Identifying vascular occlusions using spatial patterns

Abstract

Images of an individual obtained using perfusion-based or diffusion-based imaging techniques can be analyzed to identify areas of the individual's brain that have had their blood supply disrupted. The images can be used to generate an alert indicating disruption of blood flow to one or more areas of the individual's brain. The images can be used to identify vascular segments (e.g., M1, M2, M3, M4, ...) and branches (MCA, ACA, PCA) in the individual's brain in which abnormalities may exist.

Description

[Background technology]

[0001] When blood flow to a tissue is interrupted, damage to the tissue in the human body can occur. Reduced blood flow to a tissue can have several causes, such as vascular occlusion, vascular rupture, vascular narrowing, or vascular compression. The severity of damage to the tissue can depend on the degree of blood flow interruption to the tissue and the length of time blood flow to the tissue is interrupted. In situations where the blood supply to a tissue is insufficient for an extended period of time, the tissue can become infarcted. Infarcted tissue can result from the death of cells in the tissue due to lack of blood supply. [Brief explanation of the drawings]

[0002] In the drawings, which are not necessarily drawn to scale, like numerals may describe like components in different figures. To easily identify any particular element or description of an operation, the most significant digit(s) in a reference number refers to the figure number in which that element is first introduced. Some implementations are shown by way of example, and not by way of limitation. [Figure 1] FIG. 1 is a schematic diagram of a computational architecture for identifying blood vessels in the brain where abnormalities exist, according to one or more exemplary implementations. [Figure 2] 2 illustrates a computational architecture 200 for training and implementing one or more computational models for analyzing spatial patterns obtained from image data, according to one or more implementations. [Figure 3] FIG. 1 is a flow diagram of a process for identifying blood vessels in the brain where abnormalities exist, according to one or more exemplary implementations. [Figure 4] FIG. 1 is a block diagram illustrating components of a machine in the form of a computer system that can read and execute instructions from one or more machine-readable media to perform any one or more of the methods described herein, according to one or more exemplary implementations. [Figure 5] FIG. 1 is a block diagram illustrating a representative software architecture that can be used in conjunction with one or more hardware architectures described herein, according to one or more exemplary implementations. [Figure 6] Included is a first table containing patient demographics and location of vascular occlusion for a retrospective study of imaging data. [Figure 7] A second table is included containing the diagnostic performance of the readers for detecting DVO on CTA with and without Tmax. [Figure 8] A third table is included containing the diagnostic performance of the readers for detecting DVO on CTA with and without Tmax for n=118 (48 with DVO, 70 without DVO), excluding 22 patients with proximal M2-MCA occlusion. [Figure 9] Includes the first image A showing a selected slice of a normal Tmax map with no delayed regions in a patient without DVO, the second image B showing a Tmax map (selected slice) showing significant delayed regions in the left distal ACA territory, and the third image C showing an axial CTA MIP showing a culprit A4 segment ACA occlusion (red arrow) in this 72-year-old woman who presented with right leg weakness. [Figure 10] Boxplots of the time taken to interpret CTA images in patients with and without DVO are shown relative to the reference standard. The dashed red line indicates the median time, the top and bottom of the box indicate the first and third quartiles, respectively, the notches represent the 95% confidence interval of the median, and the whiskers extend to the fifth and ninety-fifth percentiles. Outliers are indicated by circles. For both readers, interpretation was significantly faster with Tmax than without Tmax (p<0.001). The spread in time also decreased with Tmax. [Figure 11]Examples of how Tmax aided in the detection of DVO on CTA include: In image A, a Tmax delay >10 seconds is evident in the left MCA superior M2 division region of a 54-year-old woman who presented with sudden aphasia onset. In image B, a sagittal CTA MIP (selected slice) shows an occlusion (red arrow), which was detected by only two readers without Tmax but by all four readers with Tmax. In image C, a wedge-shaped Tmax delay was seen in the right parietal lobe of an 83-year-old man. In image D, a right parietal M4 occlusion (red arrow) shown on an axial CTA MIP was detected by only two readers without Tmax but by all four readers with Tmax. [Figure 12] A patient selection flowchart is shown. At our institution, all patients who presented with suspected acute ischemic stroke and were within the ECR time frame (and otherwise met clinical eligibility criteria) underwent urgent multimodal CT with unenhanced CT, CTA, and CTP as standard of care. *CTA did not extend to the vertex, and therefore the anterior cerebral artery A4 and A5 segments were not covered. [Figure 13] Figure 1 shows inter-reader agreement for the presence of DVO on CTA with and without Tmax, between pairs of readers, determined using Cohen's k statistic. Without Tmax, inter-reader agreement was either moderate or substantial. Agreement increased with the use of Tmax to either substantial or near-perfect agreement for all pairs of readers. [Figure 14] Diagnostic confidence shift analysis including images A-D is shown. When the analysis was limited to patients with DVO (vs. the reference standard), there was a positive shift in confidence for all readers. More patients were identified as "very likely" to have obstruction. E-H. For patients without DVO (vs. the reference standard), confidence in the absence of DVO increased. More patients were deemed "very unlikely" to have obstruction. [Figure 15]These images show Tmax delays not attributable to DVO, leading residents to false positives. In image A, a Tmax delay >6 seconds is seen in the left distal ACA territory (arrow) on this selected slice from the Tmax map of an 83-year-old woman who presented with sudden onset of right-sided weakness. In image B, the CTA shows venous structures consistent with a developmental venous anomaly (arrow) in the corresponding territory. No distal ACA occlusion was evident. In image C, the relative cerebral blood volume map shows a blood pool (arrow) within the venous anomaly. An experienced radiologist recognized the cause of the Tmax delay and dismissed DVO. In image D, left-hemisphere Tmax delays were more pronounced in the external basin in a 70-year-old man who presented with transient right-sided weakness and speech impairment. The patient developed headache, but follow-up imaging revealed no infarction. This symptom was attributed to hemiplegic migraine. In E, a "borderline" Tmax delay is a pattern that includes deep white matter basins (red arrows) as well as outer basins (MCA-PCA basins indicated by white arrows). This results from contrast bolus dispersion, which causes delayed arrival of contrast in the most distal arterial territories. Causes of bolus dispersion include proximal (e.g., internal carotid) arterial stenosis or occlusion, inadequate cardiac output, and inadequate contrast bolus injection. [Figure 16] A table containing information about a patient with DVO at two sites is shown. [Figure 17] A table containing ROC analysis of diagnostic performance for detecting DVO on CTA for n=140 (70 with DVO, 70 without DVO) is shown. [Figure 18] A table containing ROC analysis of the diagnostic performance for detecting DVO on CTA for n=118 (48 with DVO, 70 without DVO), excluding patients with proximal M2-MCA occlusion, is shown. The prevalence of DVO in the screened population was 48 / 501 (10%). [Figure 19] A table containing ROC analysis of the diagnostic performance for detecting DVO on CTA for n=97 (27 with DVO, 70 without DVO), excluding patients with M2-MCA and P2-PCA occlusion. The prevalence of DVO in the screened population was 27 / 501 (5%). [Figure 20] Tables containing shift analysis of change in diagnostic confidence with the addition of Tmax to CTA using Wilcoxon signed rank test are shown. [Figure 21] A table containing the time (in seconds) taken to interpret the CTA with and without Tmax is shown. [Figure 22] A table containing the time (in seconds) taken to read CTA with and without Tmax for M2 (n=38) versus M3 and M4 segment (n=13) MCA occlusion is shown. [Figure 23] 1 shows a table containing false negatives for DVO on CTA with and without Tmax. [Figure 24] 1 shows a table containing false positives for DVO on CTA with and without Tmax. DETAILED DESCRIPTION OF THE INVENTION

[0003] Accurately identifying the location in the brain where the disruption of blood supply has occurred, specifically the arterial segment / branch where the vessel is occluded, is important in determining the treatment to be administered and the treatment outcome and prognosis. In some cases, the type of treatment used to restore blood supply to an area of ​​the brain may depend on the specific location of the cause of the disruption of blood flow to the tissue. In addition, the speed at which a diagnosis of the presence or absence of an occlusion and which vessel segment / branch is involved can also affect the type of treatment and the effectiveness of the treatment provided to an individual.

[0004] In situations where blood flow is disrupted due to distal vascular occlusion, such as by blockage or stenosis, accuracy in identifying the occlusion may be reduced. As used herein, distal vascular occlusion may refer to an occlusion of a vessel distal to at least one of the internal carotid artery (ICA), the M1 segment of the middle cerebral artery (MCA), or the vertebral and basilar arteries. Distal vascular occlusions may include the A2-A5 segments of the anterior cerebral artery, the M2-M4 segments of the middle cerebral artery, the P2-P4 segments of the posterior cerebral artery, the posterior inferior cerebellar artery, the anterior inferior cerebellar artery, and the superior cerebellar artery. Existing techniques for identifying distal vessels in which abnormalities may exist typically rely on data obtained from computed tomography angiography imaging techniques. However, abnormalities in distal vessels may be difficult to identify using data obtained from computed tomography angiography imaging techniques.

[0005] In various examples, the reduced accuracy characteristic of existing techniques may be the result of distal vessels having smaller diameters than proximal vessels in the brain. In addition, distal vessels are more numerous and have reduced opacity on computed tomography angiography images than proximal vessels. Furthermore, distal vessels may have greater branch diversity than proximal vessels. As a result, the reduced accuracy of identifying distal vascular abnormalities and the increased time required to identify distal vascular occlusions may delay or reduce the effectiveness of treatment provided to an individual and increase damage to the individual's brain.

[0006] In one or more implementations, image data generated by at least one of one or more perfusion-based imaging techniques or one or more diffusion-based imaging techniques can be used to identify regions of an individual's brain where vascular abnormalities may be present. The perfusion-based imaging techniques can include CT-based perfusion imaging techniques. Additionally, the perfusion-based imaging techniques can include magnetic resonance (MR)-based imaging techniques. The diffusion-based imaging techniques can include one or more additional MR-based imaging techniques.

[0007] In various embodiments, values ​​of one or more perfusion parameters may be determined. The values ​​of the one or more perfusion parameters may indicate the presence of vascular abnormalities in a region of the individual's brain. For example, the perfusion parameters may indicate a distortion in the arrival time of blood to a region of the brain. In various embodiments, the perfusion parameters may be determined based on a maximum arrival time of a tissue residue function (T) for a voxel of an image captured using one or more perfusion-based imaging techniques. max ) can be included. T max The value of is determined by one or more thresholds T max The T value can be analyzed to identify one or more areas of the brain that may be deprived of blood supply. For example, the T value of an area of ​​the brain of an individual can be analyzed to identify one or more areas of the brain that may be deprived of blood supply. max In situations where the value is at least the threshold, there may be a disruption of blood supply to the area. max The threshold value can indicate the amount of delay in the delivery of contrast agent to a brain region during a perfusion-based imaging process. Additionally, a mean tracer transit time (MTT) value can be determined. The MTT can correspond to the mean transit time of contrast agent through a brain region.

[0008] The values ​​of one or more perfusion parameters can be used to render a user interface showing values ​​of one or more perfusion parameters throughout the individual's brain. In one or more scenarios, the time that elapses before the contrast agent reaches a portion of the brain can be indicated by several different colors. In one or more examples, regions of the individual's brain where the arrival time of the contrast agent meets or exceeds a threshold time can be displayed as one or more colors that contrast with portions of the individual's brain where the arrival time of the contrast agent is less than the threshold time. In this way, a practitioner viewing the user interface can identify regions of the individual's brain that may be experiencing disruption in their blood supply. In various examples, regions of the individual's brain that have values ​​of one or more perfusion parameters that meet or exceed a threshold can be highlighted in the user interface using geometric shapes, arrows, or other indicators.

[0009] In at least some cases, T max Perfusion parameters such as ATP and MTT can provide a more accurate depiction of regions exhibiting delayed blood arrival / transit than other parameters due to the level of uniformity of perfusion parameter values ​​in the brain of an individual without an abnormality, which in turn is an indicator of hemodynamic impairment and surrogate loss of blood flow to that region of the brain. Thus, an image showing values ​​of one or more perfusion parameters throughout the brain of an individual with an abnormality can have a certain amount of contrast relative to an image of the brain of an individual without an abnormality. As a result of the contrast between areas without an abnormality and areas where an abnormality is present, areas where an abnormality is present can be more easily identified.

[0010] In various examples, a user interface can be generated that shows one or more blood vessels in which an abnormality may exist. In one or more exemplary examples, one or more images can be generated using data captured during a CT angiography imaging process. Based on the CT angiography data, a user interface can be generated that shows the blood vessels of the individual's brain. Additionally, the user interface can include indicators of blood vessels in which an abnormality may exist.

