A device and method for examining acute large vessel occlusion using binocular deflection.

Abstract

To advance an inspection capable of achieving early detection of large vessel occlusion (LVO).SOLUTION: A large vessel occlusion inspection device comprises: an image input unit that receives non-contrast CT images being a plurality of continuous sequence images that includes an anatomical object forming the upper portion of the brain and both eyes; a preprocessing unit that preprocesses the non-contrast CT images on the basis of rigid registration and affine registration; an image processing unit that performs non-rigid registration of an atlas with respect to the preprocessed non-contrast CT images, normalizes each cell in the atlas-registered non-contrast CT images to have values of 0 or 1, and extracts a region of interest from the non-contrast CT images by combining each cell of the atlas-registered non-contrast CT image through inverse transformation; and a determination unit that identifies deviation between both eyes from the non-contrast CT images from which the region of interest has been extracted, advances the LVO inspection on the basis of the deviation between both eyes, determines whether the patient is LVO-positive or negative, classifies the LVO-suspected case of the patient who has been determined as positive, and notifies medical professionals of the classified LVO-suspected case of the patient.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an acute large vessel occlusion inspection device and method using binocular deviation, and more particularly, to an acute large vessel occlusion inspection device and method using binocular deviation (Eyeball deviation), which is a biomarker identified from non-contrast computed tomography (NCCT) images of a patient, to perform an inspection of acute large vessel occlusion (Large Vessel Occlusion, LVO).

Background Art

[0002] Time is a typical factor that most affects the prognosis of patients with acute large vessel occlusion (hereinafter, "LVO"). Since 1998, in the case of acute ischemic stroke, there has been no other large vessel occlusion (Large Vessel Occlusion, LVO) reperfusion treatment method other than intravenous tissue plasminogen activator (tPA) within 4.5 hours after symptom onset. However, through the "MR CLEAN" cohort study in 2015, the treatment of LVO patients has become a turning point because it has been proven through various studies that reperfusion by mechanical thrombectomy improves the prognosis of LVO patients.

[0003] In addition, according to various research results, there is a research result that endovascular treatment is possible up to 16 hours or 24 hours for patients who meet various conditions such as the location of LVO, the stroke scale (NIHSS), and the lesion core size.

[0004] However, in reality, in order to consider the prognosis of endovascular treatment, it is generally recommended to perform treatment within 6 hours. In the case of LVO patients, it is known that appropriate treatment through such rapid diagnosis can save the patient's life and enable the patient to quickly return to daily life.

[0005] Furthermore, numerous studies have concluded that delays due to transfers to other hospitals still result in patients not receiving endovascular treatment within a reasonable timeframe, thus reducing their chances of a favorable prognosis.

[0006] On the other hand, when a patient comes to the hospital with symptoms suspected to be LVO, it is customary to first perform a non-contrast CT scan after analyzing the patient's basic condition. Generally, the first symptom confirmed by non-contrast CT is the presence or absence of cerebral hemorrhage. If no evidence of cerebral hemorrhage is found, the patient will be scanned using methods such as angiography (CT angiography, CTA), CT perfusion (CTP), or magnetic resonance diffusion imaging (MR DWI).

[0007] Such angiography (CTA), CT perfusion imaging (CTP), or magnetic resonance diffusion imaging (MR DWI) may not be feasible depending on the clinical environment or the patient's condition, such as contrast agent allergies. In contrast, non-contrast CT imaging has few limitations to consider and offers the advantage of being relatively easier to scan patients with compared to other scanning methods.

[0008] However, non-contrast CT images, compared to other scanning methods, make it difficult to clearly identify symptoms in LVO patients, leading to significant differences in symptom identification results depending on the subjective experience of radiologists. This non-objective symptom identification result is the reason why inter-influencer differences are a problem in ASPECTS (Alberta Stroke Program Early Computed Tomographic Score) scores calculated based on non-contrast CT.

[0009] Previous studies have shown that the average time from arrival at the hospital to scanning with non-contrast CT (average time of door-to-CT) and the average time from arrival to scanning using angiography (average time of door-to-CTA) were 13.4 ± 1.8 minutes and 75.5 ± 44.5 minutes, respectively.

[0010] In other words, if medical professionals can quickly analyze the condition of LVO patients, it is expected that the examination and treatment time for LVO patients will be shortened in various clinical settings. Furthermore, if LVO patients can be quickly identified through non-contrast CT images, which require less time before additional patient scans, it is expected that early identification of LVO patients will be possible not only in third-tier hospitals (advanced general hospitals) but also in primary and secondary hospitals where only non-contrast CT image scans are possible.

