determined based on the change of contrast agent concentration over time, including: obtaining the pixel density value of each two-dimensional image in the digital subtraction angiography sequence; analyzing the change of pixel density value over time in multiple consecutive two-dimensional images to construct a time-density curve; calculating the time of maximum contrast agent concentration based on the time-density curve to obtain the peak time; and determining multiple hemodynamic characteristic parameters based on the time-density curve and the peak time.

[0010] In this implementation, by extracting pixel density values ​​frame by frame and constructing time-density curves by combining continuous frame temporal analysis, data support can be provided for parameter calculation. Based on the peak time, multi-dimensional hemodynamic parameters are derived, which can comprehensively reflect the perfusion efficiency, blood flow velocity and microcirculation status of different target areas, providing a quantitative basis for subsequent assessment of brain tissue perfusion status, eliminating subjective errors of manual assessment, and improving the accuracy and consistency of low perfusion prediction.

[0011] In one possible implementation, multiple hemodynamic characteristic parameters are determined based on the time-density curve and the time to peak, including: determining the relative time to peak based on the difference between the time to peak and the time to peak in a reference region; determining the average transit time based on the ratio of the area under the time-density curve to the peak value; determining the cerebral circulation time of each arterial region based on the difference between the time to peak in the arterial region and the reference region; and determining the microvascular transit time based on the difference between the time to peak in the arterial region and the venous region; wherein the reference region is the initial segment of the internal carotid artery.

[0012] This approach comprehensively covers core dimensions such as perfusion rate, microcirculation function, and blood flow transmission delay. The parameters are logically complementary and progressive, adapting to the quantitative analysis needs of image sequences. It can correlate different physiological stages of brain tissue perfusion, providing objective quantitative evidence for distinguishing ineffective reperfusion and identifying low perfusion types postoperatively. It solves the problems of traditional single parameters, lack of microcirculation assessment, and poor comparability, achieving standardized intraoperative and postoperative real-time perfusion assessment.

[0013] In one possible implementation, for different target regions, a time difference matrix is ​​constructed based on blood flow differences, including: traversing all target regions, calculating the difference in peak time between every two target regions to obtain multiple blood flow time differences; and constructing an antisymmetric time difference matrix based on multiple blood flow time differences.

[0014] In this implementation, by traversing all target regions, calculating the peak time difference between each pair and constructing a matrix, the differences in blood perfusion timing between regions can be systematically and quantitatively characterized, and the correlation patterns and abnormal characteristics of blood perfusion between regions can be effectively explored.

[0015] In one possible implementation, spatial heterogeneity characteristic parameters are determined based on a time difference matrix, including: determining blood flow delay regions from multiple target regions based on the time difference matrix; determining the delay volume fraction based on the proportion of blood flow delay regions; determining the spatial dispersion based on the mean and standard deviation of the peak time of blood flow delay regions; and determining the spatial entropy based on the disorder of hemodynamic characteristic parameters of blood flow delay regions.

[0016] In this implementation, multiple parameters are derived based on the time difference matrix. The features have strong correlations and high data consistency, which can be directly used as effective inputs for subsequent prediction models. This provides key spatial dimension features to deeply explore the spatial patterns of perfusion anomalies and improve the accuracy of low perfusion prediction.

[0017] In one possible implementation, the method for predicting tissue hypoperfusion during or after thrombectomy further includes: generating a visualization based on hemodynamic characteristic parameters; wherein the visualization includes a parametric heatmap and a local perfusion time map; and generating and presenting a tissue hypoperfusion report by combining the predicted results of tissue hypoperfusion in the patient with the visualization.

[0018] In this implementation, hemodynamic parameters are transformed into visual graphs such as parametric heatmaps and local perfusion time maps, and a report is generated by combining the low perfusion prediction results, which is both intuitive and practical.

[0019] This disclosure also provides a device for predicting tissue hypoperfusion during or after thrombectomy. The device includes: an acquisition module for acquiring digital subtraction angiography (DSA) sequences during or after thrombectomy; wherein the DSA sequence comprises multiple consecutive two-dimensional images; a segmentation module for dividing the DSA sequence into multiple target regions based on the structure of the cerebral blood supply system; a first analysis module for determining corresponding hemodynamic characteristic parameters for each target region based on the change in contrast agent concentration over time; and a second analysis module for... Within the same target region, a time difference matrix is ​​constructed based on blood flow differences, and spatial heterogeneity feature parameters are determined based on the time difference matrix. The prediction module is used to construct a multidimensional feature vector using hemodynamic feature parameters, the time difference matrix, and spatial heterogeneity feature parameters. The multidimensional feature vector is input into the tissue hypoperfusion prediction model to obtain the prediction result of the patient's tissue hypoperfusion. The tissue hypoperfusion prediction model is trained using multidimensional feature vector samples from historical patients during or after thrombectomy and the occurrence results of tissue hypoperfusion. The prediction result of tissue hypoperfusion is an indicator information used to characterize the development trend of tissue hypoperfusion.

[0020] This disclosure also provides a computing device, which includes: a processor; a memory for storing processor-executable instructions; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the method for predicting tissue hypoperfusion during or after thrombectomy as provided in this disclosure.