[0011] By analyzing images obtained from at least one of a perfusion-based imaging technique or a diffusion-based imaging technique to identify regions where vascular abnormalities exist, rather than relying solely on images obtained using CT angiography imaging techniques, the accuracy of identifying abnormalities related to distal vessels can be increased relative to existing techniques. Increased accuracy in identifying distal vessels where abnormalities exist can provide faster and more effective treatment to individuals, thereby reducing the amount of brain tissue damaged due to the abnormality. In particular, by identifying regions of an individual's brain where blood flow is disrupted through analysis of the values ​​of one or more perfusion parameters and / or one or more diffusion parameters, the number of candidate distal vessels where abnormalities may exist is reduced. In this way, analysis to identify a given vessel where an abnormality exists can be focused on vessels supplying the specified region, rather than analyzing vessels contained in a larger portion of the individual's brain. Therefore, by reducing the regions of the brain analyzed to identify abnormalities causing disruption of blood flow to a brain region as well as the number of vessels analyzed, the probability that an abnormal vessel will be incorrectly identified as a candidate vessel or completely excluded is reduced. In addition, the time required to identify abnormalities in distal vessels is reduced. Furthermore, the number of computational resources utilized to analyze a reduced number of blood vessels to identify abnormalities can be reduced relative to existing techniques that analyze a larger number of blood vessels over a larger region of the brain.

[0012] 1 is a schematic diagram of a computational architecture 100 for identifying cerebral blood vessels containing abnormalities, according to one or more exemplary implementations. The architecture 100 may include an image processing system 102. The image processing system 102 may be implemented by one or more computing devices 104. The one or more computing devices 104 may include one or more server computing devices, one or more desktop computing devices, one or more laptop computing devices, one or more tablet computing devices, one or more mobile computing devices, or a combination thereof. In certain implementations, at least a portion of the one or more computing devices 104 may be implemented in a distributed computing environment. For example, at least a portion of the one or more computing devices 104 may be implemented in a cloud computing architecture.

[0013] The image processing system 102 may obtain first image data 106 captured regarding the individual 108. The first image data 106 may be captured by a first image data source 110. The first imaging source 110 may include a device that generates the first image data 106 utilizing one or more imaging techniques. In one or more embodiments, the first image data source 110 may implement a computed tomography (CT) imaging technique. The CT imaging technique implemented by the first image data source 110 may include a perfusion-based CT imaging technique. In one or more further embodiments, the first image data source 110 may implement a magnetic resonance (MR) imaging technique. The MR imaging technique implemented by the first image data source 110 may include a perfusion-based imaging technique. In situations where the first image data source 110 implements a CT-based imaging technique or an MR-based imaging technique, the first image data 106 may include thin-slice volumetric data.

[0014] When images are taken dynamically while a contrast agent passes through the aorta, capillary beds, and draining veins for the purpose of obtaining hemodynamic parameters such as blood flow or blood arrival time to tissues, the technique is called CT perfusion (CTP). Disruption of the flow of contrast agent to a region of the brain can indicate a lack of blood supply to that region of the brain, which can result in damage to brain tissue in that region. In one or more embodiments, the contrast agent can include iodine or gadolinium added in a carrier solution. Perfusion-based MR imaging techniques implementing contrast agents can also be used to identify regions of an individual's brain that have disrupted blood supply.

[0015] In addition, non-contrast-based imaging techniques can be implemented to identify regions of an individual's brain that may be damaged due to a lack of blood supply. For example, diffusion-based MR imaging techniques can be used to identify portions of an individual's brain that have been disrupted from blood supply and are therefore damaged. In at least some instances, damage to brain tissue may be irreversible. In one or more exemplary embodiments, diffusion-based MR imaging techniques can be implemented to identify brain tissue that has been damaged due to a lack of blood supply.

[0016] The image processing system 102 may also obtain second image data 112 captured regarding the individual 108. The second image data 112 may be captured by a second image data source 114. The first image source 110 may include a device that utilizes one or more imaging techniques to generate the first image data 106. In one or more examples, the first image data source 110 may implement a computed tomography (CT) imaging technique. The CT imaging technique implemented by the first image data source 110 may include a CT angiography imaging technique. In various examples, the second image data 112 may include thin-slice volume data. In one or more exemplary examples, a contrast agent may be delivered to the individual 108, and a CT image may be captured showing the flow of the contrast agent through the blood vessels that supply blood to the brain of the individual 108. When images are taken while the contrast agent is still in the arteries, this is referred to as CT angiography (CTA). CTA imaging techniques involve capturing images of the blood vessels that carry blood to the brain and can provide an indication of narrowing or blockage of the blood vessels.

[0017] The first image data 106 can include one or more data files containing data related to images captured by the first image data source 110, and the second image data 112 can include one or more data files containing data related to images captured by the second image data source 114. In one or more embodiments, at least one of the first image data 106 or the second image data 112 can be formatted according to one or more image data formats, such as the Digital Imaging and Communication in Medicine (DICOM) format. In one or more further embodiments, at least one of the first image data 106 or the second image data 112 can be formatted according to at least one of the portable network graphics (PNG) format, the joint photographic experts group (JPEG) format, or the high efficiency image file format (HEIF). The first image data 106 and the second image data 112 can be rendered by the image processing system 102 to generate one or more images that can be displayed on a display device. In various examples, the first image data 106 and the second image data 112 may include a series of images of the individual 108. In one or more example examples, the first image data 106 and the second image data 112 may correspond to one or more features of the head of the individual 108. In one or more scenarios, the first image data 106 and the second image data 108 may be rendered to show internal features of the head of the individual 108, such as blood vessels or brain tissue located within the head of the individual 108.

[0018] The image processing system 102 may include a spatial data analysis system 116. The spatial data analysis system 116 may implement one or more computational models to analyze values ​​of one or more perfusion parameters or one or more diffusion parameters generated from the first image data 106 to identify the location of vascular abnormalities causing a disruption of blood flow to a region of the brain of the individual 108. In various examples, the spatial data analysis system 116 may generate one or more images based on the values ​​of the one or more perfusion parameters or one or more diffusion parameters obtained from the first image data 106. The one or more images may exhibit spatial patterns corresponding to different values ​​of the one or more perfusion parameters or one or more diffusion parameters. The spatial patterns may indicate a disruption of blood flow in a region of the brain of the individual 106. Additionally, the spatial patterns obtained from the first image data 106 may be analyzed by one or more computational models to determine that the spatial patterns correspond to blockages of respective blood vessels in the brain of the individual 108.

[0019] The spatial data analysis system 116 may include a vascular anomaly detection system 118. The vascular anomaly detection system 118 may analyze information included in the first image data 106 to determine whether an anomaly is present in the blood vessels of the brain of the individual 108. In various embodiments, the vascular anomaly detection system 118 may analyze intensity values ​​of voxels in the first image data 106 for one or more regions of the brain of the individual 108 to determine one or more perfusion parameters. The values ​​of the one or more perfusion parameters may be analyzed to determine the presence or absence of an anomaly in the blood vessels of the brain of the individual 108. Additionally, the vascular anomaly detection system 118 may analyze intensity values ​​of voxels derived from one or more diffusion-related parameters to determine the presence or absence of an anomaly in the blood vessels of the brain of the individual 108.

[0020] In one or more exemplary embodiments, the vascular abnormality detection system 118 may analyze intensity information of voxels corresponding to several regions of the brain of the individual 108 over a period of time to determine values ​​of one or more perfusion parameters. In one or more embodiments, the one or more perfusion parameters may be Tmax The vascular abnormality detection system 118 may analyze the value of one or more perfusion parameters against one or more thresholds to determine the presence or absence of an abnormality in the blood vessels of the brain of the individual 108. max In scenarios where the value of T is analyzed, thresholds of at least 2 seconds, at least 4 seconds, at least 6 seconds, at least 8 seconds, at least 10 seconds, at least 12 seconds, or at least 14 seconds may be utilized. In various embodiments, a threshold of at least 2 seconds, at least 4 seconds, at least 6 seconds, at least 8 seconds, at least 10 seconds, at least 12 seconds, or at least 14 seconds may be utilized. max The value of T max For example, a value of T that meets a first threshold can be displayed as a color in the user interface. max The first voxels with a value of T can be displayed as a first color, and the second voxels with a value of T that meets the second threshold. max A second voxel having a value of may be displayed as a second color. Additionally, a tissue damage threshold may be specified that corresponds to a disruption of blood flow to a region of the brain of the individual 108 that may result in damage to tissue of the brain of the individual 108.

[0021] The image processing system 102 may also include a brain region identification system 120 that determines one or more regions of the brain of the individual 108 that may have disrupted blood flow. The brain region identification system 120 may determine that a region of the brain of the individual 108 is disrupted by identifying a group of voxels associated with a value of a perfusion parameter that is equal to or greater than a threshold. In various embodiments, the brain region identification system 120 may identify a region of the brain of the individual 108 that is associated with disruption of blood flow to the region by determining that a minimum number of adjacent voxels have a value of a perfusion parameter that meets or exceeds a threshold. In one or more exemplary embodiments, a T above a threshold may be determined. max A group of voxels having a value of may indicate a region of the brain of the individual 108 that has a disrupted blood supply. In various embodiments, a region may be identified as a region of interest that has at least a minimum threshold of a potential disrupted blood supply.

[0022] The brain region identification system 120 may also implement one or more machine learning techniques to identify one or more regions of the brain of the individual 108 that have a disrupted blood supply. In one or more embodiments, one or more machine learning techniques may be used to identify regions of tissue in the brain of the individual 108 that have at least a threshold value that may have a disrupted blood supply. In one or more embodiments, one or more convolutional neural networks may be implemented to identify regions of the brain of the individual that have a disrupted blood supply. For example, one or more classification convolutional neural networks may be implemented to identify regions of the brain of the individual that have a disrupted blood supply due to abnormalities in the blood vessels of the brain of the individual 108.

[0023] In one or more exemplary embodiments, the brain region identification system 120 can acquire several CT perfusion images as training images. The training images can include a first number of images of a brain of a first individual having one or more regions with disrupted blood supply and a second number of images of a brain of a second individual without any regions with disrupted blood supply. In one or more scenarios, the first number of images can be classified as having one or more regions with disrupted blood supply, and the second number of images can be classified as not having any regions with disrupted blood supply. In the images with disrupted blood supply, T max Based on this, the network can classify the likelihood that a particular branch / segment is occluded (or the T delay caused by other causes). In one or more embodiments, a region classified as having disrupted blood supply is one that has a T delay of at least a threshold T max and / or a minimum number of voxels having a value of MTT. Values ​​for the parameters of the one or more models generated in conjunction with one or more machine learning techniques can be determined through a training process. After the training process is complete and the one or more models are validated using additional sets of images, regions of the new images can be identified by the brain region identification system 120 as having or not having blood flow disruption using the one or more models.

[0024] Additionally, the image processing system 102 may include a blood vessel identification system 122. The blood vessel identification system 122 may identify blood vessels in which an abnormality exists that results in a disruption of blood flow to a region of the brain of the individual 108. In one or more embodiments, the image processing system 102 may acquire reference data 124. The reference data 124 may include a mapping between regions of the human brain and the blood vessels that supply blood to the respective regions. In various embodiments, the blood vessel identification system 122 may use the reference data 124 to identify blood vessels in which an abnormality exists, such as a blockage and / or stenosis. In one or more exemplary embodiments, the brain region identification system 120 may identify a region of interest in the brain of the individual 108 in which blood flow is disrupted. The blood vessel identification system 122 may then analyze the region of interest in relation to the reference data 124. For example, the blood vessel identification system 122 may use the reference data 124 to identify one or more blood vessels that supply blood to a portion of the brain that includes the region of interest. Based on the location of the region of interest within the portion of the brain, the blood vessel identification system 122 may identify blood vessels in which an abnormality exists. In various embodiments, the vessel identification system 122 can identify blood vessels in the brain of the individual 108 that have at least a threshold likelihood of having an abnormality that causes a disruption of blood flow to the region of interest.

[0025] In one or more further embodiments, the blood vessel identification system 122 can analyze the second image data 112 to identify blood vessels in which an abnormality exists. For example, the blood vessel identification system 122 can analyze the opacity of voxels in the second image data 112 to identify blood vessels in the brain of the individual 108 that have at least a threshold where an abnormality may exist. In these scenarios, the blood vessel identification system 122 can utilize the region of interest identified by the brain region identification system 120 to focus its analysis on a portion of the second image data 122 that corresponds to the region of interest. In this manner, the brain region identification system 120 has already identified the region of interest that has at least a threshold where blood flow may be disrupted, increasing the likelihood of accurately identifying the blood vessels with an abnormality. In this manner, because the analysis performed by the blood vessel identification system 122 corresponds to the region of interest and not to other regions of the brain of the individual 108, the likelihood that one or more algorithms implemented by the blood vessel identification system 122 will identify a false positive or false negative in other regions of the brain of the individual 108 is reduced.

