Method and analyzer for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid (CSF) in the extracerebral region.

Abstract

A method for calculating dementia-related indicators from brain images includes the steps of: an analyzer receiving a brain image of a subject as input; the analyzer classifying regions of interest from the brain image; the analyzer calculating the volume or area of ​​the regions of interest; and the analyzer calculating dementia-related indicators for the subject based on the volume or area.

Description

Technical Field

[0001] The technology described below is a technology for calculating dementia-related information based on brain images.

Background Art

[0002] Dementia refers to a syndrome that causes impairment of cognitive functions such as memory, language, and judgment. There are various types of dementia, and Alzheimer's disease is the most common form of dementia.

[0003] Dementia is a disease whose symptoms progress over a long period, and pathological accumulation progresses even before clinical symptoms appear. Therefore, early diagnosis of dementia is very important for delaying the onset and managing dementia symptoms.

[0004] Brain images such as magnetic resonance imaging (MRI) and positron emission tomography (PET) are used for the diagnosis of dementia. A typical factor related to dementia is the cortical thickness of the cerebral cortex.

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, there are variations in the cortical thickness depending on the imaging device (vendor) used in medical institutions. Therefore, it can be said that the cortical thickness has some limitations as an index for accurately diagnosing dementia.

[0006] An object of the present invention is to provide a method for calculating dementia-related information based on other regions of interest that can be extracted from brain images instead of the cerebral cortex from the technology described below.

Means for Solving the Problems

[0007] Method for Calculating Dementia-Related Indicators Based on the Volume of the Extracerebral Cerebrospinal Fluid Region The method for calculating dementia-related indicators based on the volume of the extracerebral cerebrospinal fluid region includes the steps of: an analyzer receiving a brain image of a subject as input; the analyzer classifying a region of interest from the brain image; the analyzer calculating the volume or area of ​​the region of interest; and the analyzer calculating a dementia-related indicator of the subject based on the volume or area.

[0008] In other contexts, a method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid includes the steps of: an analyzer receiving a brain image of a subject as input; the analyzer classifying regions of interest from the brain image; and the analyzer inputting the regions of interest into a pre-trained learning model to calculate dementia-related indicators for the subject.

[0009] The analysis device for calculating dementia-related indicators includes an input device that receives brain images of a subject, and a calculation device that divides regions of interest from the brain images and calculates dementia-related indicators based on the regions of interest. The regions of interest include extracerebral cerebrospinal fluid regions. [Effects of the Invention]

[0010] The technique described below provides, in brain imaging, the degree of brain atrophy centered on the extracerebral cerebrospinal fluid region, which is clearly distinguishable from other brain structures. Therefore, the technique described below can provide useful dementia-related information regardless of differences in the characteristics of imaging equipment. [Brief explanation of the drawing]

[0011] [Figure 1] This figure shows an example of estimating the thickness of the cerebral cortex from medical images. [Figure 2] This figure shows an example of a system that analyzes brain images to calculate dementia-related indicators. [Figure 3] This figure shows an example of the process of analyzing brain images to calculate dementia-related indicators. [Figure 4]This figure shows an example of the process for calculating dementia-related indicators based on the extracerebral cerebrospinal fluid (CSF) region and the ventricular region. [Figure 5] This figure shows the results of the evaluation of dementia-related factors in the lateral ventricular region and the extracerebral cerebrospinal fluid region. [Figure 6] This figure shows the results of the evaluation of dementia-related associations in extracerebral cerebrospinal fluid subregions. [Figure 7] This figure shows the performance evaluation results of a classifier constructed based on the lateral ventricular region and the extracerebral cerebrospinal fluid region. [Figure 8] This figure shows the performance evaluation results of a classifier that uses patient information in addition to the model shown in Figure 7. [Figure 9] This figure shows the performance evaluation results of a classifier that uses patient information in addition to the lateral ventricular region and extracerebral cerebrospinal fluid subregions. [Figure 10] This figure shows an example of the process of calculating dementia-related indicators using a learning model. [Figure 11] This figure shows an example of an analytical device for calculating dementia-related indicators. [Modes for carrying out the invention]

[0012] The technology described below can be modified in various ways and may have various embodiments; therefore, specific embodiments are illustrated in the drawings and described in detail. However, this should not be understood as limiting the technology described below to specific embodiments, but rather as including all modifications, equivalents, or substitutes that fall within the concept and scope of the technology described below.

[0013] Terms such as First, Second, A, B, etc., can be used to describe various components, but the components are not limited by such terms and are used solely for the purpose of distinguishing one component from another. For example, without exceeding the scope of the rights of the technology described below, the First component may be named the Second component, and similarly, the Second component may be named the First component. The terms and / or include combinations of multiple related described items, or any one of multiple related described items.