[0021] This disclosure also provides a computer-readable storage medium storing a computer program for executing a method for predicting tissue hypoperfusion during or after thrombectomy as provided in this disclosure. Attached Figure Description

[0022] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0023] Figure 1 A flowchart illustrating a method for predicting tissue hypoperfusion during or after thrombectomy, provided in an embodiment of this disclosure; Figure 2 A schematic diagram of the structure of a device for predicting tissue hypoperfusion during or after thrombectomy, provided in an embodiment of this disclosure; Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of the present disclosure. Detailed Implementation

[0024] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0025] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0026] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0027] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0028] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0029] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0030] Despite a large vessel recanalization rate exceeding 85% after endovascular mechanical thrombectomy (EVT) for acute ischemic stroke, approximately 50% of patients still experience "ineffective recanalization," meaning that despite large vessel recanalization, neurological function does not achieve a satisfactory prognosis. A significant reason for this is the ineffective restoration of cerebral blood flow after large vessel recanalization. Current assessments primarily rely on the eTICI scoring system, but this method is based on the surgeon's subjective visual judgment and does not reflect microcirculation perfusion. Perfusion assessment at the cerebral tissue level is typically completed 2–24 hours after EVT, making it impossible to obtain immediate results intraoperatively or postoperatively. This leads to delayed identification of hypoperfusion, missed opportunities for optimal intervention, and an inability to make timely adjustments to treatment strategies intraoperatively.

[0031] To address the aforementioned issues, this disclosure provides a method for predicting tissue hypoperfusion during or after thrombectomy. The method includes: acquiring digital subtraction angiography (DSA) sequences during or after thrombectomy; wherein the DSA sequence comprises multiple consecutive two-dimensional images; dividing the DSA sequence into multiple target regions based on the structure of the cerebral blood supply system; for each target region, determining corresponding hemodynamic characteristic parameters based on the time-varying pattern of contrast agent concentration; for different target regions, constructing a time difference matrix based on blood flow differences, and determining spatial heterogeneity characteristic parameters based on the time difference matrix; constructing a multidimensional feature vector using the hemodynamic characteristic parameters, the time difference matrix, and the spatial heterogeneity characteristic parameters; inputting the multidimensional feature vector into a tissue hypoperfusion prediction model to obtain the predicted tissue hypoperfusion result output by the tissue hypoperfusion prediction model; wherein the tissue hypoperfusion prediction model is trained using multidimensional feature vector samples from historical patients during or after thrombectomy and the corresponding tissue hypoperfusion occurrence results, and the predicted tissue hypoperfusion result serves as an indicator to characterize the development trend of tissue hypoperfusion. By acquiring and analyzing digital subtraction angiography sequences during or immediately after thrombectomy, this method enables real-time assessment of cerebral hypoperfusion, providing a valuable window for timely intervention. Target regions are segmented according to the cerebral blood supply system structure, enhancing the correlation between characteristic parameters and tissue perfusion status. A multidimensional feature vector is constructed by combining single-region hemodynamic parameters, cross-regional blood flow time difference matrices, and spatial heterogeneity parameters. The model automatically integrates and analyzes these multidimensional feature parameters and determines the likelihood of future hypoperfusion, eliminating subjective errors from manual assessment and improving the accuracy and consistency of hypoperfusion prediction. This application enables early identification of tissue hypoperfusion, guiding intraoperative decision-making and individualized treatment.

[0032] The method will be described below with reference to specific embodiments.

[0033] Figure 1 This is a flowchart illustrating a method for predicting tissue hypoperfusion during or after thrombectomy, provided in an embodiment of this disclosure. This method can be executed by a device for predicting tissue hypoperfusion during or after thrombectomy, which can be implemented using software and / or hardware, and is generally integrated into a computing device. Figure 1 As shown, the method includes: S101. Collect digital subtraction angiography sequences during or after thrombectomy.

[0034] Digital subtraction angiography (DSA) sequences are acquired in real time during endovascular thrombectomy (EVT) or immediately after thrombectomy. The DSA sequence comprises multiple consecutive two-dimensional images.

[0035] Specifically, during or after thrombectomy, a targeted contrast agent is injected via the internal carotid artery. The injection dose is 8 ml, and the injection rate is controlled at 4 ml / s. A uniform injection rate is used to ensure the stability of the intravascular contrast agent concentration and avoid perfusion artifacts caused by rate fluctuations. Furthermore, a fixed image acquisition rate of 4 frames per second (fps) is set, along with standard anteroposterior and lateral projection angles. The optimal projection angle is determined based on the patient's condition, and digital subtraction angiography sequences are acquired.

[0036] Understandably, standard anteroposterior and lateral views can ensure the basic consistency of vascular anatomy and the optimal projection angle can be determined through preoperative scanning or real-time adjustment during surgery, avoiding anatomical overlap problems of the skull, dura mater, and adjacent blood vessels, achieving clear imaging of the target blood vessels without overlap, and eliminating the interference of overlap artifacts on subsequent perfusion analysis.

[0037] Furthermore, the multi-frame two-dimensional images in the digital subtraction angiography sequence are preprocessed.

