Dementia related index calculation method and analysis device based on external cerebrospinal fluid area volume

Abstract

A method of calculating a dementia-related indicator from a brain image, comprising the steps of: an analysis device receiving a brain image input of a subject; the analysis device distinguishes a region of interest in the brain image; the analysis device calculates the volume or area of the region of interest; and calculating, by the analysis device, a dementia-related indicator of the subject based on the volume or area.

Description

Technical Field

[0001] The following technology is a technology for calculating dementia-related information based on brain images. Background Art

[0002] Dementia is a syndrome that causes cognitive function disorders such as memory, language, and judgment. Dementia includes many different types, and Alzheimer's disease is the most common dementia.

[0003] Dementia is a disease that develops over a long period, and its pathological changes accumulate before the clinical symptoms appear. Therefore, early diagnosis of dementia is crucial for delaying and managing the onset of dementia symptoms.

[0004] Brain images such as MRI (Magnetic Resonance Imaging) and PET (positron emission tomography) are used for dementia diagnosis. Representatively, cortical thickness has been used as a dementia-related factor. Summary of the Invention

[0005] Problems to be Solved by the Invention

[0006] However, the cortical thickness may vary due to differences in imaging devices (vendors) used by medical institutions. Therefore, the cortical thickness has certain limitations as an accurate indicator for diagnosing dementia.

[0007] The following technology aims to provide a technology for calculating dementia-related information based on other regions of interest that can be extracted from brain images instead of the cerebral cortex.

[0008] Means for Solving the Problems

[0009] A method for calculating dementia-related metrics from a brain image, including the following steps: an analysis device receives an input of a brain image of a subject; the analysis device distinguishes a region of interest in the brain image; the analysis device calculates the volume or area of the region of interest; and the analysis device calculates a dementia-related metric of the subject based on the volume or area.

[0010] On the other hand, a method for calculating a dementia-related metric based on the volume of the extracerebral cerebrospinal fluid region includes the following steps: an analysis device receives an input of a brain image of a subject; the analysis device distinguishes a region of interest in the brain image; and the analysis device inputs the region of interest into a pre-trained learning model to calculate a dementia-related metric of the subject

[0011] An analysis device for calculating dementia-related metrics, comprising: an input device that receives an input of a subject's brain image; and a calculation device that differentiates regions of interest in the brain image and calculates dementia-related metrics based on the regions of interest. The regions of interest include an extracerebral cerebrospinal fluid region.

[0012] Advantages of the Invention

[0013] The techniques described below provide the degree of brain atrophy centered on an extracerebral cerebrospinal fluid region that is clearly distinguishable from other brain structures in a brain image. Therefore, the techniques described below are not affected by differences in the characteristics of imaging devices and can provide reliable dementia-related information. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 An example of estimating the thickness of the cerebral cortex in a medical image;

[0015] Figure 2 An example of a system for generating dementia-related metrics by analyzing a brain image;

[0016] Figure 3 An example of a process for generating dementia-related metrics by analyzing a brain image;

[0017] Figure 4 An example of a process for calculating dementia-related metrics based on an extracerebral cerebrospinal fluid region and a ventricular region;

[0018] Figure 5 An evaluation result of the correlation of dementia assessment for the contralateral ventricular and extracerebral cerebrospinal fluid regions;

[0019] Figure 6 An evaluation result of the correlation of dementia assessment for an extracerebral cerebrospinal fluid sub-region;

[0020] Figure 7 An evaluation result of the performance of a classifier constructed based on the ventricular and extracerebral cerebrospinal fluid regions;

[0021] Figure 8 For Figure 7 An evaluation result of the performance of a classifier that additionally uses patient information in the model of;

[0022] Figure 9 An evaluation result of the performance of a classifier that uses patient information in the ventricular and extracerebral cerebrospinal fluid regions;

[0023] Figure 10 An example of a process for calculating dementia-related metrics using a learning model;

[0024] Figure 11 An example of an analysis device for calculating dementia-related metrics. DETAILED DESCRIPTION

[0025] The technologies described below can be variously modified and have various embodiments. Specific embodiments are illustrated in the accompanying drawings and described in detail. However, the technologies described below are not limited to a specific factual manner, but should include all changes, equivalents, and alternatives belonging to the ideas and technical scope of the technologies described below.