[0011] In response to these needs, there is a need for research and development of acute large vessel occlusion testing devices and methods that can perform tests to determine whether a patient has positive or negative LVO (long vessel occlusion) based on non-contrast CT images, and that can provide notification to medical professionals if a patient is determined to be LVO-positive, thereby facilitating the rapid progression of the patient's treatment. [Prior art documents] [Patent Documents]

[0012] [Patent Document 1] Korean Registered Patent Publication No. 10-1702267 (Registered January 25, 2017) [Overview of the project] [Problems that the invention aims to solve]

[0013] Therefore, the present invention was devised to solve the above-mentioned problems, and the object of the present invention is to provide an acute great vessel occlusion examination device and method utilizing binocular deflection, which is a biomarker identified from non-contrast CT images of a patient, that can be used to perform LVO examination.

[0014] Furthermore, an object of the present invention is to provide an acute large vessel occlusion testing device and method that utilizes binocular deflection to provide notification to medical professionals when a patient is determined to be LVO-positive based on the results of the LVO test, thereby guiding the rapid progress of the patient's treatment.

[0015] Furthermore, the object of the present invention is to provide an acute large vessel occlusion examination device and method that utilizes binocular deflection, which can determine whether a patient is LVO positive or negative not only through binocular deflection but also through multiple biomarkers identified from the patient's non-contrast CT images.

[0016] Specifically, the object of the present invention is to provide an acute great vessel occlusion examination device and method that utilizes binocular deflection, which can identify dense middle cerebral artery signs (DMS), early ischemic changes (EIC), and binocular deflection from a patient's non-contrast CT images, and determine whether the patient is LVO-positive or negative based on the identified multiple biomarkers.

[0017] However, the technical problems that the present invention aims to solve are not limited to those mentioned above, and other technical problems not mentioned can be clearly understood by a person with ordinary skill in the art to which the present invention pertains from the following description. [Means for solving the problem]

[0018] An acute large vessel occlusion examination apparatus according to one embodiment of the present invention for achieving the above objectives comprises: an image input unit that receives a sequence of non-contrast CT images consisting of at least a plurality of consecutive non-contrast CT images including the upper part of the brain corresponding to the entire midbrain to the cerebral cortex and an anatomical entity that constitutes both eyes; a pre-processing unit that pre-processes the non-contrast CT images based on rigid body alignment and affine alignment; an image processing unit that non-rigidly aligns at least one atlas to the non-contrast CT images pre-processed by the pre-processing unit, normalizes the atlas-aligned non-contrast CT images so that each cell has a value of 0 or 1, and extracts a region of interest from the reconstructed non-contrast CT image by combining each cell of the atlas-aligned non-contrast CT image through inverse transform; and a unit that extracts binocular bias (eyeball) from the non-contrast CT image from which the region of interest has been extracted. It may include a judgment unit that identifies deviation, performs an LVO test based on the binocular deviation to determine whether the patient is LVO positive or negative, classifies suspected LVO cases of patients determined to be LVO positive through the LVO test, and then provides notification to medical professionals.

[0019] The acute large vessel occlusion examination method according to one embodiment of the present invention, performed by the aforementioned acute large vessel occlusion examination device, comprises: a) a first step in which the video input unit receives input of a non-contrast CT image, which is a sequence of images consisting of at least a plurality of consecutive non-contrast CT images including the upper part of the brain corresponding to the entire midbrain to the cerebral cortex and an anatomical individual that constitutes both eyes; b) a second step in which the pre-processing unit pre-processes the non-contrast CT image based on rigid body alignment and affine alignment; c) a third step in which the video processing unit non-rigidly aligns at least one atlas to the non-contrast CT image pre-processed by the pre-processing unit, normalizes the atlas-aligned non-contrast CT image so that each cell has a value of 0 or 1, and extracts a region of interest from the reconstructed non-contrast CT image by combining each cell of the atlas-aligned non-contrast CT image through inverse transformation; and d) a judgment unit determines the binocular bias (eyeball) from the non-contrast CT image from which the region of interest has been extracted. A fourth step may include: identifying the deviation; performing an LVO test based on the binocular deviation to determine whether the patient is LVO positive or negative; classifying suspected LVO cases in patients determined to be LVO positive through the LVO test; and then providing notification to a medical professional. [Effects of the Invention]

[0020] According to one embodiment of the present invention, an LVO test is performed to determine whether a patient has positive or negative LVO based on non-contrast CT images, and if the patient is determined to be LVO positive, medical professionals are notified to facilitate the rapid progression of the patient's treatment.

[0021] According to one embodiment of the present invention, by utilizing dense middle cerebral artery symbols, early ischemic changes, and binocular bias identified from the patient's non-contrast CT images, it is possible to improve the limitations in determining whether a patient is LVO-positive or negative through their respective biomarkers.

[0022] However, the effects obtained by the present invention are not limited to the effects mentioned above, and other effects not mentioned may be clearly understood by those having ordinary knowledge in the technical field to which the present invention pertains from the following description.