[0038] Optionally, image stability calibration is performed on multiple frames of two-dimensional images in the digital subtraction angiography sequence. Specifically, based on a registration algorithm of vascular anatomical feature points, spatial position calibration is performed on multiple frames of two-dimensional images to eliminate image shifts caused by slight patient head movements and equipment vibrations during the procedure, ensuring that the spatial position of the same vascular segment is consistent in different time frames.

[0039] Optionally, temporal consistency correction is performed on multiple frames of two-dimensional images in a digital subtraction angiography sequence. Specifically, the timestamps of multiple frames of two-dimensional images are calibrated based on a fixed image acquisition rate to ensure uniform time intervals between consecutive two-dimensional images.

[0040] Optionally, background noise filtering can be performed on multiple frames of two-dimensional images in the digital subtraction angiography sequence. Specifically, algorithms such as Gaussian filtering and / or frequency domain denoising are used to filter various types of noise in the two-dimensional images, thereby improving the image signal-to-noise ratio.

[0041] Optionally, dynamic enhancement normalization processing is performed on multiple frames of two-dimensional images in the digital subtraction angiography sequence. Specifically, the enhancement signals of each frame and each vascular segment in the sequence are normalized using the peak contrast enhancement intensity of the internal carotid artery as a reference benchmark.

[0042] Furthermore, a quantitative perfusion map is generated using specialized image analysis software.

[0043] Specifically, the Atlas dedicated image analysis software is used to quantify the data of the digital subtraction angiography sequence and intuitively convert it into a color visualization image, resulting in a color-coded time-to-peak opacity map.

[0044] S102. Based on the structure of the brain blood supply system, the digital subtraction angiography sequence is divided into multiple target regions.

[0045] Specifically, based on the anatomical structure of the cerebral blood supply system, the vascular blood supply zones, and the hemodynamic characteristics, each frame of the two-dimensional image in the digital subtraction angiography sequence acquired after mechanical thrombectomy vessel recanalization is divided into regions.

[0046] In one possible implementation, arterial and venous regions are divided from a digital subtraction angiography sequence according to the arterial blood supply system of the brain.

[0047] Optionally, a deep learning model can be used to segment the digital subtraction angiography (DSA) sequence. Specifically, the DSA sequence is input into a semantic segmentation model for automatic pixel-level segmentation, outputting the coordinate matrix of each region. For example, the semantic segmentation model can be UNet, FCN, DeepLab series, SegNet, etc. The semantic segmentation model is pre-trained using manually labeled DSA sequence samples. The semantic segmentation model includes a connected input layer, encoder, bottleneck layer, decoder, and output layer, which are used for feature downsampling, feature fusion, feature uptesting, and pixel-level classification of the DSA sequence, respectively.

[0048] For example, the multiple arterial regions in the digital subtraction angiography sequence, according to the arterial blood supply system of the brain, include the initial segment of the internal carotid artery (ICA), the intracranial internal carotid artery segment, the M1 segment of the middle cerebral artery, the M2 segment of the middle cerebral artery, or cortical branch segments, and the venous region includes the superior sagittal sinus segment.

[0049] In this approach, by refining the arterial regions such as the initial segment of the internal carotid artery, the intracranial segment, the middle cerebral artery, and cortical branches, the differences in blood perfusion in different arterial segments can be captured in a targeted manner, providing anatomical anchors for the regional quantitative analysis of parameters and helping to identify low-perfusion sites in the arterial blood supply layer; it also incorporates the superior sagittal sinus segment venous region, taking into account the synergistic assessment of arterial blood supply and venous drainage.

[0050] S103. For each target region, determine the corresponding hemodynamic characteristic parameters based on the change of contrast agent concentration over time.

[0051] The variation of contrast agent concentration over time is directly related to the vascular anatomy and hemodynamic state. Under normal blood flow conditions, it exhibits a typical single-peak continuous curve. However, characteristic distortions in the curve's shape appear when the vessel is stenotic or recanalization is incomplete. Therefore, this application analyzes the hemodynamic characteristic parameters of each of the multiple target regions obtained in step S102, based on the variation of contrast agent concentration over time during blood flow, to characterize the perfusion status of the target region. The following explanation uses one of the multiple target regions as an example. The method for determining the hemodynamic characteristic parameters of each target region is the same.

[0052] In one possible implementation, step S103 includes: S1031. Obtain the pixel density value of each two-dimensional image in the digital subtraction angiography sequence.

[0053] Specifically, for each frame of a two-dimensional image in a digital subtraction angiography sequence, the gray value of each pixel is recorded as I (x, y, t), which represents the contrast agent concentration at that location at that moment.

[0054] Where (x, y) represents pixel coordinates, t represents the time frame sequence, t=1, 2,……,T. For multi-frame two-dimensional images, the time frame corresponding to each frame of the two-dimensional image is different.

[0055] S1032. Analyze the variation of pixel density values ​​over time in multiple consecutive two-dimensional images and construct a time-density curve.

[0056] For a target region, the construction time-density curve is as follows:

[0057] In the formula, The time-density curve, This represents the total number of pixels in the target area. The grayscale values ​​of each pixel within the target area are averaged using the formula described above to eliminate single-pixel noise.

[0058] S1033. Calculate the time of peak concentration of contrast agent based on the time-density curve to obtain the peak time.