[0026] Terms such as first, second, A, B, etc. can be used to describe various components, but the components are not limited by these terms and are only used to distinguish one component from another. For example, without departing from the scope of the technologies described below, the first component can be named the second component, and the second component can be named the first component. "And / or" represents a combination of multiple related recited items or one of the multiple related recited items.

[0027] Among the terms used in this specification, if there is no obvious difference in the context, the singular recitation includes the plural meaning. Terms such as "including" indicate the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, rather than precluding the existence or additional possibility of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0028] Before describing the accompanying drawings in detail, it should be clear that the division of components in this specification is only based on the main functions performed by each component. That is, two or more components to be described below in this specification can be combined into one component, or one component can be divided into two or more components and have more detailed functions. In addition, each component to be described below, in addition to its main function, can also perform some or all of the functions of other components, and some of the main functions of each component can be completely performed by other components.

[0029] When performing a method or an operation method, unless the context clearly indicates a specific order, the steps constituting the method can be performed in an order different from the recited order. That is, each process can be performed in the same order as specified, can actually be performed simultaneously, or can be performed in the reverse order.

[0030] Alzheimer's disease is diagnosed based on the thickness of the cerebral cortex obtained from medical images. Figure 1 An example of estimating the thickness of the cerebral cortex in a medical image. Figure 1 An example of measuring or estimating the thickness of the cerebral cortex in an MRI image. In Figure 1In [the figure], the cerebral cortex corresponds to the gray area between the solid white line and the dashed white line. Therefore, only by clearly distinguishing the white area inside the solid white line from the gray area between the solid white line and the dashed white line can the thickness of the cerebral cortex be accurately measured. On the other hand, medical imaging devices are provided by different suppliers. Even for devices produced by the same supplier, the image generation parameters may vary depending on the device type.

[0031] The researchers collected data on the patients visiting their affiliated medical institution (Samsung Seoul Hospital). 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. Among them, the normal group included amyloid-negative subjects, and the dementia group included amyloid-positive subjects.

[0032] [Table 1]

[0033]

[0034]

[0035] The above Table 1 shows the results of measuring the cerebral cortex thickness of the subjects using the imaging devices Archieva and Ingenia of the same manufacturer (Philips) used by the researchers' affiliated medical institution. Aβ(-)NC refers to the normal control group without β-amyloid accumulation, while Aβ(+)ADD refers to the Alzheimer's disease patient group with β-amyloid accumulation. In Aβ(+)ADD, there are significant differences in the cerebral cortex thickness of the MRI images taken by the Achieva and Ingenia devices. However, in Aβ(-)Nc, there are significant differences in the measured values of the temporal and occipital lobes of the two devices. That is, as the researchers expected, even for medical devices of the same supplier, when their types are different, the measured cerebral cortex thicknesses are different. This is because there are differences in the parameter values set in different imaging devices.

[0036] In Figure 1In the figure, it represents a partial area in the extracerebral cerebrospinal fluid (Extracerebral CSF) region. The extracerebral cerebrospinal fluid region is the sulcus region between gyri. The extracerebral cerebrospinal fluid region appears black or very dark in MRI images. That is, compared with the cerebral cortex region that appears gray in MRI, the extracerebral cerebrospinal fluid region is a region with an obvious visual difference. Therefore, it can be inferred that the extracerebral cerebrospinal fluid region is a region that can be relatively accurately distinguished by the image processing technology or segmentation model of a computer device.