[0026] The image processing system 102 may generate a system output 126. The system output 126 may include a vascular abnormality alert 128. The vascular abnormality alert 128 may include an audio notification, a visual notification, a text notification, a symbolic notification, a video notification, one or more combinations thereof, etc., indicating the presence of an abnormality in a blood vessel of the brain of the individual 108. The vascular abnormality alert 128 may be provided using one or more output devices of the computing device, such as a display device and / or a speaker. The vascular abnormality alert 128 may be generated based on an output from the vascular abnormality detection system 118 indicating the presence of an abnormality in the brain of the individual 108. In one or more exemplary embodiments, the vascular abnormality alert 128 may be generated in response to the vascular abnormality detection system 118 determining that at least a threshold number of voxels in one or more images of the first image data 106 have at least threshold values ​​for one or more perfusion parameters.

[0027] The system output 126 may also include a brain region identifier 130. The brain region identifier 130 may indicate a region of the individual's 108 brain that has a disrupted blood supply. In one or more embodiments, the brain region identifier 130 may be displayed within a user interface. In various embodiments, the brain region identifier may include an outline of the region of the individual's 108 brain that has a disrupted blood supply. In one or more further embodiments, the brain region identifier 130 may include a geometric shape or arrow that indicates the region of the individual's 108 brain that has a disrupted blood supply. In one or more exemplary embodiments, the brain region identifier 130 may be displayed as an overlay on an image of the individual's 108 brain, the image being obtained from at least one of the first image data 106 or the second image data 112. For example, the brain region identifier 130 may be displayed as an overlay in a user interface that indicates values ​​of one or more perfusion parameters obtained from the first image data 106. Additionally, the brain region identifier 130 may be displayed as an overlay in a user interface showing the blood vessels of the brain of the individual 112 obtained from at least one of the first image data 106 or the second image data 112 .

[0028] Additionally, the system output 126 may include a blood vessel identifier 132. The blood vessel identifier 132 may indicate a blood vessel in the brain of the individual 108 in which an abnormality is present. The blood vessel identifier 132 may be displayed as an overlay on an image of the brain of the individual 108, showing at least a portion of a blood vessel in the brain of the individual 108. In various examples, the blood vessel identifier 132 may highlight a blood vessel in which an abnormality is present. In one or more exemplary examples, the blood vessel identifier 132 may include an arrow indicating a blood vessel in the brain of the individual 108 in which an abnormality is present.

[0029] FIG. 2 illustrates a computational architecture 200 for training and implementing one or more computational models for analyzing spatial patterns obtained from image data, according to one or more implementations. The computational architecture 200 can include one or more spatial data analysis models 202. The one or more spatial data analysis models 202 can implement one or more machine learning techniques for determining abnormalities in an individual's brain. In various embodiments, the one or more spatial data analysis models 202 can identify locations of blood vessels in the individual's brain where abnormalities, such as blockages or stenoses, exist. In one or more embodiments, the one or more spatial data analysis models 202 can identify spatial patterns indicated by values ​​of perfusion or diffusion parameters corresponding to abnormalities in one or more blood vessels in the individual's brain. In one or more exemplary embodiments, the one or more spatial data analysis models 202 can be implemented in conjunction with one or more convolutional neural networks.

[0030] The one or more spatial data analysis models 202 can be trained using training data 204. The training data 204 can include image training data 206. The image training data 206 can include a number of images showing values ​​of perfusion parameters or values ​​of diffusion parameters. For example, the image training data 206 can include a number of images showing values ​​of T maxThe training images may include images corresponding to values ​​of the perfusion parameter or diffusion parameter. Different values ​​of the perfusion parameter or the diffusion parameter may be displayed differently relative to one another. For example, different values ​​of the perfusion parameter or the diffusion parameter may be displayed as different colors. As the value of the perfusion parameter or the diffusion parameter changes, the color displayed in the training images may also change. Thus, different values ​​of the perfusion parameter or the diffusion parameter may result in patterns present in the training images. The presence of vascular abnormalities in the individual's brain may generate respective patterns of values ​​of the perfusion parameter or the diffusion parameter. In one or more exemplary embodiments, a vascular abnormality located at a first location in the human brain may result in a first pattern of values ​​of the perfusion parameter or the diffusion parameter. Additionally, a further vascular abnormality located at a second location in the human brain may result in a second pattern of values ​​of the perfusion parameter or the diffusion parameter that differs from the first pattern. The image training data 206 may be used to train one or more spatial data analysis models 202 to recognize different patterns of values ​​of the perfusion parameter or the diffusion parameter corresponding to vascular abnormalities located at different locations in the human brain.

[0031] In one or more implementations, one or more spatial data analysis models can be trained by providing paired images and linguistic labels for each image. The labels can indicate where the most proximal site of arterial occlusion / stenosis is located. For example, the labels can indicate that the abnormality is located in the proximal ICA, distal ICA, M1-MCA, M2-MCA, M3-MCA, M4-MCA, vertebral artery, basilar artery, P1-PCA, P2-PCA, ..., A1-ACA, A2-ACA, ..., SCA, AICA, PICA. After training, the one or more spatial data analysis models 202 can de novo infer which arterial segment or branch is occluded based on previously untrained perfusion or diffusion learning and output the segment / branch name and the likelihood of a possible abnormality or common variant (e.g., fetal PCA).

[0032] In the exemplary embodiment of FIG. 2 , the image training data 206 may include a first image 208. The first image 208 may correspond to a perfusion parameter or diffusion parameter value of an individual in whom a cerebral vascular abnormality is not present. In these circumstances, the amount of contrast between voxels included in the first image 208 may be less than a threshold amount of contrast. Additionally, the image training data 206 may include a second image 210. The second image 210 may include a region 212 having voxels with at least a threshold amount of contrast with other regions of the brain. The region 212 may indicate voxels corresponding to a perfusion parameter or diffusion parameter value having at least a threshold amount. The presence of the region 212 may indicate a first vascular abnormality in the human brain. Furthermore, the image training data 206 may include a third image 214. The third image 214 may include an additional region 214 different from the region 212. The additional region 214 may also indicate voxels with at least a threshold amount of contrast with other regions of the brain. In one or more embodiments, the additional region 216 can indicate voxels corresponding to values ​​of the perfusion parameter or diffusion parameter having at least a threshold value. The presence of the additional region 216 can indicate an abnormality related to a second blood vessel in the human brain having a different location than the first blood vessel corresponding to region 212. The image training data 206 can include hundreds, up to thousands, up to tens of thousands, or more images corresponding to abnormalities in individual blood vessels located in respective locations in the human brain.

[0033] The training data 204 may also include classification data 218. The classification data 218 may indicate images included in the image training data 206 that correspond to abnormalities in respective blood vessels of the human brain. For example, the classification data 218 may label the first image 208 and other images included in the image training data 206 that have similar voxel values ​​as not being associated with a vascular abnormality. In addition, the classification data 218 may label the second image 210 and other images included in the image training data 206 that have a region similar to that of region 212 as being associated with an abnormality in the first blood vessel. Furthermore, the classification data 218 may label the third image 214 and other images included in the image training data 206 that have a region similar to that of further region 216 as being associated with an abnormality in the second blood vessel.

[0034] After the one or more spatial data analysis models 202 are trained, the spatial data analysis system 116 can access image data 220 of the individual 108. The image data 220 can be generated in response to one or more perfusion-based imaging techniques or one or more diffusion-based imaging techniques that capture images of the brain of the individual 108. The image data 220 can include an image 222. The image 222 can be derived from values ​​of one or more perfusion parameters or values ​​of one or more diffusion parameters. The image 222 can include a region 224 that has at least a certain amount of contrast relative to other regions of the image 222. The region 224 can indicate a disruption of blood flow to the region 224. In various embodiments, the region 224 can correspond to values ​​of one or more perfusion parameters or one or more diffusion parameters that have at least a threshold value.

[0035] One or more spatial data analysis models 202 can analyze image 222 to determine a likelihood that image 222 corresponds to one or more sets of images included in image training data 206. In various embodiments, one or more spatial data analysis models 202 can determine a likelihood that image 222 corresponds to an abnormality associated with a given blood vessel. In situations where one or more spatial data analysis models 202 determine that image 222 has at least a threshold likelihood corresponding to an abnormality in a blood vessel supplying a region of the individual's 108's brain similar to region 224, spatial data analysis system 116 can generate an additional image 228. The additional image 228 can indicate region 224 and can also indicate a location 226 of the blood vessel in which the abnormality is present. In one or more further embodiments, spatial data analysis system 116 can alternatively or additionally generate a text-based identifier of location abnormality 228. The identifier of location abnormality 228 can indicate an identifier of the blood vessel in which the abnormality is present, such as M3-MCA, P2-PCA, etc.

[0036] In situations where one or more spatial data analysis models 202 determine that the likelihood of image 222 corresponding to a vascular abnormality is less than a threshold likelihood, spatial data analysis system 116 can provide an indication of another cause of the presence of region 224 in image 222. For example, one or more spatial data analysis models 202 can determine that image 222 corresponds to a vascular variant in the brain of individual 108. Additionally, one or more spatial data analysis models 202 can determine that image 222 corresponds to a tumor present in the brain of individual 108, or that individual 108 is experiencing a migraine or stroke. In scenarios where one or more spatial data analysis models 202 determine that image 222 is less than a threshold likelihood corresponding to a cerebrovascular abnormality, spatial data analysis system 116 can analyze the resulting image from additional parameters to classify image 222. For example, one or more spatial data analysis models 202 can determine that T maxThe image showing values ​​of σ can be analyzed to determine the likelihood that the image 222 corresponds to a cerebrovascular abnormality, and the resulting image from values ​​of cerebral blood volume and / or cerebral blood flow can then be used to determine the conditions that produce the spatial pattern shown in the image 222. FIG. 3 shows a flowchart of a process for identifying cerebral blood vessels in which an abnormality is present. The process can be embodied in computer-readable instructions for execution by one or more processors, such that the operations of the process can be performed in part or entirely by the functional components of the image processing system 102. Thus, the process described below is an example with reference thereto in some circumstances. However, in other implementations, at least some of the operations of the process described with reference to FIG. 2 may be deployed on various other hardware configurations. Thus, the process described with reference to FIG. 2 is not intended to be limited to the image processing system 102 and can be performed in part or entirely by one or more additional components. While the described flowchart may depict operations as sequential processes, many of the operations can be performed in parallel or simultaneously. Additionally, the order of the operations may be rearranged. A process terminates upon completion of its operations. A process may correspond to a method, a procedure, an algorithm, or the like. The operations of the methods may be performed in whole or in part, may be performed in conjunction with some or all of the operations of other methods, and may be performed by any number of different systems, such as those described herein, or any part of a system, such as a processor included in any of the systems.

[0037] FIG. 3 is a flow diagram of a process 300 for identifying blood vessels in the brain where abnormalities are present, according to one or more exemplary implementations. In operation 302, process 300 may include acquiring image data from one or more image data sources. The image data may include images of the individual's brain. In one or more embodiments, the one or more image data sources may include one or more perfusion-based imaging systems. In one or more further embodiments, the one or more image data sources may include one or more diffusion-based imaging systems. In one or more further embodiments, the one or more image data sources may include one or more perfusion-based imaging systems and one or more diffusion-based imaging systems. The one or more perfusion-based imaging systems may include a CT perfusion imaging system. Additionally, the one or more perfusion-based imaging systems may include an MR perfusion imaging system. Furthermore, the one or more diffusion-based imaging systems may include an MR diffusion imaging system.

[0038] Process 300 may include determining a value of a perfusion parameter or a value of a diffusion parameter for voxels of one or more images at operation 304. In one or more embodiments, the value of the perfusion parameter or the value of the diffusion parameter may be determined by a system incorporating one or more deconvolution image processing techniques. In one or more exemplary embodiments, the perfusion parameter may be determined by a system incorporating one or more deconvolution image processing techniques. max In one or more further exemplary embodiments, the perfusion parameter may include a mean tracer transit time.

[0039] Additionally, in operation 306, process 300 can include determining that the plurality of values ​​of the perfusion parameter or the plurality of values ​​of the diffusion parameter has at least a threshold value. In various embodiments, the threshold value can indicate a disruption of blood flow to a portion of the individual's brain corresponding to each voxel associated with the plurality of values. In one or more embodiments, the threshold value can correspond to an amount of delay for blood to arrive at a region of the individual's brain.

[0040] Based on the plurality of values ​​of the perfusion parameter or the plurality of values ​​of the diffusion parameter having at least a threshold value, process 300 may proceed to at least one of operations 308, 310, or 312. At operation 308, an alert may be generated indicating that blood flow to a region of the individual's brain has been disrupted. The alert may include at least one of a visual alert or an audio alert. The alert may be provided to medical personnel via a computing device and / or a display device.