Brief Description of Drawings

[0023] [Figure 1] It is a block configuration diagram of an acute large vessel occlusion inspection device according to an embodiment of the present invention. [Figure 2] It is a block configuration diagram of a classification model used for model learning to embody the determination unit illustrated in FIG. 1. [Figure 3] It is a flowchart of an acute large vessel occlusion inspection method according to an embodiment of the present invention. [Figure 4A] It is a flowchart illustrating the detailed process of the patient's LVO data generation step illustrated in FIG. 3. [Figure 4B] It is a flowchart illustrating the detailed process of the patient's LVO data generation step illustrated in FIG. 3. [Figure 5] It is a flowchart illustrating the detailed process of the patient's LVO data generation step illustrated in FIG. 3. [Figure 6] It is a flowchart illustrating the detailed process of the patient's LVO data generation step illustrated in FIG. 3.

Modes for Carrying Out the Invention

[0024] In the following, embodiments of the present invention will be described in detail, with reference to the attached drawings, so that they can be easily implemented by a person with ordinary skill in the art to which the present invention pertains. However, since the description of the present invention is merely an example for structural or functional explanation, the scope of the present invention should not be construed as being limited by the embodiments described herein. That is, since embodiments can be modified in various ways and can take various forms, the scope of the present invention should be understood to include equivalents that can realize the technical idea. Furthermore, since the purposes or effects presented in the present invention do not mean that a particular embodiment should include all of them or only such effects, the scope of the present invention should not be understood as being limited by them.

[0025] The meanings of the terms used in this invention should be understood as follows:

[0026] Terms such as "first" and "second" are used to distinguish one component from another, and these terms should not limit the scope of rights. For example, the first component may be named the second component, and similarly, the second component may be named the first component. When it is mentioned that one component is "linked" to another component, it should be understood that it may be directly linked to the other component, or that other components may exist in between. Conversely, when it is mentioned that one component is "directly linked" to another component, it should be understood that there are no other components in between. Other expressions describing the relationship between components, namely "between" and "immediately between," or "adjacent to" and "directly adjacent to," should be interpreted similarly.

[0027] A singular expression should be understood to include multiple expressions unless the context clearly indicates otherwise, and terms such as “includes” or “possesses” should be understood to indicate the existence of features, numbers, steps, actions, components, parts, or combinations thereof that are implemented, without prejudice to the existence or possibility of adding one or more different features, numbers, steps, actions, components, parts, or combinations thereof.

[0028] All terms used herein have the same meaning as that generally understood by a person of ordinary skill in the art to which this invention pertains, unless otherwise defined. Terms as defined in commonly used dictionaries should be interpreted in accordance with their meaning in the context of the relevant art and should not be interpreted in an ideal or overly formal sense unless explicitly defined herein.

[0029] Acute large vessel occlusion testing device

[0030] In the present invention, non-contrast computed tomography (NCCT) refers to computed tomography images obtained without the use of a contrast agent solution. More specifically, it is preferable that the NCCT images consist of a sequence of multiple consecutive non-contrast CT images that include at least the upper part of the brain, such as the midbrain to the entire cerebral cortex of the subject (patient), and the anatomical entity that makes up both eyes.

[0031] Furthermore, non-contrast CT images can be acquired from the imaging equipment with a slice thickness of 5 mm or less so that the judgment unit 14, described later, can extract dense middle cerebral artery signs (Dense MCA Sign, DMS).

[0032] The acute large vessel occlusion examination device 10 of the present invention can be implemented with radiologic computer-aided classification and notification (CADt) software designed to enable image analysis of non-contrast CT images in order to display suspected occlusion findings of the internal carotid artery (ICA) or middle cerebral artery (MCA) in the non-contrast CT images, thereby supporting the classification of workflows for hospital networks and their users, the radiologists.

[0033] In the following, an acute large vessel occlusion examination device 10 according to one embodiment of the present invention, which can be implemented with software as described above, will be described in detail with reference to the attached drawings.

[0034] Figure 1 is a block diagram of an acute large vessel occlusion testing device according to one embodiment of the present invention.

[0035] Referring to Figure 1, the acute large vessel occlusion examination device 10 includes a video input unit 11, a pre-processing unit 12, a video processing unit 13, and a judgment unit 14.

[0036] The video input unit 11 receives input of non-contrast CT images from imaging equipment that captures non-contrast CT images, consisting of multiple non-contrast CT images that include at least the upper part of the brain, such as the midbrain to the entire cerebral cortex, and both eyes, which are the subject of imaging in which the presence or absence of LVO positivity and negativity of the patient is to be determined.

[0037] Such a video input unit 11 is not limited to acquiring multiple non-contrast CT images from imaging equipment, but can also receive non-contrast CT images consisting of multiple non-contrast CT images using a data set input method.

[0038] The preprocessing unit 12 receives non-contrast CT video from the video input unit 11 and preprocesses the multiple non-contrast CT images that make up the received non-contrast CT video.