[0037] [Table 2]

[0038]

[0039]

[0040] The above Table 2 shows the results of measuring the length (or size) of the extracerebral cerebrospinal fluid region in the population described in Table 1 using the achieva and ingenia devices. The results in Table 2 show that there is no significant difference in the length of the extracerebral cerebrospinal fluid region measured by the two devices in Aβ(+)ADD and Aβ(-)NC. That is, as predicted by the researchers, it can be inferred that compared with the cerebral cortex region, the extracerebral cerebrospinal fluid region has obvious distinguishable image features. The cerebral cortex thickness is an index indicating the degree of brain atrophy. The more severe the brain atrophy, the thinner the cerebral cortex thickness. In addition, the size of the extracerebral cerebrospinal fluid region is also correlated with the degree of brain atrophy. If the brain atrophy is severe, the size of the extracerebral cerebrospinal fluid region increases.

[0041] Dementia in the following text refers to Alzheimer's dementia.

[0042] The following described technology is a technology for diagnosing dementia or predicting the possibility of dementia onset centered on the extracerebral cerebrospinal fluid region.

[0043] The size of the extracerebral cerebrospinal fluid region can be evaluated as the spacing or distance between gyri. Or, the size of the extracerebral cerebrospinal fluid region can be evaluated as the area of a specific sulcus. The size of the area can be calculated from a two-dimensional 2D (dimension) image. Or, the size of the extracerebral cerebrospinal fluid region can be evaluated as the volume of a specific sulcus. The volume can be calculated from a 3D image or a 2D slice.

[0044] Below, the device for analyzing brain images to calculate dementia-related indicators of a subject is named the analysis device. The analysis device can be in the form of a computer device such as a PC, a smart device, a network server, a dedicated chipset for data processing, etc.

[0045] The analysis device can calculate the size or volume of a specific extracerebral cerebrospinal fluid region in a brain image using traditional image processing techniques. Alternatively, the analysis device can calculate the size or volume of a specific extracerebral cerebrospinal fluid region using a model of the deep learning network type. The analysis device can calculate the degree of brain atrophy based on the size or volume of the extracerebral cerebrospinal fluid region. Further, the analysis device can analyze the brain image to calculate the dementia diagnosis or prediction result of the subject.

[0046] The analysis device analyzes the brain image to calculate dementia-related metrics. Among them, the dementia-related metrics correspond to information or factors related to the progression of dementia. The dementia-related metrics can include at least one of the following information: the size or volume of the extracerebral cerebrospinal fluid region, the degree of brain atrophy evaluated by the size or volume of the extracerebral cerebrospinal fluid region, and the presence or absence (or degree of dementia progression) of dementia evaluated by the size / volume of the extracerebral cerebrospinal fluid region (or degree of brain atrophy).

[0047] Figure 2 An example of a system 100 for generating dementia-related metrics by analyzing brain images. In Figure 2 it shows an example where the analysis device is the computer terminal 130 and the server 140.

[0048] The medical imaging device 110 generates a brain image of a patient (e.g., an MRI image). The brain image generated by the medical imaging device 110 can be stored in a separate database such as an EMR (Electronic Medical Record) 120.

[0049] In Figure 2 User A can use the computer terminal 130 to analyze the brain image to obtain dementia-related metrics. The computer terminal 130 can receive the input of the brain image of a specific subject from the medical imaging device 110 or the EMR 120 through a wired or wireless network. In some cases, the computer terminal 130 can be a device physically connected to the medical imaging device 110. The computer terminal 130 extracts the extracerebral cerebrospinal fluid region from the brain image. The computer terminal 130 can calculate dementia-related metrics based on the extracerebral cerebrospinal fluid region. (i) The computer terminal 130 can estimate the size or volume of the extracerebral cerebrospinal fluid region. (ii) The computer terminal 130 can estimate the degree of brain atrophy according to the size or volume of the extracerebral cerebrospinal fluid region. (iii) The computer terminal 130 can estimate the presence or absence of dementia, the likelihood of dementia occurrence, or the degree of dementia progression, etc., according to the size / volume of the extracerebral cerebrospinal fluid region or the degree of brain atrophy. User A can view the analysis results on the computer terminal 130.