[0041] In operation 310, regions of the individual's brain in which blood flow has been disrupted may be identified. In one or more embodiments, the regions may correspond to voxels in one or more images having one or more perfusion parameters or one or more diffusion parameters with values ​​greater than a threshold. The regions may be shown on one or more user interfaces. In various embodiments, an overlay may show regions on the image of the individual's brain showing the blood vessels of the individual's brain. In one or more exemplary embodiments, the image of the individual's brain showing the blood vessels of the individual's brain may be generated using data acquired from a CT angiography system.

[0042] Additionally, in operation 312, blood vessels in the individual's brain that supply blood to the region where the abnormality is present can be identified. In one or more implementations, the blood vessels can be identified by analyzing reference data indicative of one or more blood vessels that supply blood to a portion of the brain that includes the region of interest. The blood vessels can be identified based on the location of the region of interest relative to the blood vessel(s) that supply blood to the region of interest. In various examples, the opacity of voxels corresponding to blood vessels located within the region can be analyzed to identify the blood vessels where the abnormality is present.

[0043] In view of the above disclosure, various aspects are described below. It should be noted that one or more features of the embodiments, either alone or in combination, may be considered within the disclosure of the present application. "Aspect 1" A method includes: accessing, by a computing system including one or more processing devices and one or more memory devices, first training data including a first plurality of images of an individual's brain, the first plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique and showing a first plurality of patterns of values ​​of one or more perfusion parameters or one or more diffusion parameters in an individual not having a brain abnormality; accessing, by the computing system, second training data including a second plurality of images of the individual's brain, the second plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique and showing a second plurality of patterns of values ​​of the one or more perfusion parameters or one or more diffusion parameters in an individual not having a brain abnormality; and generating, by the computing system, one or more computational models based on the first training data and the second training data. generating, by the computing system, one or more computational models that identify abnormalities present in blood vessels of the individual's brain; accessing, by the computing system, one or more additional images of the individual's brain, the one or more additional images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique; determining, by the computing system, one or more patterns of the one or more perfusion parameters or the one or more diffusion parameters contained in the one or more additional images; analyzing, by the computing system and using the one or more computational models, the one or more patterns in relation to the first plurality of patterns and the second plurality of patterns to determine a measure of likelihood that the abnormality is present in the individual's brain; and generating, by the computing system and based on the measure of likelihood, a user interface that includes an indication that the abnormality is present in the individual's brain.

[0044]

[0022] Aspect 2. The method of aspect 1, comprising determining a value of a perfusion parameter indicative of a time to maximum arrival of a tissue residue function for voxels of one or more images.

[0023] [Aspect 3] A method as described in aspect 2, comprising performing, by a computing system, one or more deconvolution operations on contrast agent concentration curves of voxels contained in one or more images with respect to an arterial input function to generate a tissue residue function.

[0045] "Aspect 4" A method according to any one of aspects 1 to 3, comprising determining a value of a perfusion parameter indicative of the mean transit time of a contrast agent through a region of the brain. "Aspect 5" The method according to any one of aspects 1 to 4, wherein one or more additional images show values ​​of perfusion parameters or diffusion parameters in multiple regions of the individual's brain.

[0046] "Aspect 6" A method according to any one of aspects 1 to 5, wherein the first training data includes classification data that labels a first plurality of images as being obtained from a first individual in which no abnormalities are present and labels a second plurality of images as being obtained from a second individual in which an abnormality is present in the blood vessels of the second individual's brain.

[0047] "Aspect 7" A method according to any one of aspects 1 to 6, wherein the second training data includes one or more first additional images corresponding to a first vascular abnormality that disrupts blood flow to a first region of the human brain and one or more second additional images corresponding to a second vascular abnormality that disrupts blood flow to a second region of the human brain, and the first blood vessel has a different position than the second blood vessel in the human brain.

[0048] "Aspect 8" The method of aspect 7, wherein the second training data includes first additional classification data that labels one or more first additional images as associated with a first vascular abnormality, and second additional classification data that labels one or more second additional images as associated with a second vascular abnormality.

[0049] Aspect 9: A method according to any one of aspects 1 to 8, comprising training, by a computing system, one or more convolutional neural networks using the first training data and the second training data to identify when an abnormality is present in the individual's brain based on patterns contained in images obtained from one or more perfusion-based imaging techniques or one or more diffusion-based imaging techniques.

[0050] [Embodiment 10] The method of any one of embodiments 1 to 9, wherein the one or more perfusion-based imaging techniques include a computed tomography perfusion-based imaging technique or a magnetic resonance perfusion-based imaging technique.

[0051] "Aspect 11" A method according to any one of aspects 1 to 10, comprising generating, by a computing system, a text-based identifier of the blood vessel in which the abnormality is present based on the likelihood that the abnormality is present in the individual's brain.

[0052] "Aspect 12" A method according to any one of Aspects 1 to 11, wherein the abnormality present in the blood vessels of the brain of an individual with an abnormality is a distal blood vessel of the human brain, and the distal blood vessel includes at least one of the A2 to A5 segments of the anterior cerebral artery, the M2 to M4 segments of the middle cerebral artery, the P2 to P4 segments of the posterior cerebral artery, the posterior inferior cerebellar artery, the anterior inferior cerebellar artery, or the superior cerebellar artery.

[0053] Aspect 13: A system comprising one or more hardware processors and one or more non-transitory computer-readable storage media comprising computer-readable instructions, wherein the computer-readable instructions, when executed by the one or more hardware processors, cause the one or more hardware processors to: access first training data comprising a first plurality of images of an individual's brain, the first plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique and showing a first plurality of patterns of values ​​of one or more perfusion parameters or one or more diffusion parameters in an individual without a brain abnormality; and acquire a second plurality of images of the individual's brain, the second plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique and showing a second plurality of patterns of values ​​of the one or more perfusion parameters or one or more diffusion parameters in an individual with a brain abnormality. generating, based on the first training data and the second training data, one or more computational models that identify abnormalities present in the blood vessels of the individual's brain; accessing one or more additional images of the individual's brain, the one or more additional images captured using a perfusion-based imaging technique or a diffusion-based imaging technique; determining one or more patterns of one or more perfusion parameters or one or more diffusion parameters contained in the one or more additional images; analyzing the one or more patterns in relation to the first and second plurality of patterns using the one or more computational models to determine a measure of likelihood that an abnormality is present in the individual's brain; and generating, based on the measure of likelihood, a user interface that includes an indication that an abnormality is present in the individual's brain.

[0054] "Aspect 14" The system described in Aspect 13, wherein one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations including acquiring additional image data corresponding to one or more additional images captured using computed tomography angiography imaging techniques, determining a portion of the additional image data corresponding to a region of the individual's brain in which an abnormality is present, analyzing the portion of the additional image data to identify a blood vessel supplying blood to the region, and determining that an abnormality is present in the blood vessel.

[0055] "Aspect 15" The system described in Aspect 14, wherein one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform additional operations including: determining, based on the additional image data, intensity values ​​of voxels corresponding to a plurality of candidate blood vessels that supply blood to a region of the individual's brain, wherein the blood vessel is included in the plurality of candidate blood vessels; and determining, based on the intensity values, that an abnormality exists in the blood vessel.

[0056] "Aspect 16" A system described in any one of aspects 13 to 15, wherein one or more non-transitory computer-readable storage media contain additional computer-readable instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform additional operations including identifying regions of the individual's brain corresponding to voxels having perfusion parameter values ​​or diffusion parameter values ​​that have at least a threshold value.

[0057]

[0033] "Aspect 17" One or more non-transitory computer-readable media storing computer-readable instructions that, when executed by one or more processing devices, cause the one or more processing devices to: access first training data including a first plurality of images of an individual's brain, the first plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique and showing a first plurality of patterns of values ​​of one or more perfusion parameters or one or more diffusion parameters in an individual without a brain abnormality; access second training data including a second plurality of images of the individual's brain, the second plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique and showing a second plurality of patterns of values ​​of the one or more perfusion parameters or one or more diffusion parameters in an individual with a brain abnormality; and one or more non-transitory computer-readable media configured to perform operations including: generating, based on the data and second training data, one or more computational models that identify abnormalities present in the blood vessels of the individual's brain; accessing one or more additional images of the individual's brain, the one or more additional images captured using a perfusion-based imaging technique or a diffusion-based imaging technique; determining one or more patterns of one or more perfusion parameters or one or more diffusion parameters contained in the one or more additional images; analyzing, using the one or more computational models, the one or more patterns in relation to the first plurality of patterns and the second plurality of patterns to determine a measure of likelihood that an abnormality is present in the individual's brain; and generating, based on the measure of likelihood, a user interface that includes an indication that an abnormality is present in the individual's brain.

[0058] "Aspect 18" One or more non-transitory computer-readable media described in Aspect 17, wherein an operation includes generating user interface data corresponding to a user interface showing values ​​of a perfusion parameter for voxels of image data, wherein a first range of values ​​of the perfusion parameter is displayed as a first color within the user interface and a second range of values ​​of the perfusion parameter is displayed as a second color within the user interface.

[0059] "Aspect 19" One or more non-transitory computer-readable media described in aspect 18, wherein the operation includes generating additional user interface data corresponding to an additional user interface showing intensity values ​​of voxels, the additional user interface showing blood vessels in the individual's brain, and a portion of the voxels corresponding to the blood vessels in the individual's brain having intensity values ​​greater than the additional voxels.

[0060] "Aspect 20" One or more non-transitory computer-readable media described in aspect 19, wherein the first user interface and the second user interface are configured to be displayed together, and the first user interface data and the second user interface data are configured to be rendered to generate a combined user interface including the first user interface and the second user interface, and regions of the individual's brain are highlighted in the user interface and highlighted in the additional user interface.

[0061] FIG. 4 is a block diagram illustrating components of a machine 400, according to some example implementations, capable of reading instructions from a machine-readable medium (e.g., a machine-readable storage medium) and performing any one or more of the methods described herein. Specifically, FIG. 4 illustrates a schematic diagram of the machine 400 in the example form of a computer system, within which instructions 402 (e.g., software, programs, applications, applets, apps, or other executable code) can be executed that cause the machine 400 to perform any one or more of the methods described herein. Thus, the instructions 402 can be used to implement modules or components described herein. The instructions 402 transform a general, unprogrammed machine 400 into a specific machine 400 that is programmed to perform the described and illustrated functions in the described manner. In alternative implementations, the machine 400 can operate as a standalone device or can be coupled (e.g., networked) to other machines. In a networked deployment, the machine 400 can operate in the capacity of a server machine or a client machine in a server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. Machine 400 may include, but is not limited to, a server computer, a client computer, a personal computer (PC), a tablet computer, a laptop computer, a netbook, a set-top box (STB), a personal digital assistant (PDA), an entertainment media system, a mobile phone, a smartphone, a mobile device, a wearable device (e.g., a smart watch), a smart home device (e.g., a smart appliance), other smart devices, a web appliance, a network router, a network switch, a network bridge, or any machine capable of executing, sequentially or otherwise, instructions 402 that specify actions to be taken by machine 400.Furthermore, although only a single machine 400 is illustrated, the term "machine" is also intended to include a collection of machines that individually or jointly execute instructions 402 to perform any one or more of the methods described herein.

[0062] Machine 400 may include a processor 404, memory / storage 406, and I / O components 408, which may be configured to communicate with each other, such as via a bus 410. A "processor" in this context refers to any circuit or virtual circuit (a physical circuit emulated by logic executing on the actual processor 404) that manipulates data values ​​in accordance with control signals (e.g., "commands," "opcodes," "machine code," etc.) and generates corresponding output signals that are applied to operate machine 400. In an example implementation, processor 404 (e.g., a central processing unit (CPU), a reduced instruction set computing (RISC) processor, a complex instruction set computing (CISC) processor, a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a radio-frequency integrated circuit (RFIC), another processor, or any suitable combination thereof) may include, for example, processor 412 and processor 414 capable of executing instructions 402. The term "processor" is intended to include multi-core processor 404, which may include two or more independent processors (sometimes referred to as "cores") capable of simultaneously executing instructions 402. Although FIG. 4 shows multiple processors 404, machine 400 may include a single processor 412 with a single core, a single processor 412 with multiple cores (e.g., a multi-core processor), multiple processors 312, 414 with a single core, multiple processors 312, 414 with multiple cores, or any combination thereof.