[0014] In the terminology used herein, singular expressions should be understood to include plural expressions unless the context clearly indicates otherwise, and terms such as “includes” should be understood to mean that the described features, numbers, steps, actions, components, parts, or combinations thereof exist, and not to exclude the possibility of the presence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0015] Prior to a detailed description of the drawings, it should be made clear that the classification of components in this specification is merely based on the primary function each component performs. That is, two or more components described below may be combined into one component, or one component may be subdivided into two or more components with more detailed functions. Furthermore, each component described below may perform some or all of the functions performed by other components in addition to its primary function, and it goes without saying that some of the primary functions performed by each component may be exclusively performed by other components.

[0016] Furthermore, when performing a method or operation, the steps constituting the method may be performed in an order different from the order described, unless the context clearly indicates a specific order. That is, the steps may be performed in the same order as described, substantially simultaneously, or in reverse order.

[0017] Alzheimer's disease is diagnosed based on the thickness of the cerebral cortex grasped by medical images. Figure 1 shows an example of estimating the thickness of the cerebral cortex from a medical image. In Figure 1, an example of measuring or estimating the thickness of the cerebral cortex from an MRI image is shown. In Figure 1, the cerebral cortex corresponds to the gray area between the white solid line and the dotted line. Therefore, in order to accurately measure the thickness of the cerebral cortex, it is necessary to clearly distinguish the white area inside the white solid line and the gray area between the white solid line and the dotted line. On the other hand, medical imaging devices are provided by various vendors, and even for devices manufactured by the same vendor, the parameters for image generation may differ depending on the type of device.

[0018] The researchers collected data on the population who visited the affiliated medical institution (Samsung Seoul Hospital in Korea). The population included 605 participants in the normal group and 616 participants in the dementia group who underwent examinations at Samsung Seoul Hospital between 2015 and 2021. All participants received dementia evaluations such as brain MRI and amyloid PET. Here, the normal group was composed of amyloid-negative subjects, and the dementia group was composed of amyloid-positive subjects.

[0019] [Table 1]

[0020] Table 1 above shows the results of measuring the thickness of the cerebral cortex of subjects using the same manufacturer's (Philips) imaging devices, Archieva and Ingenia, used at the medical institution to which the researchers belong. Aβ(-)NC refers to the normal control group without amyloid-beta accumulation, and Aβ(+)ADD refers to the Alzheimer's disease patients group with amyloid-beta accumulation. In Aβ(+)ADD, no significant difference was observed in cerebral cortex thickness in MRI scans taken with the Archieva and Ingenia devices. However, in Aβ(-)NC, a significant difference was observed in the measurements of the temporal and occipital lobes using both devices. In other words, as the researchers expected, even with medical devices from the same vendor, the measured values ​​of cerebral cortex thickness differed slightly depending on the type of device. This is because there are differences in the parameter settings between different imaging devices.

[0021] Figure 1 shows a portion of the extracerebral cerebrospinal fluid (Extracerebral CSF) region. The extracerebral CSF region is the sulcus region between gyrus. On MRI images, the extracerebral CSF region appears black or very dark. In other words, the extracerebral CSF region is a region that is visually clearly distinguishable from the gray cerebral cortex region on MRI images. Therefore, it can be estimated that the extracerebral CSF region is a region that can be distinguished relatively accurately using computer image processing techniques or segmentation models.

[0022] [Table 2]

[0023] Table 2 shows the results of measuring the length (or size) of the extracerebral cerebrospinal fluid (Extracerebral Cerebrospinal Fluid) region using the Achieva and Ingenia instruments for the population shown in Table 1. The results in Table 2 show that there was no significant difference in the length of the Extracerebral Cerebrospinal Fluid Region between the two instruments for Aβ(+)ADD and Aβ(-)NC. That is, as the researchers predicted, the Extracerebral Cerebrospinal Fluid Region has features that are clearly distinguishable on imaging compared to the cerebral cortex region. The thickness of the cerebral cortex is an indicator of the degree of brain atrophy. As the degree of atrophy increases, the thickness of the cerebral cortex decreases. Furthermore, the size of the Extracerebral Cerebrospinal Fluid Region also correlates with the degree of brain atrophy. As the degree of atrophy increases, the size of the Extracerebral Cerebrospinal Fluid Region also increases.

[0024] In the following, "dementia" refers to Alzheimer's disease.

[0025] The technologies described below are techniques for diagnosing dementia or predicting the likelihood of developing dementia, primarily focusing on the extracerebral cerebrospinal fluid region.

[0026] The size of the extracerebral cerebrospinal fluid (CSF) region can be evaluated by the spacing or distance between gyrus. Alternatively, the size of the extracerebral CSF region can be evaluated by the area of ​​a specific sulcus. This area can be calculated from 2D (dimension) images. Alternatively, the size of the extracerebral CSF region can be evaluated by the volume of a specific sulcus. This volume can be calculated from 3D images or 2D slices.

[0027] Hereafter, we will refer to a device that analyzes brain images to calculate dementia-related indicators for a subject as an analysis device. The analysis device can take the form of a computer device such as a PC, a smart device, a network server, or a chipset dedicated to data processing.