[0039] More specifically, the preprocessing unit 12 can generate a first non-contrast-enhanced CT image, which is a reference image, from a plurality of non-contrast-enhanced CT images, and extract feature points (landmark points) from the second non-contrast-enhanced CT image, which is the remaining floating non-contrast-enhanced CT image after removing the first non-contrast-enhanced CT image.

[0040] In the present invention, the characteristic features of the first non-contrast CT image and the second non-contrast CT image are preferably extracted from anatomical individuals within the non-contrast CT image, such as the midbrain to the entire cerebral cortex, and from individuals forming both eyes, and there are preferably at least two such individuals.

[0041] Furthermore, the preprocessing unit 12 completes the preprocessing of the first and second non-contrast CT images by matching the feature points extracted based on rigid registration and affine registration to match the second non-contrast CT image with the first non-contrast CT image and generate a preprocessed non-contrast CT image, and then transmits the preprocessed non-contrast CT image to the image processing unit 13.

[0042] In this invention, rigid body matching and affine matching refer to the process of globally matching the first non-contrast CT image and the second non-contrast CT image, treating the anatomical individual as a rigid body for matching.

[0043] The image processing unit 13 receives the pre-processed non-contrast CT image from the pre-processing unit 12, synchronizes the pre-processed non-contrast CT image with a standard atlas template, and performs the process of registering the pre-processed non-contrast CT image in the space of the standard atlas template.

[0044] In this context, "atlas template" refers to a new tool for Voyager, a cloud-based software for analyzing CT scan data.

[0045] Furthermore, the image processing unit 13 non-rigid-body aligns at least one atlas with the pre-processed non-contrast CT image registered in the space of the standard atlas template so that the atlas is as similar as possible to the pre-processed non-contrast CT image.

[0046] In the present invention, the atlas may be a pair of images of an anatomical individual, such as the upper part of the brain, including the midbrain and the entire cerebral cortex, and an individual comprising both eyes, and can be used to divide a pre-processed non-contrast CT image into anatomical individuals.

[0047] In the present invention, the video processing unit 13 can perform non-rigid body matching of the atlas based on multi-scaled and phase-based registration.

[0048] Furthermore, the atlas can be modified to correspond to the morphology of non-contrast-enhanced CT images preprocessed by non-rigid body alignment performed by the image processing unit 13, and can be aligned with groups of anatomical individuals.

[0049] The image processing unit 13 then adds the atlas alignment results to the pre-processed non-contrast CT image, thereby forming an atlas-aligned non-contrast CT image on a standard atlas template.

[0050] In the present invention, the atlas-matched non-contrast CT image may be a probability map that defines the probability that each cell constituting the pre-processed non-contrast CT image represents the volume of an anatomical individual.

[0051] Furthermore, the image processing unit 13 normalizes the atlas-aligned non-contrast CT image, so that each cell in the atlas-aligned non-contrast CT image has a value of either 0 or 1 through normalization.

[0052] In this case, a value of 1 is assigned when the volume of an anatomical individual is included in part or all of the cell's area, and a value of 0 is assigned when the volume of an anatomical individual is not included in the entire cell's area.

[0053] In this invention, the non-contrast CT image, which is aligned with the atlas, is divided not only into the vascular regions of the internal carotid artery (ICA) and middle cerebral artery (MCA) through normalization performed by the image processing unit 13, but also into the cortical and subcortical regions affected by the internal carotid artery and middle cerebral artery.

[0054] The image processing unit 13 then performs an inverse transform and normalization process to combine the cells of the non-contrast CT image that have been aligned with the atlas, which have values ​​of 0 or 1, so that the non-contrast CT image is reconstructed. Through this process, the region of interest (ROI) is extracted from the reconstructed non-contrast CT image.

[0055] In the present invention, the region of interest refers to a set of cells in a non-contrast-enhanced CT image to which an atlas assigned a value of 1 through a normalization process by the image processing unit 13 has been aligned.

[0056] Furthermore, the image processing unit 13 transmits the non-contrast CT image from which the region of interest has been extracted to the judgment unit 14.

[0057] The judgment unit 14 receives non-contrast CT images from the image processing unit 13 from which regions of interest have been extracted, and performs an LVO examination to determine whether a patient is LVO positive or negative based on the analysis of the received non-contrast CT images from which regions of interest have been extracted. The judgment unit 14 then uses an artificial intelligence model (AI Model) to learn in order to classify suspected LVO cases of patients who have been determined to be LVO positive through the LVO examination.

[0058] In the present invention, the judgment unit 14 can perform learning using an artificial intelligence model by utilizing the classification models 14a and 14b shown in Figure 2.

[0059] Figure 2 is a block diagram of the classification model used for model learning to realize the decision unit shown in Figure 1.

[0060] Referring to Figure 2, the classification model for model learning of the decision unit 14 includes a first classification model 14a and a second classification model 14b.

[0061] The first classification model 14a is an artificial intelligence model that trains the decision unit 14 to perform LVO examinations to determine whether a patient is LVO positive or negative, based on the analysis of non-contrast CT images from which regions of interest have been extracted.