[0059] The time to peak (TTP) calculated based on the time-density curve above is as follows:

[0060] S1034. Multiple hemodynamic characteristic parameters are determined based on time-density curves and peak time.

[0061] In one possible implementation, hemodynamic characteristic parameters include relative time to peak (rTTP). Specifically, the relative time to peak is determined based on the difference between the time to peak and the time to peak in the reference region.

[0062] For example, the reference region is the initial segment of the internal carotid artery, and the relative time to peak is:

[0063] In the formula, This refers to the time to peak concentration at the origin of the internal carotid artery.

[0064] In one possible implementation, hemodynamic characteristic parameters include Mean Transit Time (MTT). Specifically, the mean transit time is determined based on the ratio of the area under the time-density curve to the peak value:

[0065] In the formula, The arrival time of the contrast agent. This is the point at which the contrast agent elution ends.

[0066] In one possible implementation, hemodynamic characteristic parameters include cerebral circulation time (CCT). Specifically, the cerebral circulation time for each arterial region is determined based on the difference in peak time between the arterial region and the reference region:

[0067] In the formula, The peak time for blood flow through the kth cortical branch of the cerebral cortex is represented by the above formula, which describes the time difference of blood flowing from the internal carotid artery through the cerebral cortex.

[0068] In one possible implementation, hemodynamic characteristic parameters include microvascular transit time (mTT). Specifically, the microvascular transit time is determined based on the difference in peak time between the arterial and venous regions:

[0069] In the formula, This refers to the peak time of the superior sagittal sinus segment in the venous region.

[0070] Furthermore, after calculating hemodynamic characteristic parameters for multiple target regions of the digital subtraction angiography sequence, the hemodynamic characteristic parameters of each region are combined to generate a visualization of the digital subtraction angiography sequence.

[0071] In one possible implementation, the visualization includes a parameter heat map, which is a pseudo-color heat map.

[0072] Specifically, the parameter scalar field P(x, y) generated for each pixel in the image is as follows:

[0073] The parameter scalar field is normalized to eliminate individual baseline differences, resulting in:

[0074] In the formula, The parameter mean of the reference area, This represents the standard deviation of the reference area.

[0075] Numerical values ​​are converted to the RGB color space using linear or nonlinear color mapping functions to generate a visual heatmap, thus obtaining a parameter heatmap. For example, the color mapping function is:

[0076] In one possible implementation, the visualization includes a Local Perfusion Timing Map.

[0077] Specifically, a binarized or hierarchical perfusion delay map is constructed using the relative peak times of each region.

[0078] For example, the local perfusion time profile is as follows:

[0079] In the formula, when When the value is 1, it represents a low-perfusion region. When the value is 0, it is in the normal range.

[0080] Among them, hemodynamic parameters are transformed into visual graphs such as parametric heatmaps and local perfusion time maps, and reports are generated by combining the low perfusion prediction results, which is both intuitive and practical.

[0081] Furthermore, in one possible implementation, a spatially distributed structured parameter matrix is ​​constructed based on the multiple hemodynamic feature parameters obtained above. This spatially distributed structured parameter matrix quantifies the non-uniformity of brain tissue perfusion in space and characterizes the distribution characteristics and interrelationships of hemodynamic feature parameters in the spatial dimension of brain tissue.

[0082] Specifically, the relative peak time, average transit time, cerebral circulation time, and microvascular transit time of each target region are obtained. Using the target region as the spatial index dimension and multiple hemodynamic characteristic parameters as numerical elements, a two-dimensional or multi-dimensional matrix is ​​constructed to obtain:

[0083] By introducing spatial weights based on the probability of blood supply to blood vessels, a weighted matrix is ​​generated:

[0084] The weighted spatial distribution structured parameter matrix can enhance the focus on key blood supply areas. Specifically, the spatial distribution structured parameter matrix can characterize tissue perfusion anomalies at the overall spatial structure level, avoiding misjudgments caused by judging perfusion status based on a single region or single parameter, and improving the accuracy of low perfusion identification.

[0085] In this implementation, by extracting pixel density values ​​frame by frame and constructing time-density curves by combining continuous frame temporal analysis, data support can be provided for parameter calculation. Based on the peak time, multi-dimensional hemodynamic parameters are derived, which can comprehensively reflect the perfusion efficiency, blood flow velocity and microcirculation status of different target areas, providing a quantitative basis for subsequent assessment of brain tissue perfusion status, eliminating subjective errors of manual assessment, and improving the accuracy and consistency of low perfusion prediction.

[0086] Furthermore, it can comprehensively cover core dimensions such as perfusion rate, microcirculation function, and blood flow transmission delay. The parameters are logically complementary and progressive, adapting to the quantitative analysis needs of image sequences. It can link different physiological stages of brain tissue perfusion, providing objective quantitative evidence for distinguishing ineffective reperfusion and identifying low perfusion types after surgery. It solves the problems of traditional single parameters, lack of microcirculation assessment, and poor comparability, and achieves standardized intraoperative and postoperative real-time perfusion assessment.

[0087] S104. For different target regions, construct a time difference matrix based on blood flow differences, and determine spatial heterogeneity characteristic parameters based on the time difference matrix.