[0028] The analysis device may calculate the size or volume of a specific extracerebral cerebrospinal fluid (CSF) region from brain images using conventional image processing techniques. Alternatively, the analysis device may calculate the size or volume of a specific extracerebral cerebrospinal fluid (CSF) region using a model, which is a type of deep learning network. Based on the size or volume of the extracerebral cerebrospinal fluid region, the analysis device can calculate the degree of brain atrophy. Furthermore, the analysis device can analyze brain images to calculate a diagnosis or prediction of dementia in the subject.

[0029] The analyzer analyzes brain images to calculate dementia-related indices. Here, dementia-related indices correspond to information or factors related to the progression of dementia. Dementia-related indices may include at least one of the following pieces of information: the size or volume of the extracerebral cerebrospinal fluid (Extracerebral Cerebrospinal Fluid) region, the degree of brain atrophy as assessed by the size or volume of the Extracerebral Cerebrospinal Fluid region, and the presence or absence of dementia (or the progression of dementia) as assessed by the size / volume (or degree of brain atrophy) of the Extracerebral Cerebrospinal Fluid region.

[0030] Figure 2 shows an example of a system 100 that analyzes brain images to calculate dementia-related indicators. In Figure 2, the analysis equipment is shown as an example consisting of a computer terminal 130 and a server 140.

[0031] The medical imaging device 110 generates brain images of the patient (e.g., MRI images). The brain images generated by the medical imaging device 110 may be stored in another database, such as an electronic medical record (EMR) 120.

[0032] In Figure 2, User A can analyze brain images using the computer terminal 130 and obtain dementia-related indicators. The computer terminal 130 can receive input of brain images of a specific subject from a medical imaging device 110 or EMR 120 via a wired or wireless network. In some cases, the computer terminal 130 may be a device physically connected to the medical imaging device 110. The computer terminal 130 extracts the extracerebral cerebrospinal fluid (CSF) region from the brain images. The computer terminal 130 can calculate dementia-related indicators based on the CSF region. (i) The computer terminal 130 can estimate the size or volume of the CSF region. (ii) The computer terminal 130 can estimate the degree of brain atrophy based on the size or volume of the CSF region. (iii) Based on the size / volume of the CSF region or the degree of brain atrophy, the computer terminal 130 can estimate the presence or absence of dementia, the possibility of dementia onset, or the progression of dementia. User A can check the analysis results on the computer terminal 130.

[0033] Server 140 can receive brain images of a specific subject from a medical imaging device 110 or EMR 120. Server 140 extracts the extracerebral cerebrospinal fluid (CSF) region from the brain images. Server 140 can calculate dementia-related indices based on the CSF region. (i) Server 140 can estimate the size or volume of the CSF region. (ii) Server 140 can estimate the degree of brain atrophy based on the size or volume of the CSF region. (iii) Server 140 can estimate the presence or absence of dementia, the likelihood of developing dementia, or the progression of dementia, depending on the size / volume of the CSF region or the degree of brain atrophy. Server 140 can transmit the results of the brain image analysis to User A's terminal. User A can check the analysis results through the user terminal.

[0034] The computer terminal 130 and / or server 140 can also store the analysis results in the EMR 120.

[0035] Figure 3 shows an example of the process (200) of analyzing brain images to calculate dementia-related indicators.

[0036] The analysis device receives input of the subject's brain image (MRI image) (210).

[0037] The analysis device can preprocess the input brain image data to a certain extent (220). Data preprocessing may be a process of extracting the surface structure or model of the brain structure from the input image and generating a mask. For example, the analysis device can extract an entire brain region from a brain MRI image using a CIVET pipeline. The analysis device can extract a brain region from an MRI slice. The analysis device can extract a brain region from a series of MRI slices and extract a three-dimensional brain region. In summary, (i) the analysis device can generate a mask of an entire brain region from a brain image. Also, (ii) the analysis device can generate a mask of a specific region from a brain image. A specific region may include at least one of the following: extracerebral cerebrospinal fluid region, extracerebral cerebrospinal fluid subregion, ventricle region, and ventricular subregion. Extracerebral cerebrospinal fluid subregions and ventricular subregions will be discussed later.

[0038] On the other hand, the data preprocessing process for generating the mask may be an optional process. When using image processing techniques that automatically extract regions of interest (ROIs), the analysis device does not need to prepare the mask in advance.

[0039] The analysis device can segment entire brain regions from the subject's brain images (230). The analysis device can segment entire brain regions using a mask prepared during the data preprocessing process. Alternatively, the analysis device can segment entire brain regions using a segmentation model.