[0062] As a specific example, the first classification model 14a can be implemented using 2D and 3D convolutional neural networks (CNNs), and the 2D convolutional neural network may be EfficientNet.

[0063] Since EfficientNet is typically a type of artificial intelligence model belonging to the convolutional neural network category, a detailed explanation of it will be omitted for convenience.

[0064] The second classification model 14b is an artificial intelligence model that trains the decision unit 14 to classify suspected cases of LVO in patients who have been determined to be LVO positive.

[0065] As a specific example, the second classification model 14b can be implemented using a recurrent neural network (RNN)-based bidirectional long-term memory (Bi-LSTM) model.

[0066] Since this type of bidirectional long-term short-term memory model is typically a type of artificial intelligence model belonging to recursive neural networks, a detailed explanation of it will be omitted for convenience.

[0067] On the other hand, the first classification model 14a is an artificial intelligence model trained to perform an LVO test that determines whether a patient is LVO positive or negative, based on the judgment unit 14 identifying the characteristics of biomarkers, which are indicators of LVO positivity, from non-contrast-enhanced CT images from which the region of interest has been extracted.

[0068] In this invention, the biomarkers include the dense middle cerebral artery sign (DMS), early ischemic changes (EIC), and eyeball deviation, which are indicators of LVO positivity.

[0069] Here, dense middle cerebral artery symbols (DMS) represent a phenomenon where an artery is blocked by a thrombus, and the blocked cerebral arteries appear excessively on non-contrast CT images. Such dense middle cerebral artery symbols (DMS) are a representative biomarker that explains occlusion of the middle cerebral artery (MCA) on non-contrast CT images, but their presence or absence is limited to the slice thickness setting of the scanned non-contrast CT image. Typically, dense middle cerebral artery symbols (DMS) have a higher detection probability as the slice thickness of the non-contrast CT image decreases.

[0070] Furthermore, early ischemic changes (EIC) are a term referring to ischemic changes that occur in brain tissue after large vessel occlusion (LVO), and are identified in non-contrast CT images taken early after large vessel occlusion. In this case, if the middle cerebral artery (MCA) becomes blocked by a thrombus, early ischemic changes (EIC) may also be identified in the lower brain, such as in catalytic infection or the basal ganglia, after the stroke has occurred. Due to this phenomenon of early ischemic changes (EIC), the Alberta Stroke Program Early CT Score M1-M6 (ASPECTS M1-M6), a representative scoring system for examining areas where lesions appear, such as the middle cerebral artery (MCA) and the basal ganglia, was developed.

[0071] Regarding bilateral eye deviation, acute ischemic stroke is known as a single-sided disease, where symptoms of ischemic changes appear in the opposite hemisphere to where LVO occurred. Such ischemic changes can be confirmed by palsy, but in LVO patients, the eyes are characterized by deviation towards the hemisphere where LVO occurred. In particular, recent studies have reported classifying LVO patients mainly based on eye deviation observed in CT images. When this was used as a single index to classify LVO patients, the sensitivity was confirmed to be 71% and the specificity 77.5%.

[0072] The biomarker of the present invention is a representative indicator used to determine whether a patient is LVO-positive or LVO-negative; however, when used independently, it has limitations in determining whether a patient is LVO-positive or LVO-negative.

[0073] In the present invention, it is preferable that the judgment unit 14 is model-learned through a first classification model 14a to identify the biomarkers dense middle cerebral artery symbol (DMS), early ischemic change (EIC), and binocular deviation (Eyeball deviation) from non-contrast CT images from which the region of interest has been extracted, in order to improve the limitations of each biomarker and determine whether a patient is LVO positive or negative.

[0074] Furthermore, if the determination unit 14 determines that the patient is LVO positive through the LVO test, it notifies the medical professional treating the patient to ensure that the patient's treatment proceeds quickly.

[0075] In this invention, the notification that the decision unit 14 provides to the medical professional can be provided to the medical professional via a server or application that is accessible via a terminal (e.g., smartphone, PC, tablet, etc.).

[0076] Furthermore, the information provided by the judgment unit 14 to medical professionals includes suspected cases of LVO in patients identified as positive for LVO through LVO testing, and compressed non-contrast CT images of patients that can be previewed so that medical professionals can refer to the patient's condition during the treatment process.

[0077] Furthermore, suspected LVO cases in patients can be generated and provided to medical professionals in a data format that includes binocular deviation, LVO probability values ​​for each hemisphere, and LVO discrimination results, as illustrated in Figure 3.

[0078] Methods for examining acute large vessel occlusion

[0079] In the following, the process of the acute large vessel occlusion examination method S10 according to one embodiment of the present invention, which is performed by the acute large vessel occlusion examination device 10, will be described in detail.