[0050] Server 140 can receive brain images of a specific subject from medical imaging device 110 or EMR 120. Server 140 extracts the extracerebral cerebrospinal fluid region from the brain images. Server 140 can calculate dementia-related metrics based on the extracerebral cerebrospinal fluid region. (i) Server 140 can estimate the size or volume of the extracerebral cerebrospinal fluid region. (ii) Server 140 can estimate the degree of brain atrophy according to the size or volume of the extracerebral cerebrospinal fluid region. (iii) Server 140 can estimate whether dementia is present, the likelihood of dementia onset, or the degree of dementia progression, etc., based on the size / volume of the extracerebral cerebrospinal fluid region or the degree of brain atrophy. Server 140 can transmit the results of analyzing the brain images to the terminal of User A. User A can view the analysis results through the user terminal.

[0051] Computer terminal 130 and / or server 140 can store the analysis results in EMR 120.

[0052] Figure 3 An example of process 200 for generating dementia-related metrics by analyzing brain images.

[0053] The analysis device receives an input 210 of the brain image (MRI image) of the subject.

[0054] The analysis device can perform certain preprocessing 220 on the input brain image data. The data preprocessing can be a process of extracting the surface structure or model of the brain structure from the input image to generate a mask. For example, the analysis device can use the CIVET pipeline to extract the entire brain region from the brain MRI image. The analysis device can extract the brain region on the MRI slice. The analysis device can extract brain regions from consecutive MRI slices to extract the brain region in three-dimensional shape. In summary, (i) the analysis device can generate a mask for the entire brain region from the brain image. In addition, (ii) the analysis device can also generate a mask for a specific brain region from the brain image. The specific region can include at least one of the extracerebral cerebrospinal fluid region, the extracerebral cerebrospinal fluid sub-region, the ventricle region, and the ventricle sub-region. The extracerebral cerebrospinal fluid sub-region and the ventricle sub-region will be described in detail later.

[0055] On the other hand, the data preprocessing process for generating the mask can be an optional process. If the image processing technology used can automatically extract the region of interest (ROI), the analysis device may not implement the preparation of the mask.

[0056] The analysis device can segment 230 the entire brain region in the brain image of the subject. The analysis device can use the mask generated during the data preprocessing process to segment the entire brain region. Or, the analysis device can use a segmentation model to segment the entire brain region.

[0057] The analysis device can segment ROI240 across the entire brain region. The analysis device can segment the ROI using the mask generated during the data preprocessing. Alternatively, the analysis device can segment the ROI using a segmentation model. The ROI includes the extracerebral cerebrospinal fluid region. The ROI can include other regions outside the extracerebral cerebrospinal fluid region. In addition, the ROI can be composed of sub-regions that can be extracted from the extracerebral cerebrospinal fluid region. The specific ROI will be detailed later.

[0058] The analysis device can calculate the volume of the segmented ROI250. The analysis device can use commercially available programs or algorithms to calculate the volume of the three-dimensional ROI. Alternatively, the analysis device can also calculate the area of the two-dimensional ROI in the MRI slices. The analysis device can calculate the area of the ROI in all slices or selected specific slices.

[0059] The analysis device can estimate the degree of brain atrophy260 based on the volume or area of the ROI. Figure 3 The dementia-related indicator is the degree of brain atrophy. There is a certain correlation between the volume of the ROI or the area of a specific location and the degree of brain atrophy. The correlation between the volume (or area) of the ROI and the degree of brain atrophy can be pre-tabulated. In this case, the analysis device can estimate the degree of brain atrophy of the subject based on the calculated volume (or area) of the ROI. Alternatively, the analysis device can use a function with the calculated ROI volume (or area) as a variable to estimate the degree of brain atrophy of the subject. The function can be a publicly available mathematical formula at this time. Alternatively, the function can be a mathematical formula obtained through regression analysis at this time.