[0040] The analyzer can segment ROIs across the entire brain region (240). The analyzer can segment ROIs using masks prepared during the data preprocessing process. Alternatively, the analyzer can segment ROIs using a segmentation model. ROIs include extracerebral cerebrospinal fluid (CSF) regions. ROIs may include regions other than extracerebral cerebrospinal fluid regions. ROIs may also consist of subregions that can be extracted from extracerebral cerebrospinal fluid regions. Specific ROIs will be discussed later.

[0041] The analyzer can calculate the volume of a divided ROI (250). The analyzer can calculate the volume of a 3D ROI using commercially available programs and algorithms. On the other hand, the analyzer can also calculate the area of ​​a 2D ROI from an MRI slice. The analyzer can calculate the area of ​​an ROI from all slices or from selected specific slices.

[0042] The analytical device can estimate the degree of brain atrophy based on the volume or area of ​​the ROI (260). The dementia-related index in Figure 3 is the degree of brain atrophy. The volume of the ROI or the area of ​​a specific support point(s) has a certain correlation with the degree of brain atrophy. The correlation between the volume (or area) of the ROI and the degree of brain atrophy can be prepared in advance in the form of a table. In this case, the analytical device can estimate the degree of brain atrophy of the subject based on the calculated volume (or area) of the ROI. Alternatively, the analytical device can estimate the degree of brain atrophy of the subject using a function in which the calculated volume (or area) of the ROI is a variable. In this case, the function may be a publicly available formula. Alternatively, in this case, the function may be a formula prepared in advance by regression analysis.

[0043] The ROI includes the extracerebral cerebrospinal fluid (CSF) region. Furthermore, the ROI may include at least a portion of individual subregions of the extracerebral CSF region. The ROI may also include the ventricular region, which is relatively clearly distinguishable because it is not white or gray on MRI. Furthermore, the ROI may include at least a portion of individual subregions of the ventricular region. The ROI may be any of the various regions, or any possible combination of the various regions. Candidate ROIs are shown in Table 3 below.

[0044] [Table 3]

[0045] Possible ROIs are as follows: (i) The ROI may be at least one of the entire extracerebral cerebrospinal fluid region and the entire ventricular region. (ii) The ROI may also be at least one of the entire extracerebral cerebrospinal fluid region and a subregion of the ventricular region. (iii) Furthermore, the ROI may be at least one of the extracerebral cerebrospinal fluid subregion and the entire ventricular region. (iv) Furthermore, the ROI may also be at least one of the extracerebral cerebrospinal fluid subregion and a subregion of the ventricular region.

[0046] Figure 4 shows an example of the process (300) for calculating dementia-related indicators based on the extracerebral cerebrospinal fluid region and the ventricular region. Figure 4 shows an example of calculating dementia-related indicators using sub-regions within the ROI.

[0047] The analysis device receives input of the subject's brain image (MRI image) (310).

[0048] The analysis device can preprocess the input brain image data to a certain extent (320). Data preprocessing may involve extracting surface structures or models of brain structures from the input images to generate masks. The analysis device can generate masks of entire brain regions from brain images. The analysis device can also generate masks of specific regions from brain images. Specific regions may include at least one of the extracerebral cerebrospinal fluid regions, extracerebral cerebrospinal fluid subregions, ventricular regions, and ventricular subregions.

[0049] On the other hand, the data preprocessing process for generating the mask may be a selective process. When using an image processing technique that automatically extracts ROIs, the analysis device does not need to prepare the mask in advance.

[0050] The analysis device can segment the entire ROI region from the subject's brain image (330). The analysis device can segment the entire brain region using a mask prepared during the data preprocessing process. Alternatively, the analysis device can segment the entire brain region using a segmentation model.

[0051] The analyzer can isolate a specific ROI from an entire brain region. Using a mask, the analyzer can isolate at least one of the extracerebral cerebrospinal fluid subregions (340). The analyzer can also isolate at least one of the ventricular subregions using a mask (350). The analyzer can also isolate a target ROI using a segmentation model.

[0052] The analyzer can calculate the volume of a divided ROI (360). The analyzer can calculate the volume of a 3D ROI using commercially available programs and algorithms. Alternatively, the device can calculate the area of ​​a 2D ROI from an MRI slice. The device can calculate the ROI area from the entire slice or from selected specific slices.

[0053] The analyzer can normalize each brain region using the total brain volume (Intracranial Volume, ICV) of the subject (370). The analyzer can correct the size or volume of the target ROI to be constant based on the ICV. The correction process based on the ICV is a selective process.

[0054] The analyzer can estimate the degree of brain atrophy based on the volume or area of ​​the ROI (380). The analyzer can estimate the degree of brain atrophy based on the volume or area corrected for ICV. For example, the analyzer can normalize the volume (or area) of the ROI by dividing it by the ICV.

[0055] The dementia-related indicator in Figure 4 represents the degree of brain atrophy. The analysis device can estimate the degree of brain atrophy in a subject by comparing the calculated ROI volume (or area) with a pre-prepared reference. Alternatively, the analysis device can estimate the degree of brain atrophy in a subject using a function with the calculated ROI volume (or area) as a variable.