[0080] Furthermore, it is preferable that the acute large vessel occlusion testing device 10 is learned and trained to automatically perform the acute large vessel occlusion testing method S10 described later, and so the learning and training processes of the acute large vessel occlusion testing method S10 and the acute large vessel occlusion testing device 10 may be the same.

[0081] Figure 3 is a flowchart of a method for examining acute large vessel occlusion according to one embodiment of the present invention.

[0082] Referring to Figure 3, the acute large vessel occlusion examination method S10 includes a video input step S11, a video processing step S12, biomarker extraction steps S13-S15, LVO probability value calculation step S16, and LVO examination and suspected case provision step S17.

[0083] In the video input step S11, the video input unit 11 can receive non-contrast CT images captured with a slide thickness of 5 mm or less through the imaging equipment or data set input method.

[0084] In the image processing step S12, the preprocessing unit 12 can preprocess multiple non-contrast CT images that constitute the non-contrast CT image received from the image input unit 11.

[0085] In this case, the preprocessing unit 12 can preprocess (primary matching) the non-contrast CT image based on rigid body matching and affine matching in the image processing step S12.

[0086] Furthermore, in the image processing step S12, the image processing unit 13 performs non-rigid body matching of at least one atlas with respect to the non-contrast CT image preprocessed in the space of the standard atlas template, based on multi-scaled and phase-based registration, so that the atlas becomes as similar as possible to the preprocessed non-contrast CT image.

[0087] Furthermore, the image processing unit 13 can normalize the atlas-aligned non-contrast CT image in the image processing step S12 such that each cell in the atlas-aligned non-contrast CT image has a value of 0 or 1.

[0088] Furthermore, in the image processing step S12, the image processing unit 13 combines each cell of the non-contrast-enhanced CT image, which has an atlas with a value of 0 or 1, through a normalization process via inverse transformation, thereby reconstructing the non-contrast-enhanced CT image. Through this, it is possible to extract a region of interest (ROI) from the reconstructed non-contrast-enhanced CT image.

[0089] In biomarker identification steps S13 to S15, the decision unit 14 can identify each biomarker based on a convolutional neural network (CNN) when performing LVO testing.

[0090] The biomarker identification steps S13 to S15 may include a first biomarker identification step S13, a second biomarker identification step S14, and a third biomarker identification step S15.

[0091] In the first biomarker identification step S13, the decision unit 14 can use a pre-trained 3D CNN model to separate the two eyes (eyeballs) from the non-contrast CT image from which the region of interest has been extracted, and classify the deflection of the separated eyes into one of three classes: forward, left, or right.

[0092] At this time, the judgment unit 14 calculates probability values ​​for three classes (forward, left, and right) for each of the left and right eyes in the first biomarker identification step S13, with values ​​between 0 and 1. Based on identifying the direction of the class with the highest probability value among the three classes as the direction of the left and right eyes, the deviation of both eyes can be classified.

[0093] In the second biomarker identification step S14, the decision unit 14 can use a pre-trained first 2D CNN model to examine the region from the longitudinal section of the internal carotid artery (ICA) up to the M1 segment of the middle cerebral artery (MCA) to identify dense middle cerebral artery symbols (DMS).

[0094] In this case, it is preferable that the determination unit 14 identifies the dense middle cerebral artery symbols (DMS) in the left hemisphere and the right hemisphere, respectively, in the second biomarker identification step S14.

[0095] In the third biomarker identification step S15, the decision unit 14 can use a pre-trained first 2D CNN model to examine the region where occlusion is expected to occur in the longitudinal section of the internal carotid artery (ICA) up to the M2 segment of the middle cerebral artery (MCA), thereby identifying early ischemic changes (EIC).

[0096] In this case, it is preferable that the judgment unit 14 identifies early ischemic changes (EIC) in the left hemisphere and the right hemisphere, respectively, in the third biomarker identification step S15.

[0097] In the LVO probability value calculation step S16, the determination unit 14 concatenates the features of the first 2D CNN model and the second 2D CNN model, inputs the concatenated features into the RNN model, and determines whether the LVO discrimination result for each hemisphere is positive or negative.

[0098] Furthermore, the determination unit 14 can calculate the probability values ​​of LVO for each hemisphere in the left and right hemispheres in the LVO probability value calculation step S16, based on the LVO discrimination results for each hemisphere, with the values ​​ranging from 0 to 1.

[0099] In the LVO testing and suspected case provision step S17, the judgment unit 14 extracts the characteristics of the biomarkers identified from the biomarker extraction steps S13 to S15 and the LVO probability value calculation step S16 to perform a test to determine whether or not LVO is positive or negative. After generating suspected LVO cases for patients determined to be LVO positive through the test, these cases can be provided to medical professionals.

[0100] In this case, the LVO suspected case is generated in a data format that includes binocular deviation, LVO probability values ​​for each hemisphere, and LVO discrimination results, but the LVO discrimination results can be calculated differently depending on the binocular deviation and LVO probability values ​​for each hemisphere in the LVO examination and suspected case provision step S17.