[0060] The ROI includes the extracerebral cerebrospinal fluid region. Further, the ROI can include at least a part of the individual sub-regions of the extracerebral cerebrospinal fluid region. In addition, the ROI can include the ventricular region that has a relatively distinct difference due to non-white or non-gray in the MRI. Further, the ROI can include at least a part of the individual sub-regions of the ventricular region. The ROI can be any one of various regions, or any one of the possible combinations of various regions. The candidate ROIs are shown in Table 3 below.

[0061] [Table 3]

[0062]

[0063] The possible ROIs are as follows: (i) The ROI can be at least one of the entire extracerebral cerebrospinal fluid region and the entire ventricular region. (ii) In addition, the ROI can be at least one of the entire extracerebral cerebrospinal fluid region and the ventricular sub-region. (iii) In addition, the ROI can be at least one of the entire extracerebral cerebrospinal fluid sub-region and the entire ventricular region. (iv) In addition, the ROI can be at least one of the entire extracerebral cerebrospinal fluid sub-region and the ventricular sub-region.

[0064] Figure 4 This is an example of process 300 for calculating dementia-related metrics based on extracranial cerebrospinal fluid regions and ventricular regions. Figure 4 This is an example of using sub-regions in the ROI to calculate dementia-related metrics.

[0065] The analysis device receives the input 310 of the subject's brain image (MRI image).

[0066] The analysis device can perform certain preprocessing 320 on the input brain image data. The data preprocessing can be a process of extracting the surface structure or model of the brain structure from the input image to generate a mask. The analysis device can generate a mask for the entire brain region from the brain image. In addition, the analysis device can also generate a mask for a specific brain region from the brain image. The specific region can include at least one of extracranial cerebrospinal fluid regions, extracranial cerebrospinal fluid sub-regions, ventricular regions, and ventricular sub-regions.

[0067] On the other hand, the data preprocessing process for generating the mask can be an optional process. If the image processing technology used can automatically extract the ROI, the analysis device may not implement the preparation of the mask.

[0068] The analysis device can segment the entire ROI region 330 in the subject's brain image. The analysis device can use the mask generated during the data preprocessing process to segment the entire brain region. Alternatively, the analysis device can use a segmentation model to segment the entire brain region.

[0069] The analysis device can segment specific ROIs in the entire brain region. The analysis device can use the mask to segment at least one of the extracranial cerebrospinal fluid sub-regions 340. In addition, the analysis device can use the mask to segment at least one of the ventricular sub-regions 350. The analysis device can also use a segmentation model to segment the target ROI.

[0070] The analysis device can calculate the volume of the segmented ROI 360. The analysis device can use commercially available programs or algorithms to calculate the volume of the three-dimensional ROI. Alternatively, the analysis device can also calculate the area of the two-dimensional ROI in the MRI slices. The analysis device can calculate the area of the ROI in all slices or selected specific slices.

[0071] The analysis device can normalize each brain region using the subject's intracranial volume (ICV) 370. The analysis device can make certain corrections to the size or volume of the target ROI based on the ICV. The correction process based on the ICV can be an optional process.

[0072] The analysis device can estimate the degree of brain atrophy based on the volume or area of the ROI 380. The analysis device can estimate the degree of brain atrophy based on the volume or area corrected according to the ICV. For example, the analysis device can normalize the volume by dividing the volume (or area) of the ROI by the ICV.

[0073] Figure 4 The dementia-related index is the degree of brain atrophy. The analysis device can estimate the degree of brain atrophy of the subject by comparing the calculated volume (or area) of the ROI with a pre-prepared reference value. Alternatively, the analysis device can estimate the degree of brain atrophy of the subject using a function with the calculated ROI volume (or area) as a variable.

[0074] The researchers screened the ROIs used to calculate the dementia-related indices.