[0056] The researchers selected a return on investment (ROI) to calculate dementia-related indicators.

[0057] The researchers examined the dementia association between the lateral ventricle and the entire extracerebral cerebrospinal fluid (CSF) region, both subregions of the ventricle. Figure 5 shows the results of the dementia association assessment for the lateral ventricle and extracerebral CSF regions. In Figure 5, A(-)NC represents the normal group without amyloid-beta accumulation, and A(+)ADD represents the Alzheimer's disease patient group with amyloid-beta accumulation. The results in Figure 5 show that the lateral ventricle and extracerebral CSF regions each represent a ROI. (i) The lateral ventricle region can serve as an indicator to distinguish between the normal group and the patient group (p<0.001). Also, (ii) the entire extracerebral CSF region can serve as an indicator to distinguish between the normal group and the patient group (p<0.001).

[0058] The researchers examined the association of each extracerebral cerebrospinal fluid (Extracerebral Cerebrospinal Fluid) subregion with dementia. Figure 6 shows the results of the assessment of dementia association for each Extracerebral Cerebrospinal Fluid subregion. The Extracerebral Cerebrospinal Fluid subregions include the frontal (F), temporal (T), parietal (P), and occipital (O) regions. In Figure 6, A(-)NC represents the normal group without amyloid-beta accumulation, and A(+)ADD represents the Alzheimer's disease patient group with amyloid-beta accumulation. The results in Figure 6 show that each Extracerebral Cerebrospinal Fluid subregion represents a ROI (p<0.001).

[0059] The researchers constructed a model (classifier) ​​to calculate dementia-related indicators based on selected ROIs. The classifier was implemented using a machine learning model. The researchers used 70% of the aforementioned population data as training data and the remaining 30% as validation data. They performed logistic regression using the glm function in R. However, the classifier could also be implemented using other models, such as deep learning models.

[0060] Figure 7 shows the performance evaluation results of classifiers constructed based on the lateral ventricular region and the extracerebral cerebrospinal fluid region. Figure 7 shows the performance of two models (Model 1 and Model 2). Models 1 and 2 are models that calculate dementia-related indices using only brain images. Model 1 is a model that uses the entire lateral ventricular region and the extracerebral cerebrospinal fluid region as ROIs. Model 2 is a model that uses the lateral ventricular region and sub-regions of the extracerebral cerebrospinal fluid region as ROIs. The area under the receiver operating characteristic (ROC) curve (AUC) of Model 1 was 0.808. The sub-regions of the extracerebral cerebrospinal fluid region used in Model 2 are the frontal region (F), temporal region (T), parietal region (P), and occipital region (O). The AUC of Model 2 was 0.854. Model 2 showed slightly higher performance than Model 1, but both Model 1 and Model 2 are sufficiently significant for diagnosing or predicting dementia.

[0061] Figure 8 shows the performance evaluation results of a classifier that uses patient information in addition to the entire lateral ventricular region and extracerebral cerebrospinal fluid region. Figure 8 shows the performance of two models (Model 3 and Model 4). Models 3 and 4 are models that calculate dementia-related indices using brain images and patient information. Models 3 and 4 are pre-trained models using brain images and patient information.

[0062] Model 3 is a model that further incorporates patient information (age, sex, education level) in addition to Model 1. Specifically, Model 3 calculates dementia-related indices using (i) the entire lateral ventricular region and extracerebral cerebrospinal fluid region extracted from brain images, and (ii) patient information (age, sex, education level). The AUC of Model 3 was 0.829.

[0063] Model 4 is a model that, in addition to Model 1, further uses patient information (age, sex, education level) and clinical information (APOE e4). Specifically, Model 3 is a model that calculates dementia-related indices using (i) the entire lateral ventricular region and extracerebral cerebrospinal fluid region extracted from brain images, (ii) patient information (age, sex, education level), and (iii) the patient's clinical information (APOE e4). APOE e4 represents genotype information of genes associated with dementia. The AUC of Model 4 was 0.883.

[0064] Both Model 3 and Model 4 demonstrated higher performance than Model 1. Therefore, both Model 3 and Model 4 are sufficiently significant for diagnosing or predicting dementia.

[0065] Figure 9 shows the performance evaluation results of a classifier that uses patient information in addition to the lateral ventricular region and extracerebral cerebrospinal fluid subregions. Figure 9 shows the performance of two models (Model 5 and Model 6). Models 5 and 6 are models that calculate dementia-related indices using brain images and patient information. Models 5 and 6 are pre-trained models using brain images and patient information.

[0066] Model 5 is a model that calculates dementia-related indicators using (i) ventricular subregions and extracerebral cerebrospinal fluid subregions (F, P, T, O) extracted from brain images, and (ii) patient information (age, sex, education level). The AUC of Model 5 was 0.889.