[0101] Figures 4A to 6 are flowcharts illustrating the detailed process of the LVO data generation steps for the patient shown in Figure 3.

[0102] Referring to Figure 4A, the judgment unit 14 can analyze whether eye deviation to the left or right is confirmed in both eyes (S17a), or whether eye deviation to different directions other than forward is confirmed in both eyes (S17b).

[0103] In step S17b, the deviation of the eyes in different directions means that both eyes are directed to the left or right, excluding the forward direction, such as the left eye turning to the left and the right eye turning to the right.

[0104] In this case, if deviations in different directions are confirmed in both eyes, S17b-YES, the judgment unit 14 prevents the binocular deviations extracted through the LVO examination from being reflected in the data of suspected LVO cases, S17c.

[0105] In contrast, if a deviation in the same direction is confirmed in both eyes, S17b-NO, the judgment unit 14 ensures that the binocular deviation extracted through the LVO examination is reflected in the data of the suspected LVO case, S17d.

[0106] Referring to Figure 4B, the determination unit 14 can determine in step S17d whether the LVO probability values ​​of each hemisphere exceed the reference critical value in step S17e.

[0107] At this time, if the LVO probability values ​​in each hemisphere exceed the reference critical value (S17e-YES), the judgment unit 14 determines the patient to be LVO positive based on the LVO probability values ​​in each hemisphere (S17f), and such positive LVO determination results are reflected in the data for suspected LVO cases.

[0108] In contrast, if the LVO probability values ​​in each hemisphere are below the reference critical value, S17-NO, the judgment unit 14 determines the patient to be LVO-negative based on the LVO probability values ​​in each hemisphere, S17g, and such negative LVO determination results are reflected in the data for suspected LVO cases.

[0109] Referring to Figure 5, the judgment unit 14 can determine S17h whether the LVO probability value of the hemisphere in which the eye deviation was confirmed exceeds the reference critical value when a deviation to the left or right side, excluding the forward direction, is confirmed in only the left or right eye.

[0110] In this case, the hemisphere in which eye deviation is observed refers to the left or right hemisphere corresponding to the direction in which the left or right eye is pointing.

[0111] At this time, if the LVO probability value of the hemisphere in which ocular deviation was confirmed exceeds the reference critical value, S17i-YES, the judgment unit 14 determines the patient to be LVO positive based on the LVO probability value of the hemisphere in which ocular deviation was confirmed, S17j, and such positive LVO determination results are reflected in the data of suspected LVO cases.

[0112] In contrast, if the LVO probability value of the hemisphere in which the eye deflection was confirmed is below the reference critical value, S17i-NO, the judgment unit 14 can determine whether the LVO probability value of the hemisphere in which the eye deflection was confirmed is less than 0.1 and whether the LVO probability value of the opposite hemisphere is 0.9 or greater, S17k.

[0113] At this time, if the LVO probability value of the hemisphere in which the eye deviation was confirmed is not less than 0.1, or if the LVO probability value of the opposite hemisphere is not 0.9 or greater, then S17k-NO, the judgment unit 14 determines the patient to be LVO positive based on the LVO probability value of the hemisphere in which the eye deviation was confirmed, S17j, and such positive LVO determination results are reflected in the data of suspected LVO cases.

[0114] In contrast, if the LVO probability value of the hemisphere in which ocular deviation is confirmed is less than 0.1 and the LVO probability value of the opposite hemisphere is 0.9 or greater, then S17k-YES, the judgment unit 14 ignores the ocular deviation results extracted through the LVO test and determines the patient to be LVO positive based on the LVO probability values ​​of each hemisphere, S17l, so that such positive LVO determination results are reflected in the data of suspected LVO cases.

[0115] Referring to Figure 6, the judgment unit 14 can determine S17m whether the LVO probability values ​​of each hemisphere exceed the reference critical value when at least one deviation of the left eye or right eye is not confirmed through the LVO test, or when the deviation of both eyes is classified as forward.

[0116] At this point, if the LVO probability values ​​in each hemisphere exceed the reference critical value, S17n-YES, the judgment unit 14 classifies the patient as LVO positive, S17o, and such positive LVO classification results are reflected in the data for suspected LVO cases.

[0117] In contrast, if the LVO probability values ​​in each hemisphere are below the reference critical value, S17n-NO, the judgment unit 14 classifies the patient as LVO-negative, S17p, and such negative LVO classification results are reflected in the data for suspected LVO cases.

[0118] A detailed description of preferred embodiments of the present invention, as disclosed above, is provided so that those skilled in the art can embody and practice the invention. While preferred embodiments of the invention have been described above with reference, those skilled in the art will understand that the invention can be modified and altered in various ways without departing from the scope of the invention. For example, those skilled in the art can use the configurations described in the above embodiments in combination with each other. Thus, the invention is not intended to be limited to the embodiments presented herein, but rather to grant the broadest possible scope consistent with the principles and novel features disclosed herein.