[0088] Traverse all target regions and calculate the target region. and target area The difference in peak time yields multiple blood flow time differences:

[0089] The time-delay matrix is ​​constructed based on multiple blood flow time differences as follows:

[0090] The time difference matrix, being an antisymmetric matrix, characterizes the lag relationship in blood flow between different brain regions and reflects abnormalities in the conduction sequence of cerebral hemodynamics. Understandably, it allows for the identification of blood flow delay regions from multiple target areas based on the time difference matrix.

[0091] In this implementation, by traversing all target regions, calculating the peak time difference between each pair and constructing a matrix, the differences in blood perfusion timing between regions can be systematically and quantitatively characterized, and the correlation patterns and abnormal characteristics of blood perfusion between regions can be effectively explored.

[0092] Furthermore, spatial heterogeneity characteristic parameters are determined based on the time difference matrix.

[0093] In one possible implementation, the spatial heterogeneity characteristic parameter includes a delay volume fraction. Specifically, the delay volume fraction is determined based on the proportion of the blood flow delay region:

[0094] In the formula, The total pixel area of ​​the delay region. This refers to the total area of ​​the blood supply zone to the cerebral hemispheres.

[0095] In one possible implementation, the spatial heterogeneity characteristic parameter includes spatial dispersion. Specifically, the spatial dispersion is determined based on the mean and standard deviation of the peak time in the blood flow delay region:

[0096] In the formula, This represents the average time to peak flow in the delayed blood flow region. The standard deviation of the peak time in the blood flow delay region, and the spatial dispersion. The coefficient of variation describes the non-uniformity of the perfusion distribution.

[0097] In one possible implementation, the spatial heterogeneity characteristic parameter includes spatial entropy. Here, spatial entropy represents the degree of disorder in the hemodynamic characteristic parameters of the blood flow delay region.

[0098] Optionally, one hemodynamic characteristic parameter can be selected to calculate the spatial entropy, or multiple parameters can be selected to calculate the corresponding spatial entropy separately.

[0099] For example, let's take selecting a hemodynamic characteristic parameter as an example. The hemodynamic characteristic parameter is divided into multiple preset intervals according to its value range, and corresponding parameter histograms are constructed. The frequency of each parameter value is counted, and the probability distribution after frequency normalization is calculated. The probabilities corresponding to each interval are denoted as follows: The spatial entropy is calculated based on the probability distribution as follows:

[0100] Similarly, the spatial entropy of each hemodynamic characteristic parameter is calculated using the method described above.

[0101] In this implementation, multiple parameters are derived based on the time difference matrix. The features have strong correlations and high data consistency, which can be directly used as effective inputs for subsequent prediction models. This provides key spatial dimension features to deeply explore the spatial patterns of perfusion anomalies and improve the accuracy of low perfusion prediction.

[0102] S105. Construct a multidimensional feature vector using hemodynamic feature parameters, time difference matrix, and spatial heterogeneity feature parameters. Input the multidimensional feature vector into the tissue hypoperfusion prediction model to obtain the prediction results of tissue hypoperfusion in patients output by the tissue hypoperfusion prediction model.

[0103] In one possible implementation, a multidimensional feature vector is constructed using hemodynamic feature parameters, time difference matrix, and spatial heterogeneity feature parameters.

[0104] Specifically, the multidimensional feature vector is represented as:

[0105] in, These represent the values ​​corresponding to multiple target regions. The maximum value of the multi-time difference matrix. It is the mean of the multi-time difference matrix.

[0106] In another possible implementation, a spatially distributed structured parameter matrix is ​​further introduced into the multidimensional feature vector.

[0107] In another possible implementation, an eTICI score is further introduced into the multidimensional feature vector.

[0108] Specifically, the vascular recanalization status after thrombectomy is hierarchically labeled based on digital subtraction angiography sequences. For example, an eTICI score of 3 is defined as complete vascular recanalization, and an eTICI score of 2b–2c is defined as good recanalization. The labeling results are quantified, and multidimensional feature vectors are introduced for model prediction. The eTICI score, as an auxiliary feature of global vascular recanalization, complements local tissue perfusion parameters, thereby improving the accuracy and stability of predicting post-thrombectomy tissue hypoperfusion.

[0109] Furthermore, the multidimensional feature vector is input into the pre-trained tissue hypoperfusion prediction model to obtain the prediction results of the patient's tissue hypoperfusion. The prediction results of tissue hypoperfusion serve as indicator information characterizing the development trend of tissue hypoperfusion.

[0110] Optionally, the predicted result of tissue hypoperfusion is the perfusion risk value of tissue hypoperfusion, where the perfusion risk value is 0-1.

[0111] The tissue hypoperfusion prediction model is trained using multidimensional feature vector samples from historical patients during or after thrombectomy and the corresponding tissue hypoperfusion occurrence results.

[0112] For example, the prediction model for low perfusion is a logistic regression model. The logistic regression model includes an input layer, a feature preprocessing layer, a logistic regression layer, and an output layer.

[0113] The training set data is input into the logistic regression model. The input layer receives multi-dimensional feature vector samples, and the feature preprocessing layer standardizes the multi-dimensional feature vector samples.