[0067] Model 6 is a model that calculates dementia-related indicators using (i) ventricular subregions and extracerebral cerebrospinal fluid subregions (F, P, T, O) extracted from brain images, (ii) patient information (age, sex, education level), and (iii) patient clinical information (APOE e4). The AUC of Model 6 was 0.932.

[0068] Both Model 5 and Model 6 demonstrated higher performance than Model 2. Therefore, both Model 5 and Model 6 are sufficiently significant for the diagnosis or prediction of dementia.

[0069] In summary, the ROI for dementia-related indicators was found to be significant for all of the following: (i) the entire extracerebral cerebrospinal fluid (CSF) region, (ii) all sub-regions of the extracerebral cerebrospinal fluid (CSF) region, (iii) a portion of a sub-region of the extracerebral cerebrospinal fluid (CSF) region, (iv) the entire extracerebral cerebrospinal fluid (CSF) region plus the lateral ventricular region, (v) all sub-regions of the extracerebral cerebrospinal fluid (CSF) region plus the lateral ventricular region, (vi) a portion of a sub-region of the extracerebral cerebrospinal fluid (CSF) region plus the lateral ventricular region, and (vii) at least a portion of a sub-region of the extracerebral cerebrospinal fluid (CSF) region plus the ventricular sub-region. On the other hand, the classifier that experimentally used sub-regions of the extracerebral cerebrospinal fluid (CSF) region as the ROI showed slightly higher performance than the classifier that used the entire extracerebral cerebrospinal fluid (CSF) region. Furthermore, the classifier that used only brain MRI also showed sufficiently significant performance. In addition, it was confirmed that the performance of the classifier improved when additional information such as patient information was used in addition to brain MRI images.

[0070] Figure 10 shows an example of the process (400) for calculating dementia-related indicators using a learning model. Figure 10 shows a case where a learning model is used in both the process of extracting ROIs from brain MRI images and the process of predicting dementia-related indicators based on ROIs.

[0071] The analysis device receives input of the subject's brain image (MRI image) (410).

[0072] The analysis device can input the input brain image into a pre-trained segmentation model (420). The segmentation model can be implemented using models of various types and structures. For example, the segmentation model may be a U-net-based model. Depending on its type and learning process, the segmentation model can segment various ROIs. The segmentation model can segment ROIs from a 3D brain image. Alternatively, the segmentation model can segment ROIs from individual 2D slices. As mentioned above, parts of various regions can be used as ROIs. Segmentation models can be prepared in advance for each specific ROI. For example, the segmentation model may vary, such as (i) a model that segments the entire extracerebral cerebrospinal fluid region, (ii) a model that segments at least a part of an extracerebral cerebrospinal fluid subregion, (iii) a model that segments the entire extracerebral cerebrospinal fluid region + lateral ventricular region, or (iv) a model that segments at least a part of an extracerebral cerebrospinal fluid subregion + lateral ventricular region. A segmentation model can be pre-built as a model that divides the data into one of several ROIs, depending on the training data and the learning process.

[0073] The analytical instrument can acquire the results (ROI classifications) output by the segmentation model (430).

[0074] The analysis device inputs the ROI into a pre-trained classification model (440). The classification model calculates dementia-related indicators by inputting images, or images and patient information.

[0075] A classification model is a machine learning model. Therefore, a classification model can be any one of various types of models. For example, a classification model may be a model implemented using one of the following methods: decision tree, random forest (RF), K-nearest neighbor (KNN), Naive Bayes, support vector machine (SVM), artificial neural network (ANN), or regression model. There are also various types of ANN models. For example, a classification model may be a model based on a convolutional neural network (CNN).

[0076] The analyzer can acquire dementia-related indicators output by the classification model (450). Dementia-related indicators may include information such as ROI volume, degree of brain atrophy, or progression of dementia.

[0077] The classification model can calculate dementia-related indicators using only ROI. Furthermore, the classification model can calculate dementia-related indicators by inputting ROI and patient information (age, sex, education level, etc.). Additionally, the classification model can calculate dementia-related indicators by inputting ROI, patient information (age, sex, education level, etc.), and clinical information (APOE e4, etc.).

[0078] On the other hand, unlike in Figure 10, the learning model can also be used in either the process of extracting ROIs or the process of predicting dementia-related indicators. That is, (i) the analyzer can extract ROIs using a segmentation model and estimate the degree of brain atrophy by calculating the volume of the extracted ROIs. The volume calculation and determination of the degree of brain atrophy are as described in Figures 3 and 4. Alternatively, (ii) the analyzer can extract ROIs using a mask and input the extracted ROIs into the classification model in Figure 10 to calculate dementia-related indicators. The process of extracting ROIs using a mask is as described in Figures 3 and 4.

[0079] Figure 11 shows an example of an analytical device 500 for calculating dementia-related indicators. The analytical device 500 corresponds to the analytical devices described above (130 and 140 in Figure 1). The analytical device 500 can be implemented in various physical forms. For example, the analytical device 500 can take the form of a computer device such as a PC, a network server, or a chipset dedicated to data processing.