[0119] The present invention can be embodied in other specific forms without departing from the technical spirit and essential features of the invention. Therefore, the above detailed description should not be interpreted restrictively in all respects and should be considered illustrative. The scope of the invention should be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention. The invention is not limited to the embodiments presented herein, but seeks to provide the broadest possible scope consistent with the principles and novel features disclosed herein. Furthermore, examples can be formed by combining claims that are not explicitly referenced in the claims, or by including new claims through amendments after filing. [Explanation of symbols]

[0120] 10. Acute large vessel occlusion testing device 11. Video Input Section 12 Pre-processing section 13. Video Processing Section 14 Judgment Department 14a First classification model 14b Second classification model

Claims

1. A video input unit that receives a sequence of non-contrast CT images, which are sequences of images consisting of at least multiple consecutive non-contrast CT images including the upper part of the brain corresponding to the entire midbrain and cerebral cortex, and the anatomical individuals that make up both eyes, A preprocessing unit that preprocesses the non-contrast CT images based on rigid body matching and affine matching, An image processing unit that performs non-rigid body alignment of at least one atlas on the non-contrast CT image preprocessed by the preprocessing unit, normalizes each cell of the non-contrast CT image aligned with the atlas so that it has a value of 1 when it contains the volume of the anatomical individual and a value of 0 when it does not contain the volume of the anatomical individual, and extracts a region of interest from the non-contrast CT image from which each cell of the non-contrast CT image aligned with the atlas that had a value of 0 or 1 through the normalization process has been restored to its original scale through inverse transformation, An acute large vessel occlusion examination device utilizing binocular deviation, comprising: a judgment unit that identifies binocular deviation (eyeball deviation) from non-contrast CT images from which the region of interest has been extracted; proceeds with LVO testing based on the binocular deviation to determine whether the patient is LVO positive or negative; classifies patients determined to be LVO positive through the LVO testing as suspected LVO cases; and provides notification to medical professionals.

2. The unit that makes the determination said, The acute large vessel occlusion examination device utilizing binocular bias according to claim 1, characterized in that it performs model learning to classify the progress of the LVO examination and the suspected LVO cases.

3. The unit that makes the determination said, The acute large vessel occlusion examination device utilizing binocular bias according to claim 2, characterized in that model learning is performed through a first classification model which is a 2D and 3D convolutional neural network and a second classification model which is a recursive neural network-based bidirectional long short-term memory model.

4. The first classification model described above is The acute large vessel occlusion examination device utilizing binocular bias according to claim 3, characterized in that it is an artificial intelligence model that identifies the characteristics of a biomarker that is a sign indicator of LVO positivity from non-contrast CT images from which the region of interest has been extracted.

5. The aforementioned biomarker is The acute great vessel occlusion examination device utilizing binocular deflection according to claim 4, characterized by including dense middle cerebral artery signs (Dense MCA Sign, DMS), early ischemic changes (EIC), and binocular deflection.

6. The unit that makes the determination said, The LVO examination apparatus for acute great vessel occlusion using binocular deflection according to claim 5, characterized in that the LVO examination is performed on the basis of identifying the biomarker from a non-contrast CT image from which the region of interest has been extracted.

7. The unit that makes the determination said, In the aforementioned LVO test, the deviation of both eyes is classified into one of three classes: forward, left, or right. Based on the LVO test described above, the LVO discrimination result for each hemisphere is determined to be positive or negative based on the presence of dense middle cerebral artery symbols and early ischemic changes. The acute great vessel occlusion examination device utilizing binocular deviation according to claim 5, characterized in that, in the LVO examination, the LVO probability values ​​for each hemisphere are calculated to be between 0 and 1 based on the LVO discrimination results for each hemisphere.

8. The aforementioned LVO suspicion case is, The acute great vessel occlusion examination device utilizing binocular deviation according to claim 7, characterized in that the data is generated in a data format that includes the deviation of both eyes, the LVO probability values ​​of each hemisphere, and the LVO discrimination results of each hemisphere.

9. The aforementioned notification is, The acute great vessel occlusion examination device utilizing binocular deflection according to claim 8, characterized in that it includes suspected cases of LVO in the patient and a compressed non-contrast CT image of the patient that can be previewed.

10. The aforementioned video processing unit, The acute great vessel occlusion examination device utilizing binocular deflection according to claim 1, characterized in that each cell in the non-contrast CT image aligned with the atlas is assigned a value of 1 when the volume of the anatomical individual is included in part or all of the cell's region, and is assigned a value of 0 when the volume of the anatomical individual is not included in the entire cell's region.

11. The aforementioned atlas, The acute great vessel occlusion examination apparatus utilizing binocular deflection according to claim 1, characterized in that it is a pair of images of an anatomical individual used to divide the pre-processed non-contrast CT image into the anatomical individual.

Images