[0114] For example, z-score standardization is applied to multidimensional feature vector samples to eliminate differences in feature dimensions and improve the convergence speed and prediction stability of the logistic regression model.

[0115] Furthermore, a multi-output logistic regression is employed in the logistic regression layer. The output of the linear regression is mapped to the [0,1] interval using the sigmoid function. For example, a three-branch output logistic regression is used in the logistic regression layer, outputting the probability value of low perfusion, the low perfusion risk level, and the probability distribution corresponding to the level, respectively. The three branches share the weight matrix of the underlying multidimensional features.

[0116] Furthermore, the above prediction results are output at the output layer.

[0117] The steps for training a logistic regression model include: A dataset with a pre-built logistic regression model is used. Multidimensional feature vector samples from historical patients during or after embolization are collected as feature data, and the corresponding postoperative tissue hypoperfusion results are collected as feature labels. The dataset is constructed by combining the feature data and feature labels.

[0118] In one possible implementation, the degree of vascular recanalization in historical patients is assessed and classified according to the eTICI (eTraining-Induced Perfusion Index). The eTICI score is then input into a logistic regression model for training. Here, the eTICI score serves as a structural indicator at the macroscopic level of recanalization, a sample group label to distinguish between complete and incomplete recanalization cases, a covariate in model construction, and is jointly input into the hypoperfusion prediction model along with tissue-level perfusion indicators based on quantitative DSA parameters. It also serves as a stratified reference in the model output interpretation stage to distinguish patient subtypes with adequate vascular recanalization but still exhibiting tissue-level hypoperfusion.

[0119] Divide the dataset into a training set and a validation set. For example, use 80% of the dataset as the training set and 20% as the validation set.

[0120] With the goal of minimizing the binary cross-entropy loss, we optimize the core hyperparameters of a logistic regression model using 5-fold cross-validation. Specifically, the training set is randomly divided into 5 mutually exclusive subsets of equal sample size. One subset is used as the validation set, and the remaining 4 subsets are used as the training set. The model is trained, and the validation set performance is calculated. This process is repeated 5 times to obtain 5 sets of validation set performance results. The average value is taken as the final performance of the hyperparameter combination. All hyperparameter combinations to be optimized are traversed, and the hyperparameter combination with the best average performance is selected.

[0121] The logistic regression model is trained using gradient descent. Specifically, the weights and biases are updated using gradient descent until the model converges, at which point training stops and is validated using a validation set.

[0122] Finally, the system combines the predicted results of tissue hypoperfusion in patients with visualizations to generate and present a tissue hypoperfusion report. Specifically, the system automatically generates a brain tissue-level hypoperfusion risk, probability heatmap, and spatial consistency assessment. Wedge-shaped hypoperfusion areas are highlighted if they exist. A structured report is generated, including DSA parameter values, predicted hypoperfusion probability, thresholds, and visualizations (in color). Output formats include DICOM-SR and PDF.

[0123] In the implementation of this application, the correlation between DSA parameters and postoperative perfusion is studied, a threshold model is established, and reperfusion status is objectively distinguished. This overcomes the limitations of traditional delayed assessment, enabling real-time judgment of cerebral blood flow perfusion status after thrombectomy and avoiding operator variability. Simultaneously, it improves upon the deficiencies of existing reperfusion assessments, addressing the qualitative limitations of eTICI, predicting neurological functional prognosis, and enhancing prediction accuracy, sensitivity, and specificity. This helps in the early identification of hypoperfusion, guiding targeted interventions, ultimately reducing the proportion of adverse outcomes and improving clinical outcomes. The assessment system is transformed from assessing cerebral blood structure to assessing cerebrovascular function, demonstrating greater potential for clinical translation.

[0124] To achieve the above embodiments, this disclosure also proposes a device for predicting tissue hypoperfusion during or after thrombectomy.

[0125] Figure 2 This is a schematic diagram of a device for predicting tissue hypoperfusion during or after thrombectomy, provided in an embodiment of this disclosure. This device can be implemented by software and / or hardware, and is generally integrated into a computing device. Figure 2 As shown, the device for predicting tissue hypoperfusion during or after thrombectomy includes: The acquisition module 201 is used to acquire digital subtraction angiography sequences during or after thrombectomy; wherein the digital subtraction angiography sequence includes multiple consecutive two-dimensional images.

[0126] The segmentation module 202 is used to segment the digital subtraction angiography sequence into multiple target regions based on the structure of the brain blood supply system.

[0127] The first analysis module 203 is used to determine the corresponding hemodynamic characteristic parameters for each target region based on the change of contrast agent concentration over time.

[0128] The second analysis module 204 is used to construct a time difference matrix based on blood flow differences for different target regions, and to determine spatial heterogeneity characteristic parameters based on the time difference matrix.

[0129] The prediction module 205 is used to construct a multidimensional feature vector using hemodynamic feature parameters, time difference matrix and spatial heterogeneity feature parameters. The multidimensional feature vector is input into the tissue hypoperfusion prediction model to obtain the prediction result of the patient's tissue hypoperfusion. The tissue hypoperfusion prediction model is trained using multidimensional feature vector samples and tissue hypoperfusion occurrence results from historical patients during or after thrombectomy. The prediction result of tissue hypoperfusion is an indicator information used to characterize the development trend of tissue hypoperfusion.