[0063] The memory / storage device 406 may include a memory, such as a main memory 416 or other memory storage device, and a storage unit 418, both of which are accessible to the processor 404, such as via a bus 410. The storage unit 418 and the main memory 416 store instructions 402 that embody any one or more of the methods or functions described herein. The instructions 402 may also reside, completely or partially, within the main memory 416, within the storage unit 418, within at least one of the processors 404 (e.g., within a processor's cache memory), or any suitable combination thereof, during execution of the instructions 402 by the machine 400. Thus, the main memory 416, the storage unit 418, and the memory of the processor 404 are examples of machine-readable media. A "machine-readable medium," also referred to herein as a "computer-readable storage medium," in this context refers to a component, device, or other tangible medium capable of temporarily or permanently storing instructions 402 and data, and may include, but is not limited to, random-access memory (RAM), read-only memory (ROM), buffer memory, flash memory, optical media, magnetic media, cache memory, other types of storage devices (e.g., erasable programmable read-only memory (EEPROM)), and / or any suitable combination thereof. The term "machine-readable medium" may be interpreted to include a single medium or multiple media (e.g., a centralized or distributed database, or associated caches and servers) capable of storing instructions 402. The term "machine-readable medium" shall also be taken to include any medium or combination of media capable of storing instructions 402 (e.g., code) for execution by machine 400, such that the instructions 402, when executed by one or more processors 404 of machine 400, cause machine 400 to perform any one or more of the methods described herein.Thus, "machine-readable medium" refers to a single storage device or device, as well as a "cloud-based" storage system or storage network that includes multiple storage devices or devices. The term "machine-readable medium" excludes the signal itself.

[0064] The I / O components 408 may include a wide variety of components for receiving input, providing output, generating output, transmitting information, exchanging information, capturing measurements, and the like. The particular I / O components 408 included in a particular machine 400 will depend on the type of machine. For example, a portable device such as a mobile phone will likely include a touch input device or other such input mechanism, while a headless server machine will likely not include such a touch input device. It will be understood that the I / O components 408 can include many other components not shown in FIG. 8 . The I / O components 408 are grouped according to function merely to simplify the following description, and this grouping is in no way limiting. In various exemplary implementations, the I / O components 408 may include a user output component 420 and a user input component 422. The user output components 420 may include visual components (e.g., a display such as a plasma display panel (PDP), a light emitting diode (LED) display, a liquid crystal display (LCD), a projector, or a cathode ray tube (CRT)), acoustic components (e.g., speakers), tactile components (e.g., vibration motors, resistance mechanisms), other signal generators, etc. The user input components 422 may include alphanumeric input components (e.g., a keyboard, a touchscreen configured to receive alphanumeric input, a photo-optical keyboard, or other alphanumeric input component), point-based input components (e.g., a mouse, touchpad, trackball, joystick, motion sensor, or other pointing device), tactile input components (e.g., physical buttons, a touchscreen that provides the position or force of a touch or touch gesture, or other tactile input component), audio input components (e.g., a microphone), etc.

[0065] In further example implementations, I / O component 408 may include a biometric component 424, a motion component 426, an environmental component 428, or a position component 430, among a wide variety of other components. For example, biometric component 424 may include components for detecting expressions (e.g., hand expressions, facial expressions, vocal expressions, gestures, or eye tracking), measuring biosignals (e.g., blood pressure, heart rate, body temperature, sweat, or brain waves), identifying people (e.g., voice identification, retinal identification, face identification, fingerprint identification, or brainwave-based identification), etc. Motion component 426 may include an acceleration sensor component (e.g., an accelerometer), a gravity sensor component, a rotation sensor component (e.g., a gyroscope), etc. The environmental components 428 may include, for example, a lighting sensor component (e.g., a light meter), a temperature sensor component (e.g., one or more thermometers that detect ambient temperature), a humidity sensor component, a pressure sensor component (e.g., a barometer), an acoustic sensor component (e.g., one or more microphones that detect background noise), a proximity sensor component (e.g., an infrared sensor that detects nearby objects), a gas sensor (e.g., a gas detection sensor that detects the concentration of harmful gases or measures pollutants in the air for safety purposes), or other components that can provide an indication, measurement, or signal corresponding to the surrounding physical environment. The location components 430 may include a location sensor component (e.g., a GPS receiver component), an altitude sensor component (e.g., an altimeter or barometer that detects air pressure from which altitude can be obtained), an orientation sensor component (e.g., a magnetometer), etc.

[0066] Communications can be implemented using a wide variety of technologies. I / O component 408 can include a communications component 432 operable to couple machine 400 to a network 434 or a device 436. For example, communications component 432 can include a network interface component or other suitable device for interfacing with network 434. In further embodiments, communications component 432 can include a wired communications component, a wireless communications component, a cellular communications component, a near field communications (NFC) component, a Bluetooth® component (e.g., Bluetooth Low Energy), a Wi-Fi® component, and other communications components that provide communications via other modalities. Device 436 can be another machine 400 or any of a wide variety of peripheral devices (e.g., a peripheral device coupled via USB).

[0067] Further, communication component 432 may detect an identifier or may include a component operable to detect an identifier. For example, communication component 432 may include a radio frequency identification (RFID) tag reader component, an NFC smart tag detection component, an optical reader component (e.g., an optical sensor for detecting one-dimensional barcodes such as Universal Product Code (UPC) barcodes, multidimensional barcodes such as Quick Response (QR) Code, Aztec Code, Data Matrix, Dataglyph, MaxiCode, PDF417, Ultra Code, UCC RSS-2D barcodes, and other optical codes), or an acoustic detection component (e.g., a microphone for identifying tagged audio signals). Additionally, various information can be obtained via communication component 432, such as location via Internet Protocol (IP) geolocation, location via Wi-Fi signal triangulation, location via detecting NFC beacon signals that can indicate a specific location.

[0068] A "component" in this context refers to a device, physical entity, or logic having boundaries defined by function or subroutine calls, branch points, APIs, or other techniques that provide division or modularization of specific processing or control functions. Components can be combined with other components via their interfaces to perform machine processes. A component may also be a packaged functional hardware unit designed for use with other components, and typically a part of a program that performs specific functions of associated functionality. A component can constitute either a software component (e.g., code embodied on a machine-readable medium) or a hardware component. A "hardware component" is a tangible unit capable of performing specific operations and can be configured or arranged in a specific physical manner. In various exemplary implementations, one or more computer systems (e.g., standalone computer systems, client computer systems, or server computer systems) or one or more hardware components of a computer system (e.g., a processor or group of processors) can be configured as hardware components that operate by software (e.g., an application or application portion) to perform specific operations as described herein.

[0069] A hardware component may also be implemented mechanically, electronically, or any suitable combination thereof. For example, a hardware component may include dedicated circuitry or logic permanently configured to perform specific operations. A hardware component may be a dedicated processor, such as a field-programmable gate array (FPGA) or an ASIC. A hardware component may also include programmable logic or circuitry temporarily configured by software to perform specific operations. For example, a hardware component may include software executed by a general-purpose processor 404 or other programmable processor. Once configured by such software, the hardware component becomes a specific machine (or a specific component of machine 400) uniquely tailored to perform the configured function, and is no longer a general-purpose processor 404. It will be appreciated that the decision to implement a hardware component mechanically, with dedicated and permanently configured circuitry, or with temporarily configured (e.g., configured by software) circuitry may be influenced by cost and time considerations. Thus, the phrase "hardware component" (or "hardware-implemented component") should be understood to encompass a tangible entity, an entity that is physically constructed, permanently configured (e.g., hardwired), or temporarily configured (e.g., programmed) to operate in a particular manner or to perform particular operations described herein. Considering implementations in which the hardware components are temporarily configured (e.g., programmed), each of the hardware components need not be configured or instantiated at any one time.For example, if the hardware components include a general-purpose processor 404, processor 404, configured by software to be a special-purpose processor, then the general-purpose processor 404, processor 404, can be configured at different times (e.g., with different hardware components) as different special-purpose processors. The software accordingly configures the particular processor 412, 414, or processor 404, e.g., to configure a particular hardware component at one time and a different hardware component at a different time.

[0070] Hardware components can provide information to and receive information from other hardware components. Thus, the described hardware components can be considered to be communicatively coupled. When multiple hardware components are present simultaneously, communication can be achieved through signal transmission between two or more of the hardware components (e.g., via appropriate circuits and buses). In implementations in which multiple hardware components are configured or instantiated at different times, communication between such hardware components can be achieved, for example, through the storage and retrieval of information in memory structures to which the multiple hardware components have access. For example, one hardware component can perform an operation and store the output of that operation in a memory device to which the one hardware component is communicatively coupled. An additional hardware component can then later access the memory device to retrieve and process the stored output.

[0071] Hardware components may also initiate communications with input or output devices and operate on resources (e.g., collections of information). Various operations of the example methods described herein may be performed, at least in part, by one or more processors 404 that are temporarily or permanently configured (e.g., by software) to perform the associated operations. Whether temporarily or permanently configured, such processors 404 may constitute processor-implemented components that operate to perform one or more operations or functions described herein. As used herein, a "processor-implemented component" refers to a hardware component implemented using one or more processors 404. Similarly, the methods described herein may be at least in part processor-implemented, with a particular processor 412, 414, or processor 404 being an example of hardware. For example, at least some of the operations of the methods may be performed by one or more processors 404 or processor-implemented components. Furthermore, one or more processors 404 may also operate to support execution of the associated operations in a "cloud computing" environment or as "software as a service" (SaaS). For example, at least some of the operations may be performed by a group of computers (e.g., machine 300 including processor 404, as an example), and these operations are accessible via network 434 (e.g., the Internet) and via one or more appropriate interfaces (e.g., APIs). Performance of certain operations may be distributed among processors and may reside within a single machine 400 as well as spread across multiple machines. In some example implementations, processor 404 or processor-implemented components may be located in a single geographic location (e.g., in a home environment, an office environment, or a server farm). In other example implementations, processor 404 or processor-implemented components may be distributed across several geographic locations.

[0072] FIG. 5 is a block diagram illustrating a system 500 including an exemplary software architecture 502 that can be used in conjunction with various hardware architectures described herein. FIG. 4 is a non-limiting example of a software architecture, and it will be appreciated that many other architectures can be implemented to facilitate the functionality described herein. The software architecture 502 can execute on hardware such as the machine 400 of FIG. 4, which includes, among other things, a processor 404, memory / storage 406, and input / output (I / O) components 408. A representative hardware layer 504 is shown and may represent, for example, the machine 400 of FIG. 4. The representative hardware layer 504 includes a processing unit 506 having associated executable instructions 508. The executable instructions 508 represent executable instructions of the software architecture 502, including implementations of methods, components, and the like, described herein. The hardware layer 504 also includes at least one memory or storage module memory / storage 510 that also has the executable instructions 508. The hardware layer 504 may also include other hardware 512.

[0073] In the example architecture of FIG. 5 , software architecture 502 can be conceptualized as a stack of layers, with each layer providing a specific function. For example, software architecture 502 can include layers such as operating system 514, libraries 516, framework / middleware 518, application 520, and presentation layer 522. In operation, application 520 or other components within the layers can invoke API calls 524 through the software stack and receive messages 526 in response to API calls 524. The illustrated layers are representative in nature, and not all software architectures have all layers. For example, some mobile or dedicated operating systems may not provide framework / middleware 518, while others may provide such a layer. Other software architectures may include additional or different layers.

[0074] The operating system 514 may manage hardware resources and provide common services. The operating system 514 may include, for example, a kernel 528, services 530, and drivers 532. The kernel 528 may serve as an abstraction layer between the hardware layer and other software layers. For example, the kernel 528 may be responsible for memory management, processor management (e.g., scheduling), component management, networking, security configuration, etc. The services 530 may provide other common services to the other software layers. The drivers 532 are responsible for controlling or interfacing with the underlying hardware. For example, the drivers 532 may include a display driver, a camera driver, a Bluetooth driver, a flash memory driver, a serial communication driver (e.g., a Universal Serial Bus (USB) driver), a Wi-Fi driver, an audio driver, a power management driver, etc., depending on the hardware configuration.

[0075] Libraries 516 provide a common infrastructure used by applications 520 and / or other components or layers. Libraries 516 provide functions that allow other software components to perform tasks more easily than by directly interfacing with underlying operating system 514 functions (e.g., kernel 528, services 530, drivers 532). Libraries 516 may include system libraries 534 (e.g., standard C libraries) that may provide functions such as memory allocation functions, string manipulation functions, mathematical functions, etc. Additionally, libraries 516 may include API libraries 536 such as media libraries (e.g., libraries that support the presentation and manipulation of various media formats such as MPEG4, H.264, MP3, AAC, AMR, JPG, PNG, etc.), graphics libraries (e.g., OpenGL frameworks that can be used to render two-dimensional and three-dimensional graphical content on a display), database libraries (e.g., SQLite, which can provide various relational database functions), and web libraries (e.g., WebKit, which can provide web browsing functions). The library 516 may also include a wide variety of other libraries 538 to provide many other APIs to the application 520 and other software components / modules.