[0080] The analysis device 500 may include a storage device 510, a memory 520, an arithmetic unit 530, an interface device 540, a communication device 550, and an output device 560.

[0081] The memory device 510 can store brain images (MRI images) of a subject generated by a medical imaging device.

[0082] The storage device 510 can store patient information and clinical information about the subject. The patient information and clinical information are as described above.

[0083] The storage device 510 can store code or programs for calculating dementia-related indicators from brain images.

[0084] The memory device 510 can store masks (or multiple masks) extracted from brain images.

[0085] The memory device 510 can store segmentation models for extracting ROIs from brain images. These segmentation models are pre-trained models.

[0086] The memory device 510 can store a classification model that calculates dementia-related indicators upon receiving an ROI input. The classification model is a pre-trained model.

[0087] The memory device 510 can store dementia-related indicators of the subject.

[0088] The memory 520 can store data and information generated during the process in which the analysis device 500 calculates dementia-related indicators from brain images.

[0089] The interface device 540 is a device that receives certain commands and data input from an external source. The interface device 540 can receive input of a subject's brain images from a physically connected input device or external storage device. The interface device 540 can receive input of a subject's patient information and / or clinical information from a physically connected input device or external storage device. The interface device 540 may transmit dementia-related indices calculated based on brain images to an external object.

[0090] The communication device 550 refers to a configuration that receives and transmits certain information via a wired or wireless network. The communication device 550 can receive brain images of a subject from an external object. The communication device 550 can receive patient information and / or clinical information of a subject from an external object. The communication device 550 can also transmit dementia-related indices calculated based on brain images to an external object such as a user terminal.

[0091] The interface device 540 and the communication device 550 are configured to send and receive certain data from a user or other physical object, and are therefore sometimes collectively referred to as input / output devices. From the standpoint of their information or data input function, the interface device 540 and the communication device 550 may also be referred to as input devices.

[0092] The output device 560 is a device that outputs certain information. The output device 560 can output interfaces required during the data processing process, brain images, ROIs (Regions of Interest) separated from brain images, and dementia-related indices calculated based on the ROIs.

[0093] The computing unit 530 can pre-process the subject's brain images to a consistent level. The data pre-processing process is as described in Figure 3. Through the data pre-processing process, the computing unit 530 can generate masks (multiple masks are possible) for dividing the entire brain region and ROI.

[0094] The computing device 530 can separate an entire brain region from a brain image using a mask for the entire brain region. Furthermore, the computing device 530 can separate a target ROI from the entire brain region using a mask for a specific ROI. The ROI can be diverse, as described above. For example, the ROI may be the entire extracerebral cerebrospinal fluid region, an extracerebral cerebrospinal fluid subregion, the entire extracerebral cerebrospinal fluid region plus the lateral ventricle region, or an extracerebral cerebrospinal fluid subregion plus the lateral ventricle region.

[0095] The computing unit 530 can also divide the ROI from the input brain image using a segmentation model.

[0096] The computing device 530 can calculate the volume or area of ​​the divided ROI. The computing device 530 can calculate the volume or area of ​​the ROI using one of the commercially available programs for calculating the volume of brain regions. Furthermore, the computing device 530 can normalize the initially calculated ROI volume based on ICV.

[0097] The calculation device 530 can estimate the degree of brain atrophy based on the volume or area of ​​the final ROI. Alternatively, the calculation device 530 can estimate the degree of dementia of the subject based on the volume or area of ​​the final ROI. Alternatively, the calculation device 530 can estimate the degree of dementia of the subject based on the degree of brain atrophy of the subject.

[0098] The computing device 530 can input ROI into a pre-built classification model to calculate dementia-related indicators. Furthermore, the computing device 530 can input ROI and patient information (age, gender, education level, etc.) into the classification model to calculate dementia-related indicators. Additionally, the computing device 530 can input ROI, patient information (age, gender, education level, etc.), and clinical information (APOE e4, etc.) into the classification model to calculate dementia-related indicators.

[0099] The arithmetic unit 530 is a device that processes data and performs certain calculations, such as a processor, an application processor (AP), or a chip into which a program is embedded.

[0100] Furthermore, the brain image processing method, dementia-related index calculation method, or dementia prediction method described above can be implemented using a program (or application) that includes an executable algorithm that can be run on a computer. The program may be stored and provided on a temporary or non-transitory computer-readable medium.

[0101] Non-temporary computer-readable media refers to media that store data semi-permanently and can be read by devices, rather than media that store data for a short moment, such as registers, caches, and memory. Specifically, the various applications or programs mentioned above may be stored and provided on non-temporary computer-readable media such as compact discs (CDs), digital versatile discs (DVDs), hard disks, Blu-ray discs, Universal Serial Bus (USB), memory cards, read-only memory (ROMs), programmable read-only memory (PROMs), erasable PROMs (EPROMs), or electrically erasable PROMs (EEPROMs), or flash memory.