[0075] The researchers determined the correlation between the lateral ventricle in the ventricular sub-region and the entire extracerebral CSF region and dementia. Figure 5 This is the evaluation result of the correlation of dementia assessment for the contralateral ventricle and the extracerebral CSF region. In Figure 5 A(-)NC is the normal group without β-amyloid accumulation, and A(+)ADD is the Alzheimer's disease patient group with β-amyloid accumulation. From Figure 5 the results, it can be seen that the lateral ventricle region and the extracerebral CSF region can be used as ROIs respectively. (i) The lateral ventricle region can be used as an index to distinguish the normal group and the patient group (P<0.001). In addition, (ii) the entire extracerebral CSF region can also be used as an index to distinguish the normal group and the patient group (P<0.001).

[0076] The researchers determined the correlation between each extracerebral CSF sub-region and dementia. Figure 6 This is the evaluation result of the correlation of dementia assessment for the extracerebral CSF sub-region. The extracerebral CSF sub-region includes the frontal region (frontal, F), temporal region (temporal, T), parietal region (parietal, P), and occipital region (occipital, O). In Figure 6 A(-)NC is the normal group without β-amyloid accumulation, and A(+)ADD is the Alzheimer's disease patient group with β-amyloid accumulation. From Figure 6 the results, it can be seen that each extracerebral CSF sub-region can be used as an ROI (p<0.001).

[0077] The researchers constructed a model (classifier) for calculating dementia-related metrics based on the selected ROIs. The classifier was implemented in the form of a machine learning model. The researchers used 70% of the data of the above-mentioned population as training data and the remaining 30% as validation data. The researchers used glm() in R for logistic regression. However, the classifier can also be implemented in the form of other models, such as deep learning models.

[0078] Figure 7 Performance evaluation results for the classifier constructed based on the lateral ventricle and extra-cerebral cerebrospinal fluid regions. Figure 7 Shows the performance of two models (Model 1 and Model 2). Model 1 and Model 2 are models that calculate dementia-related metrics using only brain images. Model 1 is a model that uses the lateral ventricle and the entire extra-cerebral cerebrospinal fluid region as ROIs. Model 2 is a model that uses the lateral ventricle and sub-regions of the extra-cerebral cerebrospinal fluid as ROIs. The ROC (Area Under the ROC Curve) of Model 1 is 0.808. The sub-regions of the extra-cerebral cerebrospinal fluid used in Model 2 are the frontal region (F), temporal region (T), parietal region (P), and occipital region (O). The ROC (Area Under the ROC Curve) of Model 2 is 0.854. Although the performance of Model 2 is slightly higher than that of Model 1, both Model 1 and Model 2 are sufficient for diagnosing or predicting dementia.

[0079] Figure 8 Performance evaluation results for the classifier that additionally uses patient information in the lateral ventricle and the entire extra-cerebral cerebrospinal fluid region. Figure 8 Shows the performance of two models (Model 3 and Model 4). Model 3 and Model 4 are models that calculate dementia-related metrics using brain images and patient information. Model 3 and Model 4 are models pre-trained using brain images and patient information.

[0080] Model 3 is a model that uses additional patient information (age, gender, education level) on the basis of Model 1. That is, Model 3 is a model that calculates dementia-related metrics using (i) the lateral ventricle and the entire extra-cerebral cerebrospinal fluid region extracted from brain images and (ii) patient information (age, gender, education level). The AUC of Model 3 is 0.829.

[0081] Model 4 is a model that also uses additional patient information (age, gender, education level) and clinical information (APOE e4) on the basis of Model 1. That is, Model 3 is a model that calculates dementia-related indicators using (i) the lateral ventricles and the entire extracerebral cerebrospinal fluid region extracted from brain images, (ii) patient information (age, gender, education level), and (iii) patient clinical information (APOE e4). APOE e4 refers to the genotype information of the gene related to dementia. The AUC of Model 4 is 0.883.

[0082] The performances of both Model 3 and Model 4 are higher than that of Model 1. Therefore, both Model 3 and Model 4 are sufficient for diagnosing or predicting dementia.