[0130] In one possible implementation, the partitioning module 202 includes: The first segmentation submodule is used to segment multiple arterial regions from the digital subtraction angiography sequence according to the arterial blood supply system of the brain. The arterial regions include the initial segment of the internal carotid artery, the intracranial segment of the internal carotid artery, the M1 segment of the middle cerebral artery, the M2 segment of the middle cerebral artery, or cortical branch segments.

[0131] The second segmentation submodule is used to segment the venous region from the digital subtraction angiography sequence according to the venous drainage system of the brain's blood supply system. The venous region includes the superior sagittal sinus segment.

[0132] In one possible implementation, the first analysis module 203 includes: The acquisition submodule is used to acquire the pixel density value of each two-dimensional image in the digital subtraction angiography sequence.

[0133] The analysis submodule is used to analyze the variation of pixel density values ​​over time in multiple consecutive frames of two-dimensional images, and to construct a time-density curve.

[0134] The calculation submodule is used to calculate the moment when the contrast agent reaches its maximum concentration based on the time-density curve, thus obtaining the peak time.

[0135] The determination submodule is used to determine multiple hemodynamic characteristic parameters based on the time-density curve and time to peak.

[0136] In one possible implementation, determining the submodules includes: The first determining unit is used to determine the relative peak time based on the difference between the peak time and the peak time of the reference area.

[0137] The second determining unit is used to determine the average transit time based on the ratio of the area to the peak value of the time-density curve.

[0138] The third determining unit is used to determine the cerebral circulation time of each arterial region based on the difference between the peak time of the arterial region and the reference region.

[0139] The fourth determining unit is used to determine the microvascular transit time based on the difference in peak time between the arterial region and the venous region.

[0140] In one possible implementation, the second analysis module 204 includes: The traversal submodule is used to traverse all target regions, calculate the difference in peak time between every two target regions, and obtain multiple blood flow time differences.

[0141] A submodule is constructed to build an antisymmetric time difference matrix based on multiple blood flow time differences.

[0142] In one possible implementation, the second analysis module 204 includes: The fifth determining unit is used to determine the blood flow delay region from multiple target regions based on the time difference matrix.

[0143] The sixth determining unit is used to determine the delay volume fraction based on the proportion of the blood flow delay region.

[0144] The seventh determining unit is used to determine the spatial dispersion based on the mean and standard deviation of the peak time in the blood flow delay region.

[0145] The eighth determining unit is used to determine the spatial entropy based on the degree of disorder of hemodynamic characteristic parameters in the blood flow delay region.

[0146] In one possible implementation, the device further includes: The visualization module is used to generate visualization maps based on hemodynamic characteristic parameters. The visualization maps include parameter heatmaps and local perfusion time maps. The module combines the predicted results of tissue hypoperfusion in patients with the visualization maps to generate and present a tissue hypoperfusion report.

[0147] The device for predicting tissue hypoperfusion during or after thrombectomy provided in this disclosure can execute the method for predicting tissue hypoperfusion during or after thrombectomy provided in any embodiment of this disclosure, and has the corresponding functional modules and beneficial effects for executing the method.

[0148] To implement the above embodiments, this disclosure also proposes a computer program product, including a computer program / instruction, which, when executed by a processor, implements the method for predicting tissue hypoperfusion during or after thrombectomy as described in the above embodiments.

[0149] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of the present disclosure.

[0150] The following is a detailed reference. Figure 3 The diagram illustrates a structural schematic suitable for implementing the computing device 300 in the embodiments of this disclosure. The computing device 300 in the embodiments of this disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3 The computing device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0151] like Figure 3As shown, the computing device 300 may include a processor (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 302 or a program loaded from memory 308 into random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the computing device 300. The processor 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0152] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows computing device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 A computing device 300 with various devices is shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0153] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from memory 308, or installed from ROM 302. When the computer program is executed by processor 301, it performs the functions defined in the method for predicting tissue hypoperfusion during or after thrombectomy according to embodiments of this disclosure.

[0154] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0155] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0156] The aforementioned computer-readable medium may be included in the aforementioned computing device; or it may exist independently and not assembled into the computing device.

[0157] The aforementioned computer-readable medium carries one or more programs that, when executed by the computing device, cause the computing device to perform the aforementioned method for predicting tissue hypoperfusion during or after thrombectomy.

[0158] The computing device can be programmed with computer program code in one or more programming languages ​​or a combination thereof to perform the operations of this disclosure. These programming languages ​​include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0159] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0160] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0161] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0162] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0163] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this disclosure.

[0164] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0165] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for predicting tissue hypoperfusion during or after thrombectomy, characterized in that, The method includes: Digital subtraction angiography sequences were acquired during or after thrombectomy in the patient; wherein the digital subtraction angiography sequences included multiple consecutive two-dimensional images; Based on the structure of the brain blood supply system, the digital subtraction angiography sequence is divided into multiple target regions; For each target region, the corresponding hemodynamic characteristic parameters are determined based on the change of contrast agent concentration over time. For different target regions, a time difference matrix is ​​constructed based on blood flow differences, and spatial heterogeneity characteristic parameters are determined based on the time difference matrix. A multidimensional feature vector is constructed using the hemodynamic feature parameters, the time difference matrix, and the spatial heterogeneity feature parameters. This multidimensional feature vector is then input into a tissue hypoperfusion prediction model to obtain the predicted tissue hypoperfusion outcome for the patient. The tissue hypoperfusion prediction model is trained using multidimensional feature vector samples from historical patients during or after thrombectomy and the corresponding tissue hypoperfusion outcomes. The predicted tissue hypoperfusion outcome serves as an indicator to characterize the development trend of tissue hypoperfusion.