[0102] Temporary computer-readable media refers to various types of random access memory (RAM), including static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synclink DRAM (SLDRAM), and direct rambus RAM (DRAM).

[0103] This embodiment and the drawings attached to this specification only clearly illustrate a part of the technical concept included in the aforementioned technology, and it will be obvious that any modifications and specific examples that can be easily inferred by a person skilled in the art within the scope of the technical concept included in the specification and drawings of the aforementioned technology are all included within the scope of the rights of the aforementioned technology.

Claims

1. The analysis device receives brain image input from the subject, The analysis device performs the step of separating the region of interest from the brain image, The analytical device calculates the volume or area of ​​the region of interest, The analysis device includes the step of calculating dementia-related indicators for the subject based on the volume or area, The aforementioned area of ​​interest is a method for calculating dementia-related indices based on the volume of the extracerebral cerebrospinal fluid (Extracerebral) region, including the Extracerebral cerebrospinal fluid (Extracerebral) region.

2. The method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid region according to claim 1, wherein the dementia-related indicators include at least one of the degree of brain atrophy, the presence or absence of dementia, and the progression of dementia.

3. The method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid region according to claim 1, wherein the analytical device divides the region of interest using a mask generated by preprocessing the brain image.

4. The aforementioned area of ​​interest is, Including the ventricular region, The method for calculating dementia-related indicators based on the volume of the extracerebral cerebrospinal fluid region according to claim 1, wherein the ventricular region includes the lateral ventricle.

5. The method for calculating dementia-related indicators based on the volume of the extracerebral cerebrospinal fluid region according to claim 1, wherein the region of interest is a subregion included in the extracerebral cerebrospinal fluid region, and the subregion includes at least one of the frontal region, the temporal region, the parietal region, and the occipital region.

6. The analysis device receives brain image input from the subject, The analysis device performs the step of separating the region of interest from the brain image, The analysis device includes the step of inputting the domain of interest into a pre-trained learning model to calculate dementia-related indicators for the subject, The aforementioned area of ​​interest is a method for calculating dementia-related indices based on the volume of the extracerebral cerebrospinal fluid (Extracerebral) region, including the Extracerebral cerebrospinal fluid (Extracerebral) region.

7. The method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid regions according to claim 6, wherein the analysis device inputs the brain images into a pre-trained segmentation model to divide the region of interest.

8. The method for calculating dementia-related indicators based on the volume of an extracerebral cerebrospinal fluid region according to claim 6, wherein the region of interest includes the entire extracerebral cerebrospinal fluid region and the lateral ventricular region.

9. The method for calculating dementia-related indicators based on the volume of the extracerebral cerebrospinal fluid region according to claim 6, wherein the region of interest is a subregion included in the extracerebral cerebrospinal fluid region, and the subregion includes at least one of the frontal region, the temporal region, the parietal region, and the occipital region.

10. The method for calculating dementia-related indicators based on the volume of an extracerebral cerebrospinal fluid region according to claim 9, wherein the region of interest further includes the lateral ventricular region.

11. The analysis device further inputs additional information about the subject into the learning model to calculate the dementia-related indicators. The method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid region according to claim 9, wherein the additional information includes at least one of the subject's age, sex, education level, and APOE e4 genotype.

12. An input device that receives brain images of the subject, The system includes a computing device that divides regions of interest from the brain image and calculates dementia-related indicators based on the regions of interest, The aforementioned area of ​​interest includes an analytical device for calculating dementia-related indicators, specifically the extracerebral cerebrospinal fluid (ECF) region.

13. The analysis device for calculating dementia-related indicators according to claim 12, wherein the calculation device divides the region of interest using a mask generated by preprocessing the brain image.

14. The calculation device is an analytical device that calculates dementia-related indicators according to claim 12, wherein the calculation device inputs the brain images into a pre-trained segmentation model to divide the regions of interest.

15. The calculation device calculates the dementia-related index based on the volume or area of ​​the region of interest, as described in claim 12.

16. The analysis device for calculating dementia-related indicators according to claim 12, wherein the computing device inputs the domain of interest into a pre-trained learning model to calculate the dementia-related indicators.

17. The analytical device for calculating dementia-related indicators according to claim 12, wherein the region of interest includes the entire extracerebral cerebrospinal fluid region and the lateral ventricular region.

18. The analyzer for calculating dementia-related indicators according to claim 12, wherein the region of interest is a subregion included in the extracerebral cerebrospinal fluid region, and the subregion includes at least one of the frontal region, the temporal region, the parietal region, and the occipital region.

19. The analyzer for calculating dementia-related indicators according to claim 12, wherein the region of interest is a subregion included in the extracerebral cerebrospinal fluid region, and the subregion includes at least one of the frontal region, the temporal region, the parietal region, and the occipital region, and the lateral ventricular region.

Images