[0083] Figure 9 It is the performance evaluation result of the classifier that uses patient information in the lateral ventricles and the extracerebral cerebrospinal fluid region. Figure 9 It represents the performances of two models (Model 5 and Model 6). Model 5 and Model 6 are models that calculate dementia-related indicators using brain images and patient information. Model 5 and Model 6 are models pre-trained using brain images and patient information.

[0084] Model 5 is a model that calculates dementia-related indicators using (i) the ventricular sub-regions and the extracerebral cerebrospinal fluid region (F, P, T, O) extracted from brain images and (ii) patient information (age, gender, education level). The AUC of Model 5 is 0.889.

[0085] Model 6 is a model that calculates dementia-related indicators using (i) the ventricular sub-regions and the extracerebral cerebrospinal fluid sub-regions (F, P, T, O) extracted from brain images, (ii) patient information (age, gender, education level), and (iii) patient clinical information (APOE e4). The AUC of Model 6 is 0.932.

[0086] The performances of both Model 5 and Model 6 are higher than that of Model 2. Therefore, both Model 5 and Model 6 are sufficient for diagnosing or predicting dementia.

[0087] In summary, the ROIs related to dementia-related metrics include: (i) the entire extracerebral cerebrospinal fluid (ECSF) region; (ii) entire sub-regions of the ECSF region; (iii) partial regions within sub-regions of the ECSF region; (iv) the entire ECSF region + lateral ventricles; (v) entire sub-regions of the ECSF region + lateral ventricles; (vi) partial regions within sub-regions of the ECSF region + lateral ventricles, and (vii) at least partial regions within sub-regions of the ECSF region + sub-regions of the ventricles. On the other hand, experiments have shown that the performance of a classifier using a sub-region of the ECSF region as the ROI is slightly higher than that of a classifier using the entire ECSF region. In addition, a classifier using only brain MRI also has a sufficiently significant performance. Further, when using additional information such as patient information in addition to brain MRI images, the performance of the classifier can be improved.

[0088] Figure 10 An example of process 400 for calculating dementia-related metrics using a learning model. Figure 10 This is the case where both the process of extracting ROIs from brain MRI images and the process of predicting dementia-related metrics based on the ROIs use a learning model.

[0089] The analysis device receives an input 410 of a brain image (MRI image) of a subject.

[0090] The analysis device can input the input brain image into a pre-trained segmentation model 420. The segmentation model can be implemented as various types or structures of models. For example, the segmentation model can be a U-net-based model. The segmentation model can segment various ROIs according to its type or learning process. The segmentation model can segment ROIs from three-dimensional brain images. Alternatively, the segmentation model can also segment ROIs from a single two-dimensional slice. As described above, the ROIs can use partial regions in various regions. The segmentation model can be pre-constructed according to specific ROIs. For example, the segmentation model can include various models: (i) a model for segmenting the entire extracerebral cerebrospinal fluid region; (ii) a model for segmenting at least a part of the sub-region of the extracerebral cerebrospinal fluid; (iii) a model for segmenting the entire extracerebral cerebrospinal fluid region + the lateral ventricle region; (iv) a model for segmenting at least a part of the sub-region of the extracerebral cerebrospinal fluid + the lateral ventricle region, etc. The segmentation model can be pre-constructed as a model for segmenting a certain one of various ROIs according to the training data and the training process.

[0091] The analysis device can obtain the result (ROI discrimination) output by the segmentation model 430.

[0092] The analysis device inputs the ROI into a pre-trained classification model 440. The classification model is a model that inputs an image or images and patient information to calculate dementia-related metrics.

[0093] The classification model is a machine learning model. Therefore, the classification model can be one of various types of models. For example, the classification model can be a model implemented in one of the following ways: decision tree, random forest (RF), KNN (K-nearest neighbor), Bayes, SVM (support vector machine), ANN (artificial neural network), regression model, etc. There are also various types of models in ANN. For example, the classification model can be a model based on CNN (Convolutional Neural Network).

[0094] The analysis device can obtain the dementia-related index 450 output by the classification model. The dementia-related index can be information such as the volume of the ROI, the degree of brain atrophy, or the degree of dementia progression.