2. The method for predicting tissue hypoperfusion during or after thrombectomy according to claim 1, characterized in that, The structure based on the brain blood supply system divides the digital subtraction angiography sequence into multiple target regions, including: According to the arterial blood supply system of the brain blood supply system, the digital subtraction angiography sequence is divided into multiple arterial regions, including the origin of the internal carotid artery, the intracranial internal carotid artery segment, the M1 segment of the middle cerebral artery, the M2 segment of the middle cerebral artery, or cortical branch segments. The venous region is divided from the digital subtraction angiography sequence according to the venous drainage system of the brain blood supply system, and the venous region includes the superior sagittal sinus segment.

3. The method for predicting tissue hypoperfusion during or after thrombectomy according to claim 2, characterized in that, The determination of corresponding hemodynamic characteristic parameters based on the change of contrast agent concentration over time includes: Obtain the pixel density value of each two-dimensional image in the digital subtraction angiography sequence; The pixel density values ​​of multiple consecutive two-dimensional images are analyzed over time to construct a time-density curve. The time of peak concentration of contrast agent is calculated based on the time-density curve to obtain the peak time. Based on the time-density curve and the time to peak, several hemodynamic characteristic parameters are determined.

4. The method for predicting tissue hypoperfusion during or after thrombectomy according to claim 3, characterized in that, The determination of multiple hemodynamic characteristic parameters based on the time-density curve and the time to peak flow includes: The relative peak time is determined based on the difference between the peak time and the peak time of the reference area. The average transit time is determined based on the ratio of the area under the time-density curve to the peak value. The cerebral circulation time of each arterial region is determined based on the difference between the peak time of the arterial region and the reference region. The microvascular transit time is determined based on the difference between the peak time of the arterial region and the venous region. The reference region is the initial segment of the internal carotid artery.

5. The method for predicting tissue hypoperfusion during or after thrombectomy according to claim 3, characterized in that, The construction of a time difference matrix based on blood flow differences for different target regions includes: Traverse all the target regions and calculate the difference in peak time between every two target regions to obtain multiple blood flow time differences; An antisymmetric time difference matrix is ​​constructed based on multiple blood flow time differences.

6. The method for predicting tissue hypoperfusion during or after thrombectomy according to claim 3 or 5, characterized in that, The determination of spatial heterogeneity characteristic parameters based on the time difference matrix includes: Blood flow delay regions are determined from multiple target regions based on the time difference matrix; The delay volume fraction is determined based on the proportion of the blood flow delay region; Spatial dispersion is determined based on the mean and standard deviation of the peak time in the blood flow delay region; Spatial entropy is determined based on the degree of disorder of the hemodynamic characteristic parameters in the blood flow delay region.

7. The method for predicting tissue hypoperfusion during or after thrombectomy according to claim 1, characterized in that, The method further includes: A visualization is generated based on the hemodynamic characteristic parameters; wherein the visualization includes a parametric heatmap and a local perfusion time map; The predicted results of tissue hypoperfusion in patients and the visualization are combined to generate and present a tissue hypoperfusion report.

8. A device for predicting tissue hypoperfusion during or after thrombectomy, characterized in that, The device includes: The acquisition module is used to acquire digital subtraction angiography sequences during or after thrombectomy; wherein the digital subtraction angiography sequence includes multiple consecutive two-dimensional images; A segmentation module is used to divide the digital subtraction angiography sequence into multiple target regions based on the structure of the brain blood supply system; The first analysis module is used to determine the corresponding hemodynamic characteristic parameters for each target region based on the change of contrast agent concentration over time. The second analysis module is used to construct a time difference matrix based on blood flow differences for different target regions, and to determine spatial heterogeneity characteristic parameters based on the time difference matrix. The prediction module is used to construct a multidimensional feature vector using the hemodynamic feature parameters, the time difference matrix, and the spatial heterogeneity feature parameters. The multidimensional feature vector is then input into a tissue hypoperfusion prediction model to obtain the prediction result of tissue hypoperfusion in the patient output by the tissue hypoperfusion prediction model. The tissue hypoperfusion prediction model is trained using multidimensional feature vector samples from historical patients during or after thrombectomy and the occurrence of tissue hypoperfusion. The prediction result of tissue hypoperfusion is an indicator used to characterize the development trend of tissue hypoperfusion.

9. A computing device, characterized in that, The computing device includes: a processor; a memory for storing executable instructions of the processor; the processor for reading the executable instructions from the memory and executing the instructions to implement the method for predicting tissue hypoperfusion during or after thrombectomy as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method for predicting tissue hypoperfusion during or after thrombectomy as described in any one of claims 1 to 7.

Images