[0076] The framework / middleware 518 (sometimes called middleware) provides a higher-level common infrastructure that can be used by the applications 520 or other software components / modules. For example, the framework / middleware 518 can provide various graphical user interface functionality, high-level resource management, high-level location services, etc. The framework / middleware 518 can provide a wide range of other APIs that can be utilized by the applications 520 or other software components / modules, some of which may be specific to a particular operating system 514 or platform.

[0077] The applications 520 include built-in applications 540 and third-party applications 542. Representative examples of built-in applications 540 may include, but are not limited to, a contacts application, a browser application, a book reader application, a location application, a media application, a messaging application, or a game application. The third-party applications 542 may include applications developed using the ANDROID™ or IOS™ software development kit (SDK) by an entity other than the vendor of a particular platform, or may be mobile software running on a mobile operating system such as IOS, ANDROID, WINDOWS™ Phone, or other mobile operating system. The third-party applications 542 may invoke API calls 524 provided by the mobile operating system (e.g., operating system 514) to facilitate the functionality described herein.

[0078] Applications 520 can use built-in operating system facilities (e.g., kernel 528, services 530, drivers 532), libraries 516, and frameworks / middleware 518 to generate a UI for interacting with a user of the system. Alternatively or additionally, in some systems, interaction with the user may occur through a presentation layer, such as presentation layer 522. In these systems, application / component "logic" can be separated from the aspects of the application / component that interact with the user.

[0079] Changes and modifications can be made to the disclosed implementations without departing from the scope of the disclosure. These and other changes or modifications are intended to be included within the scope of the disclosure, as expressed in the following claims. [Example]

[0080] Example 1 Background and Objectives Detecting distal intracranial arterial occlusions on CTA is difficult but increasingly relevant for clinical decision-making. Our objective was to determine whether the use of time-to-maximum arrival (Tmax) maps of the tissue residue function obtained from CTP would improve diagnostic performance for detecting these occlusions.

[0081] Materials and Methods This retrospective study included 70 consecutive patients with distal arterial occlusion and 70 randomly selected controls who underwent multimodal CT with CTA and CTP for suspected acute ischemic stroke. Four readers with different levels of experience independently interpreted the CTA in two separate sessions, with and without Tmax maps, and recorded the presence or absence of occlusion, diagnostic confidence, and interpretation time. The accuracy of detecting distal occlusion was assessed using receiver operating characteristic (ROC) analysis, and the areas under the ROC curves (AUC) were compared to assess whether accuracy improved with the use of Tmax. Changes in diagnostic confidence and interpretation time were assessed using the Wilcoxon signed-rank test.

[0082] result The mean sensitivity for detecting occlusion on CTA increased from 70.7% to 90.4% with the use of Tmax maps. Diagnostic accuracy improved significantly for all four readers (p<0.001), with AUCs increasing by 0.186, 0.136, 0.114, and 0.121, respectively. Diagnostic confidence and speed also increased significantly.

[0083] conclusion All evaluated metrics of diagnostic performance for detecting distal arterial occlusion were improved with the use of T maps, encouraging their use to aid CTA interpretation by both experienced and inexperienced readers. These findings demonstrate the additional diagnostic value of including CTP in acute stroke imaging protocols.

[0084] Introduction Intravenous thrombolysis is the mainstay of treatment for arterial occlusions distal to the internal carotid artery, the M1 segment of the middle cerebral artery (MCA), and the vertebral and basilar arteries. These occlusions are called "distal vessel occlusions" (DVO) to distinguish them from proximal large-vessel occlusions. Although demonstration of DVO is not a prerequisite for thrombolysis, detection of DVO is becoming increasingly relevant to clinical decision-making.

[0085] The primary reason is that endovascular thrombetomy (EVT) can be used to treat occlusions involving large and medium-sized distal arteries in carefully selected patients. Evidence exists that EVT improves functional outcomes compared with standard medical management in patients with occlusions of the M2 segment of the MCA. Therefore, EVT is increasingly being considered for M2 occlusions, and it is also safe and technically feasible for occlusions involving the M3 segment of the MCA, the anterior cerebral artery (ACA), or the posterior cerebral artery (PCA). Advances in endovascular device technology have led to the development of smaller, more navigable stent retrievers and thrombus aspiration devices that can reach smaller distal arteries, including the M4 segment of the MCA and the A4 segment of the ACA. Because these DVOs can cause severe neurological deficits if delivered to normal brain regions, EVT may be warranted to achieve rapid reperfusion. It is also the only option for reperfusion in patients who are ineligible for thrombolysis. Therefore, distal vascular EVT is considered the "next promising frontier" for stroke treatment and is the subject of current research. Because demonstration of target arterial occlusion is required for triage to EVT, rapid and accurate detection of DVO is crucial to ensure timely treatment.

[0086] Detecting DVO also allows for accurate diagnosis, which in turn is important for prognosis and ongoing management, including workup for embolic sources and secondary prevention. If the treatment window extends beyond 4.5 hours, targeted DVO detection may become a prerequisite for thrombolysis to avoid futile treatment and justify the increased risk of thrombolysis.

[0087] CTA has become a routine part of acute stroke imaging protocols. Its primary purpose is to identify patients with proximal large-vessel occlusions for triage to EVT. DVOs are more difficult to detect on CTA than these proximal occlusions due to the smaller caliber, greater number, and more inadequate opacification of the distal arteries. Reported sensitivity is as low as 33%, with 35% of M2-segment MCA occlusions being missed at the time of initial CTA evaluation in one recent study.

[0088] CTP is now widely included in acute stroke CT protocols. The time to maximum arrival of the tissue residue function (Tmax) is a parameter that can always be obtained from CTP when deconvolution-based postprocessing is used. Tmax has been well established for identifying salvageable ischemic penumbra in patients with proximal vascular occlusion. In our clinical practice, we observed that Tmax delay within a vascular territory indicates severe stenosis or occlusion of the supplying artery. This information can then be used to detect and localize distal arterial occlusions on CTA. Otherwise, these occlusions may be missed or difficult to find. Despite its real-world value in routine clinical practice, no previous studies have evaluated and quantified the diagnostic utility of Tmax for detecting intracranial arterial occlusions.

[0089] The purpose of this study was to evaluate the added value of Tmax maps and to verify our clinical impression that they facilitate the detection of distal occlusion on CTA. We hypothesized that the diagnostic accuracy, speed, and confidence for detecting DVO on CTA would increase with the use of Tmax in readers with different levels of experience.

[0090] method Patient Selection Five hundred and one consecutive patients who visited our primary stroke center between January 1, 2017, and December 31, 2018, and underwent multimodal CT for suspected stroke were screened using our Picture Archiving and Communication System and Electronic Medical Records. Raw and postprocessed images were evaluated for technical adequacy by a neuroradiologist with 9 years of post-fellowship experience. Patients who met the following inclusion criteria were retrospectively identified: a. age ≥ 18 years; b. multimodal CT with CTA and CTP; and c. symptom onset or last known presentation within 24 hours. Exclusion criteria were: a. technically inadequate CTP or CTA (insufficient contrast bolus or substantial motion); b. unavailable thin-slice CTA images; and c. occlusion of the internal carotid artery, M1 segment of the MCA, vertebral artery, or basilar artery (excluded to allow specific evaluation of diagnostic performance for detecting more distal occlusions). One hundred twenty-eight patients were excluded: 84 with large vessel occlusion and 42 with technically inadequate CTA or CTP (patient selection flow chart shown in Supplementary Material Fig. 12).

[0091] Multimodal stroke CT scans of all consecutive patients who met the inclusion criteria were reviewed by a neuroradiologist who had access to all clinical records and images. All consecutive patients with DVO were identified and included in the study. An equal number of patients without any vascular occlusion were randomly selected from the remaining patients and included in the study. Data processing, scan anonymization, and randomization were performed by this neuroradiologist.

[0092] DVO was defined as arterial occlusion involving the A2-A5 segments of the ACA, the M2-M4 segments of the MCA, the P2-P4 segments of the PCA, or the PICA, AICA, or SCA. Proximal M2 occlusions are difficult to classify due to the large anatomical variability between patients in size and dominance. Although some may be considered proximal or large vessel occlusions, they are not recognized as such by the AHA guidelines and are more difficult to detect on CTA than M1 occlusions; therefore, they are included as "DVO" in this study.

[0093] This study was approved by the local institutional review board, which granted a waiver of written consent based on a retrospective study design and de-identification of all data. This investigator-initiated study received no financial support.

[0094] CT image acquisition, reconstruction, and post-processing All patients were scanned with a 256-slice multidetector CT (iCt 256, Philips Healthcare, Cleveland, OH, USA). Our prescribed multimodal "stroke CT" protocol consisted of non-enhanced CT (NECT), followed by CTP, and then CTA. Scanning techniques and parameters are detailed in the Supplementary Material.

[0095] For CTP, images were acquired axially, reconstructed with 10 mm slice thickness, and processed using a commercially available software platform (RAPID 4.9, iSchemaView, Menlo Park, California) that uses a delay-insensitive deconvolution algorithm with automated arterial input function selection. The software calculates Tmax values ​​for each image voxel over the range of 0 to 12 seconds in 2-second increments and displays them on a color-scale map.

[0096] Spiral CTA images were reconstructed axially in 0.8 mm slices, as well as in three planes (axial, coronal, and sagittal), 4 mm thick multiplanar reformats (MPR), and 10 mm thick maximum intensity projections (MIP).

[0097] reference standard Two neuroradiologists (with 9 and 20 years of post-fellowship experience, respectively) consensually read the CTAs using a systematic approach in conjunction with all available clinical and imaging data, including NECT and all CTP parametric maps (CBF, CBV, MTT, and Tmax) and any available follow-up scans. These "expert reads" served as the reference standard.

[0098] Image Review CTA images were independently interpreted by four readers with different levels of experience: a second-year radiology resident, a neuroradiology fellow, an attending radiologist (with a 2-year post-cardiovascular fellowship), and an imaging scientist. These readers had 18 months, 6 years, 8 years, and 20 years of experience, respectively, in interpreting acute stroke images. All readers had prior knowledge of the major cerebral arteries, their segments, and their supply territories, acquired through routine radiological training and clinical practice. No additional training was provided for this study. Details of the patient's presenting neurologic deficits were provided (to reflect clinical practice), but readers were blinded to all other clinical and follow-up imaging data. In no case did a CTP acquisition not cover the area supplied by the occlusion detected by the neuroradiologist.

[0099] Interpretation was performed in two separate sittings, 2 months apart, to eliminate memory and learning effects. Raw and reconstructed CTA and NECT data were made available at each sitting and viewed using a public domain DICOM viewer (Horos, v3.3.5, www.horosproject.org). Scans were anonymized and presented in random order. Interpreters were allowed to manipulate the provided NECT and CTA data (e.g., perform windowing and MIPS) as they would in routine clinical practice. In the first sitting, Tmax maps were provided for the first half of the patient cohort but not for the second half. This was reversed in the second sitting.

[0100] The readers were asked to: 1. Any region T max Review the Tmax map, if available, before interpreting the CTA to determine if a delay is present (Figure 9).

[0101] 2. Evaluate the CTA and record the presence and location of DVO(s) and mark the location on the thin-slice CTA images. If a Tmax map is available, the following approach has been proposed to identify the location of the occlusion on the CTA.

[0102] If present, narrow down the side, main vessel territory, and potentially occluded segments using a distribution Tmax delay that fits the arterial territory. Perform a focused search.

[0103] b. If this fails, gradually widen the search, as there is considerable anatomical variability in the territories supplied by the major intracranial arteries and their segments. Use a 5-point Likert scale to rank diagnostic confidence: 1 (very unlikely obstruction), 2 (unlikely obstruction), 3 (uncertain), 4 (likely obstruction), and 5 (very likely obstruction).

[0104] statistical analysis All statistical analyses were performed using MedCalc (MedCalc Statistical Software Version 17.2, 64 bit, Ostend, Belgium).

[0105] The diagnostic performance of each reader for detecting DVO on CTA was assessed against the reference standard (expert reading) using receiver operating characteristic (ROC) analysis. A true positive required that both the presence and location of DVO be correctly identified. The change in accuracy with the addition of Tmax was assessed by pairwise comparison of the area under the receiver operating characteristic (ROC) curve (AUC) using the De Long algorithm.

[0106] The added value of Tmax to diagnostic confidence was assessed by shift analysis using the Wilcoxon signed-rank test to assess the significance of any change. Fleiss's k statistic (kF) was used to assess inter-reader agreement. Cohen's k statistic (k) was used to determine agreement between each pair of readers.

[0107] Using the Wilcoxon signed-rank test, T max We determined whether the addition of CT scans significantly affected the time it took to interpret a CTA. Confidence intervals were calculated using bootstrapping with replacement on 10,000 samples. A level of 0.001 was interpreted as indicating significance for all tests, except for the confidence shift analysis, where a level of 0.05 was applied.

[0108] result CTA was analyzed from 140 patients (median age 73 years, IQR 64-83), of whom 77 were men and 70 had DVO (including 22 with proximal M2 occlusion). Details of patient baseline characteristics and vascular occlusion are provided in Figure 6.

[0109] Diagnostic accuracy The results of the ROC analysis for detecting DVO on CTA are shown in Figure 7. Across all readers, sensitivity and specificity increased with the addition of Tmax, and accuracy (as measured by AUC) increased significantly (p<0.001). The mean sensitivity for detecting DVO increased from 70.7% to 90.4% with the addition of Tmax, while mean specificity increased from 87.5% to 95.7%.

[0110] The analysis was repeated after excluding 22 patients with proximal M2-MCA occlusion (including 10 with occlusion of the proximal trunk of the dominant or codominant M2 segment, which could be considered a proximal vessel) (Figs. 8 and 18). The mean sensitivity for detecting DVO on CTA was significantly higher than that of Tmax (Fig. 1). max The sensitivity increased from 61.0% without to 86.5% without. Thus, the increase in sensitivity was greater when proximal M2 occlusion was included. After excluding 43 patients with either M2-MCA or P2-PCA occlusion (Figure 19), the diagnostic sensitivity for detecting more distal occlusion could be isolated, and the mean sensitivity was 1.0% for T max The sensitivity increased from 42.6% without to 81.5% with Tmax. Thus, the sensitivity of detecting DVO on CTA alone was much lower than when M2 and P2 occlusion were included. However, the increase in sensitivity, and therefore the increase in AUC, with the addition of Tmax was greater. In all three analyses, when Tmax was used, the AUC increased significantly for all readers (p<0.001), with the greatest improvement occurring when only the most "distal" DVO was considered (i.e., after exclusion of M2 or P2 occlusion).

[0111] Inter-reader agreement for CTA improved with the addition of Tmax, from kF = 0.61 (CI95 = 0.54-0.68) to kF = 0.79 (CI95 = 0.72-0.86). When CTAs were read with Tmax, there was greater agreement between pairs of readers than without Tmax (Figure 13).

[0112] Diagnostic reliability The increase in confidence with the addition of Tmax is shown in Figure 11 and Supplementary Material Figure 20. In patients deemed to have DVO relative to the reference standard, diagnostic confidence that obstruction was present increased significantly for all four readers (p<0.05). Each reader had fewer false negatives and deemed more patients to have very likely obstruction.

[0113] All readers were more confident in rejecting DVO on CTA when Tmax was used. The increase in confidence reached significance (p<0.05) for all readers except for residents, who had a large number of false positives (n=8). The number of patients considered to have a very low probability of obstruction increased for all readers.

[0114] CTA interpretation time CTAs were read significantly faster (p<0.001) with the use of Tmax (Figure 21), with median reading times 1.6 times faster for fellows and 3.3 times faster for scientists using Tmax. A boxplot of CTA reading times for two timed readers is shown (Figure 10). Patients were dichotomized into those with and without DVO relative to the reference standard. In both groups, reading times were significantly (p<0.001) shorter with Tmax than without Tmax, indicating that DVO was detected and rejected faster. The median time to detect M2 occlusion was shorter than for M3 and M4 occlusion, but this did not reach significance (Figure 22).

[0115] Post-mortem analysis False negatives and false positives are detailed in Figures 23 and 24, respectively, in the supplemental material. The number of false negatives decreased for all readers with the addition of Tmax. Four proximal M2 occlusions were missed by one or more readers on CTA without Tmax (Figure 11, images A and B). With Tmax, only one was missed by one reader. Fewer "distal" DVOs were missed on CTA when Tmax was used (Figure 11, images C and D). M4 segment MCA occlusions remained a challenge, but fellows and radiologists each missed five, even with Tmax. There were too few distal ACA occlusions in the cohort for meaningful analysis. All eight PCA occlusions distal to the P2 segment were detected on CTA with Tmax.

[0116] There were several false positives for DVO on CTA without Tmax. All were associated with small-caliber distal vessels, particularly branches and bifurcations. A few false positives on CTA with Tmax were also associated with small-caliber, poorly opacified distal vessels. Tmax delay was present in all but one of these cases but did not fit into the territory of the intracranial arteries. A recurrent cause of false positives in two inexperienced readers was Tmax delay in the deep white matter and outer watersheds ("borderline pattern") (Supplementary Material, Figure 15, Image E).

[0117] Fourteen patients had two DVOs (Figure 16). The Tmax benefit was greater with the two experienced readers. Both occlusions were detected and correctly identified in three additional patients by the radiologist and in seven additional patients by the scientist.

[0118] Consideration The added value of Tmax maps to the diagnostic ability to detect DVO on CTA was evaluated in this study. Diagnostic accuracy, confidence, and speed were shown to improve significantly with the addition of Tmax for readers with different levels of experience in interpreting stroke images. The beneficial effect of Tmax in aiding the detection of DVO on CTA was greater for more distal occlusions.

[0119] CTP is now widely included in acute stroke CT protocols. Its primary purpose is to identify patients with proximal arterial occlusions with salvageable brain tissue and who could therefore benefit from EVT. It also provides information about the macrovascular structure that can be exploited to improve the detection of vascular occlusions. In one previous study, which was not specifically designed to assess DVO and included only a small number of these distal occlusions, perfusion maps were shown to improve the detection of intracranial arterial occlusions. Another important distinction between this previous study and ours is that Tmax was not used.

[0120] Tmax is used in some EVT studies to identify salvageable ischemic penumbra and is now routinely available on most CTP post-processing software platforms. It can also be used to assess early reperfusion due to treatment. Tmax reflects the delay in contrast arrival within tissue relative to a proximal arterial reference point. This arterial reference point, called the arterial input function (AIF), is an essential part of deconvolution-based perfusion analysis. Intracranial arterial occlusion prolongs arterial transit time, and therefore Tmax, within the supplied territory. Time-to-peak (TTP), another time-based parameter that has been used to assess penumbra, is also prolonged in the presence of arterial transit delay. However, unlike Tmax, it is not obtained by deconvolution and therefore is not corrected for the shape of the contrast bolus. Therefore, Tmax is less sensitive than TTP to bolus delays proximal to the AIF and more specific to arterial transit delays between the AIF and tissue caused by occlusion. The distribution of Tmax delays can be used to narrow down the laterality, primary region, probable segment (e.g., M2 vs. M3), and likely location of arterial occlusion. A more focused search can then be performed. The alternative of systematically examining all cerebral arteries to the most distal identifiable level is very time-consuming and therefore not feasible under clinical time constraints. Clinical information regarding neurological deficits narrows the "search area" but is not always available or reliable. Tmax maps are objective and are consistently available when CTP is performed. To our knowledge, this is the first study to evaluate the utility of Tmax for detecting intracranial arterial occlusion.

[0121] As hypothesized, the diagnostic performance for detecting DVO on CTA significantly improved with the use of Tmax. The effect size was large enough to demonstrate a significant improvement using a p cutoff of 0.001, even with a sample size of 140, including 70 patients with DVO. Because information on neurological deficits was provided, the beneficial effect of Tmax was additive to any benefit provided by clinical records. The sensitivity of detecting DVO on CTA alone is likely lower; therefore, when reliable clinical information is unavailable, the improvement in performance with Tmax may be even greater.

[0122] Importantly, fewer proximal M2 occlusions were missed. EVT is increasingly being considered for patients with M2 occlusions because it may improve their functional outcomes. M2 occlusions are easily detected by experienced neuroradiologists but, as shown in this study, can be missed by residents and general radiologists. Thirty-five percent of M2 occlusions were missed during CTA evaluation in a previous study conducted at a primary stroke center. Residents read the majority of CTAs performed at stroke centers, but these scans are typically interpreted by general radiologists at primary stroke centers. Therefore, improving the detection of M2 occlusions by these less experienced readers has high clinical relevance for ensuring that patients do not miss potentially beneficial treatments.

[0123] T max The increased sensitivity, and therefore the benefit, of adding T was greater for more distal occlusions. The low sensitivity of CTA alone to detect M4 and distal ACA occlusions can be explained by the small caliber and large number of these vessels, making detection, even with accurate clinical records, equivalent to "finding a needle in a haystack." maxThe map narrowed the search area and increased the chance of finding the culprit occlusion on CTA. Despite the improved sensitivity, some distal anterior circulation occlusions were still missed by the readers. It is important to note that despite a clear regional Tmax delay suggesting DVO, the occlusion may not be clearly visible on CTA (due to the small internal diameter and insufficient opacification of the occluded artery). If an alternative cause cannot be found, it may be reasonable to diagnose a possible distal occlusion in these cases. Readers were better at detecting distal PCA occlusions than MCA occlusions. This may be due to the smaller spatial extent and fewer branches of the PCA compared with the MCA. Sensitivity for CTA with Tmax was imperfect and cannot be used to definitively exclude DVO. Notably, sensitivity was not related to the reader's experience level. Scientists and residents had the highest sensitivity both with and without Tmax. A possible explanation is that these readers had a more systematic approach to reading CTAs.

[0124] The specificity for detecting DVO on CTA exceeded 95% with the addition of Tmax for all readers except residents, which is important to ensure that futile and potentially harmful reperfusion is avoided in patients without DVO if the findings are used to guide treatment. Specificity was related to the level of experience in reading CTA and CTP. With more experience, advanced readers were able to distinguish between occlusions and poorly opacified distal vessels and to recognize and dismiss delayed Tmax that did not fit the vessel territory.

[0125] CTA interpretation was significantly faster with the addition of Tmax. As expected, DVO was detected more quickly due to the narrower search area, but occlusions were also rejected more quickly. Faster diagnosis not only expedites treatment but also improves workflow efficiency. The latter is important in clinical practice, especially in the context of a busy comprehensive stroke center. An important caveat to using Tmax to speed interpretation is that the reader must continue to systematically review the CTA for other incidental findings and avoid succumbing to the "streetlight effect."

[0126] Diagnostic confidence in the presence or absence of DVO on CTA increased with Tmax for all readers. One possible reason is that Tmax maps provide additional evidence to confirm or refute findings from CTA alone. Diagnostic confidence is greater when two separate tests indicate either the presence or absence of occlusion. Another factor likely contributing to the improvement in all metrics of diagnostic performance is the ease of interpretation of Tmax maps. They have a high contrast-to-noise ratio (CNR). While CBF and CBV have substantial gray-white matter contrast, Tmax values ​​do not vary with histology and have a rather flat contrast in normal brains. Therefore, even small or subtle areas of Tmax delay are highly visible and can be reliably and confidently detected and characterized. This, in turn, contributes to increased sensitivity when Tmax findings are used to guide the search for DVO on CTA. Conversely, maps without Tmax delay appear monochromatic, making it easier to interpret as normal and dismiss occlusion with high confidence. While DVO findings on CTA can be subtle due to the small diameter and poor opacification of distal vessels, findings with Tmax are more obvious and therefore less likely to be interpreted differently. Therefore, the use of Tmax has led to greater consistency in CTA interpretation and higher inter-reader agreement.

[0127] An alternative method that has been reported to improve the detection of DVO is wavelet transform angiography (waveletCTA), which is also obtained by post-processing of CTP data (21). Its clinical uptake is limited by the need for thin-slice CTP and specialized post-processing software that is not widely available. In contrast, most CTP post-processing software packages do not support T max Maps can be generated and generated from thick-slice CTP, making the use of Tmax feasible in routine clinical practice even in small peripheral centers.

[0128] An important limitation to note here is that CTP is recommended by guidelines only in the late (6–24 h) time frame. Because CTP incurs additional radiation dose and its role in patients within 6 h of stroke onset has not been established, not all stroke centers routinely perform CTP in the early time frame. CTP also has diagnostic utility beyond histological classification, aiding in the diagnosis of stroke and some pseudostrokes, such as migraine, and therefore is recommended in some centers.

[0129] A limitation of using Tmax maps to aid in the detection of DVO on CTA is that non-regional Tmax delays can result in false positives and lower specificity. Experienced readers dismissed DVO in these cases, suggesting that such errors can be avoided by training in Tmax map interpretation, specifically to distinguish Tmax delays in vascular territories from delays that are either transregional (e.g., due to migraine), artifactual, or "borderline" in distribution. "Borderline" Tmax delays arise from contrast bolus dispersion, for example, due to insufficient infusion, proximal arterial stenotic occlusive disease, and impaired cardiac output. This manifests as Tmax delays in arterial watershed regions. CTP acquisitions with limited brain coverage can result in false negatives if the areas supplied by occluded vessels are not included. This is avoided by whole-brain CTP acquisitions on modern multi-detector array CT scanners. When only limited coverage is feasible, false negatives can be avoided by targeting CTP acquisition to clinical findings (e.g., posterior circulation).

[0130] A limitation of this study is that only one CT scanner and CTP post-processing software package were used. This may limit the generalizability of the findings. The algorithms used by different post-processing software packages are variable, and most, if not all, use deconvolution-based post-processing, which is a requirement for obtaining Tmax. Furthermore, even among packages using deconvolution-based post-processing, there are substantial differences in the resulting perfusion parameters, resulting in variability in the quantification of the infarct core and penumbra. While variability is unlikely to affect qualitative assessment, differences in the display of parametric maps (e.g., color scale) may affect visual assessment. In turn, this may affect the diagnostic performance for the detection of regional Tmax delay, which warrants further evaluation in future studies.

[0131] A potential limitation of this study is that the prevalence of DVO in the study cohort (50%) was higher than in the population of stroke patients undergoing multimodal CT. The true prevalence of DVO is likely much lower, closer to the 14% observed in the cohort screened for this study. The use of a "balanced" sample may bias the absolute sensitivity and specificity for detecting DVO on CTA, both with and without Tmax. Therefore, these values ​​should be interpreted with caution. However, the primary objective of this study was to evaluate the relative change in diagnostic performance for detecting DVO on CTA when Tmax is added, rather than the absolute diagnostic performance.

[0132] Another potential limitation of this study is the choice of reference standard: expert reading of CTA instead of digital subtraction angiography (DSA), the reference standard for detecting intracranial vascular lesions. The use of CTA was justified because DVOs can recanalize or transfer between angiographic modalities, rendering DSA inaccurate as a reference standard, especially when thrombolysis is administered. The authors acknowledge that some very distal DVOs may have been missed by all readers, including expert readers. This affects the determination of absolute accuracy. However, this is less relevant to the comparative evaluation of diagnostic performance for CTA with and without Tmax, which was the purpose of this study.

[0133] conclusion When Tmax maps were used to focus the search for vascular occlusion, DVO was detected with greater accuracy, reliability, and speed on CTA. While the beneficial effect was greater for more distal occlusions, Tmax also aids in detecting M2 occlusions, which is clinically important because they are considered targets for EVT. Our findings demonstrate the significant added value of CTP beyond tissue classification, potentially benefiting the management of more patients than those with merely proximal arterial occlusions. By demonstrating that Tmax can be utilized to improve the detection of vascular occlusions by trainees and general practitioners, our findings encourage the inclusion of CTP in acute stroke imaging protocols at both comprehensive and primary stroke centers.

[0134] Supplementary Materials Part S1: Randomization procedure for selecting 70 "control" patients without vascular occlusion The 303 patients without vascular occlusion were entered into an Excel spreadsheet. A random number was generated for each patient using Excel's "randomize" function. The "data sort" function was then used to order the patients from lowest to highest according to their randomly assigned numbers. The first 70 patients were selected.

[0135] The same randomization procedure in Excel was used to determine the order in which the 140 patients included in this study were presented to the readers. The same order was maintained for the first and second readings.

[0136] Part S2: NECT, CTP, and CTA techniques and scan parameters Unenhanced CT was acquired in helical mode with the following parameters: 0.625 mm slice collimation, a spiral pitch factor of 0.283, a tube voltage of 120 kVp, and an image matrix of 512 × 512. Images were reconstructed with 1 mm overlapping sections using the UB kernel, and axial, coronal, and sagittal multiplanar reconstructions were performed with a slice thickness of 4 mm.

[0137] For CTP, 50 mL of nonionic contrast agent (350 mg iodine / mL, Iohexol Omnipaque 350, GE Healthcare, Wisconsin, USA) was intravenously injected, followed by 50 mL of normal saline. A flow rate of 5 mL / s was used for both injections. CTP acquisition parameters were as follows: 35 consecutive scans, 2.05 s average temporal resolution, 80 kVp tube voltage, 160 mA tube current, 500 ms gantry rotation time, 80 mm z-axis coverage, 1.5 mm slice collimation, and a 512 × 512 acquisition matrix. Images were reconstructed using an iterative reconstruction (iDose) factor of 4 with a 10 mm slice thickness.

[0138] CTA was performed using 80 mL of the same non-ionic contrast agent, injected intravenously at a rate of 5 mL / s, followed by 40 mL of saline at 6 mL / s. Acquisition parameters were as follows: craniocaudal coverage from the aortic arch to the vertex, 100 kVp tube voltage with tube current modulation, 0.625 mm slice collimation width, 512 × 512 image matrix, and 0.618 helical pitch factor. The following reconstruction parameters were used: iterative reconstruction (iDose) factor 5 and convolution kernel B. Contrast bolus triggering was performed at the aortic arch.

Claims

1. accessing, by a computing system including one or more processing devices and one or more memory devices, first training data including a first plurality of images of a brain of a first individual, the first plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique, and showing a first plurality of patterns of values ​​of one or more perfusion parameters or one or more diffusion parameters in the first individual in the absence of a brain abnormality; accessing, by the computing system, second training data including a second plurality of images of a brain of a second individual, the second plurality of images being captured using the perfusion-based imaging technique or the diffusion-based imaging technique and showing a second plurality of patterns of values ​​of the one or more perfusion parameters or the one or more diffusion parameters in the second individual in the presence of a brain abnormality; generating, by the computing system, one or more computational models based on the first training data and the second training data, the one or more computational models identifying abnormalities present in blood vessels of the individual's brain; accessing, by the computing system, one or more additional images of the individual's brain, wherein the one or more additional images are captured using a perfusion-based imaging technique or a diffusion-based imaging technique; determining, by the computing system, one or more patterns of the one or more perfusion parameters or the one or more diffusion parameters included in the one or more additional images; analyzing, by the computing system and using the one or more computational models, the one or more patterns in relation to the first plurality of patterns and the second plurality of patterns to determine a measure of the likelihood that an abnormality is present in the brain of the individual; generating, by the computing system and based on the measure of likelihood, a user interface including an indication that an abnormality is present in the brain of the individual.

2. The method of claim 1 , comprising determining a value of a perfusion parameter indicative of a time to maximum arrival of a tissue residue function for a voxel of the one or more images.

3. 3. The method of claim 2, further comprising performing, by the computing system, one or more deconvolution operations on contrast agent concentration curves of voxels included in the one or more images with respect to an arterial input function to generate the tissue residue function.

4. The method of claim 1 , comprising determining a value of a perfusion parameter indicative of a mean transit time of a contrast agent through the region of the brain.

5. The method of claim 1 , wherein the one or more additional images show the values ​​of the perfusion parameter or the values ​​of the diffusion parameter in multiple regions of the brain of the individual.

6. 2. The method of claim 1, wherein the first training data includes first classification data that labels the first plurality of images as being acquired from the first individual in whom no abnormalities are present, and the second training data includes second classification data that labels the second plurality of images as being acquired from the second individual in whom an abnormality is present in the blood vessels of the brain of the second individual.

7. 2. The method of claim 1, wherein the second training data includes one or more first additional images corresponding to a first vascular abnormality that disrupts blood flow to a first region of a brain in a first portion of the second individual and one or more second additional images corresponding to a second vascular abnormality that disrupts blood flow to a second region of a brain in a second portion of the second individual, wherein the first blood vessel has a different location than the second blood vessel.

8. 8. The method of claim 7, wherein the second training data includes first additional classification data that labels the one or more first additional images as being associated with the first vascular abnormality, and second additional classification data that labels the one or more second additional images as being associated with the second vascular abnormality.

9. 10. The method of claim 1, comprising training, by the computing system and using the first training data and the second training data, one or more convolutional neural networks to identify when abnormalities are present in the individual's brain based on patterns in images obtained from one or more perfusion-based imaging techniques or one or more diffusion-based imaging techniques.

10. The method of claim 1 , wherein the one or more perfusion-based imaging techniques include a computed tomography perfusion-based imaging technique or a magnetic resonance perfusion-based imaging technique.

11. 2. The method of claim 1, comprising generating, by the computing system and based on the likelihood that the abnormality is present in the brain of the individual, a text-based identifier of the blood vessel in which the abnormality is present.

12. 2. The method of claim 1, wherein the abnormality present in a blood vessel of the brain of the individual in which the abnormality is present is a distal blood vessel of the human brain, and the distal blood vessel comprises at least one of the A2-A5 segments of the anterior cerebral artery, the M2-M4 segments of the middle cerebral artery, the P2-P4 segments of the posterior cerebral artery, the posterior inferior cerebellar artery, the anterior inferior cerebellar artery, or the superior cerebellar artery.

13. one or more hardware processors; one or more non-transitory computer-readable storage media containing computer-readable instructions, the computer-readable instructions, when executed by the one or more hardware processors, causing the one or more hardware processors to: accessing first training data comprising a first plurality of images of an individual's brain, the first plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique, and showing a first plurality of patterns of values ​​of one or more perfusion parameters or one or more diffusion parameters in an individual in the absence of a brain abnormality; accessing second training data comprising a second plurality of images of the individual's brain, the second plurality of images being captured using the perfusion-based imaging technique or the diffusion-based imaging technique and showing a second plurality of patterns of values ​​of the one or more perfusion parameters or the one or more diffusion parameters in an individual in the presence of a brain abnormality; generating one or more computational models based on the first training data and the second training data, the one or more computational models identifying abnormalities present in blood vessels of the individual's brain; accessing one or more additional images of the individual's brain, wherein the one or more additional images are captured using a perfusion-based imaging technique or a diffusion-based imaging technique; determining one or more patterns of the one or more perfusion parameters or the one or more diffusion parameters contained in the one or more additional images; analyzing the one or more patterns in relation to the first plurality of patterns and the second plurality of patterns using the one or more computational models to determine a measure of the likelihood that an abnormality is present in the brain of the individual; generating a user interface including an indication that an abnormality is present in the brain of the individual based on the measure of likelihood.

14. The one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to: acquiring additional image data corresponding to one or more additional images captured using a computed tomography angiography imaging technique; determining a portion of the additional image data corresponding to a region of the brain of the individual where the abnormality is present; and analyzing the portion of the additional image data to identify blood vessels supplying the region; and determining that the abnormality is present in the blood vessel.

15. The one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to: determining intensity values ​​of voxels corresponding to a plurality of candidate vessels supplying blood to the region of the brain of the individual based on the additional image data, wherein the vessel is included in the plurality of candidate vessels; and determining that the abnormality is present in the blood vessel based on the intensity value.

16. The one or more non-transitory computer-readable storage media include additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to:

14. The system of claim 13, further comprising: identifying regions of the brain of the individual corresponding to voxels having values ​​of the perfusion parameter or the diffusion parameter that are at least a threshold value.

17. One or more non-transitory computer-readable media storing computer-readable instructions that, when executed by one or more processing devices, cause the one or more processing devices to: accessing first training data comprising a first plurality of images of an individual's brain, the first plurality of images being captured using a perfusion-based imaging technique or a diffusion-based imaging technique, and showing a first plurality of patterns of values ​​of one or more perfusion parameters or one or more diffusion parameters in an individual in the absence of a brain abnormality; accessing second training data comprising a second plurality of images of the individual's brain, the second plurality of images being captured using the perfusion-based imaging technique or the diffusion-based imaging technique and showing a second plurality of patterns of values ​​of the one or more perfusion parameters or the one or more diffusion parameters in an individual in the presence of a brain abnormality; generating one or more computational models based on the first training data and the second training data, the one or more computational models identifying abnormalities present in blood vessels of the individual's brain; accessing one or more additional images of the individual's brain, wherein the one or more additional images are captured using a perfusion-based imaging technique or a diffusion-based imaging technique; determining one or more patterns of the one or more perfusion parameters or the one or more diffusion parameters contained in the one or more additional images; analyzing the one or more patterns in relation to the first plurality of patterns and the second plurality of patterns using the one or more computational models to determine a measure of the likelihood that an abnormality is present in the brain of the individual; and generating a user interface including an indication that an abnormality is present in the brain of the individual based on the measure of likelihood.

18. operations include generating first additional user interface data corresponding to a first additional user interface indicative of values ​​of the perfusion parameter for voxels of the one or more additional images; 18. The one or more non-transitory computer-readable media of claim 17, wherein a first range of values ​​of the perfusion parameter is displayed as a first color in the first additional user interface and a second range of values ​​of the perfusion parameter is displayed as a second color in the first additional user interface.

19. 20. The one or more non-transitory computer-readable media of claim 18, wherein the operations include generating second additional user interface data corresponding to a second additional user interface showing intensity values ​​of voxels, the second additional user interface showing blood vessels of the brain of the individual.

20. 20. The one or more non-transitory computer-readable media of claim 19, wherein the first additional user interface and the second additional user interface are configured to be displayed together and the first additional user interface data and the second additional user interface data are configured to be rendered to generate a combined user interface comprising the first additional user interface and the second additional user interface, and wherein a region of the brain of the individual is highlighted in the combined user interface comprising the first additional user interface and the second additional user interface.

Images