[0095] The classification model can generate the dementia-related index only using the ROI. In addition, the classification model can input the ROI and patient information (age, gender, education level, etc.) into the classification model to generate the dementia-related index. Further, the classification model can also input the ROI, patient information (age, gender, education level, etc.) and clinical information (APOE e4, etc.) into the classification model to generate the dementia-related index.

[0096] On the other hand, different from Figure 10 only one of the process of extracting the ROI or the process of predicting the dementia-related index uses the learning model. That is, (i) the analysis device can use the segmentation model to extract the ROI and calculate the volume of the extracted ROI to estimate the degree of brain atrophy. Calculating the volume and determining the degree of brain atrophy are as Figures 3 to 4 described. Or, (ii) the analysis device can also use the mask to extract the ROI and input the extracted ROI into the Figure 10 classification model to calculate the dementia-related index. The process of using the mask to extract the ROI is as Figures 3 to 4 described.

[0097] Figure 11 is an example of the analysis device 500 for calculating the dementia-related index. The analysis device 500 corresponds to the above analysis device ( Figure 1 130 and 140 in). The analysis device 500 can be physically implemented in various forms. For example, the analysis device 500 can be in the form of a computer device such as a PC, a server on the network, a dedicated chipset for data processing, etc.

[0098] The analysis device 500 can include a storage device 510, a memory 520, a computing device 530, an interface device 540, a communication device 550, and an output device 560.

[0099] The storage device 510 can store the brain image (MRI image) of the subject generated by the medical imaging device.

[0100] The storage device 510 can store the patient information and clinical information of the subject. The patient information and clinical information are as described above.

[0101] The storage device 510 can store the code or program for calculating dementia-related metrics from the brain image.

[0102] The storage device 510 can store the mask extracted from the brain image.

[0103] The storage device 510 can store the segmentation model for extracting the ROI from the brain image. The segmentation model can be a pre-trained model.

[0104] The storage device 510 can store the classification model for obtaining the input of the ROI to calculate dementia-related metrics. The classification model can be a pre-trained model.

[0105] The storage device 510 can store the dementia-related metrics of the subject.

[0106] The memory 520 can store the data and information generated during the process of the analysis device 500 calculating the dementia-related metrics from the brain image.

[0107] The interface device 540 is a device for receiving certain command and data inputs from the outside. The interface device 540 can receive the brain image of the subject from the physically connected device or an external storage device. The interface device 540 can receive the patient information and / or clinical information of the subject from the physically connected device or an external storage device. The interface device 540 can also transfer the dementia-related metrics calculated based on the brain image to an external object.

[0108] The communication device 550 refers to a configuration for receiving and transmitting certain information through a wired or wireless network. The communication device 550 can receive the brain image of the subject from an external object. The communication device 550 can receive the patient information and / or clinical information of the subject from an external object. The communication device 550 can also transfer the dementia-related metrics calculated based on the brain image to an external object such as a user terminal.

[0109] The interface device 540 and the communication device 550 are configured to send or receive certain data with a user or other physical objects, so they can be collectively referred to as input / output devices. When limited to the information or data input function, the interface device 540 and the communication device 550 can also be called input devices.

[0110] The output device 560 is a device for outputting certain information. The output device 560 can output the interface required for the data processing process, brain images, ROIs segmented from the brain images, dementia-related metrics calculated based on the ROIs, etc.

[0111] The computing device 530 can perform certain preprocessing on the brain images of the subject. The data preprocessing process is as Figure 3 described. The computing device 530 can generate masks for segmenting the entire brain region and ROIs through the data preprocessing process.

[0112] The computing device 530 can use the mask of the entire brain region to segment the entire brain region from the brain image. Further, the computing device 530 can use the mask of a specific ROI to segment the target ROI from the entire brain region. As described above, the ROI can be of various types. For example, the ROI can be any one of the entire extracerebral cerebrospinal fluid region, extracerebral cerebrospinal fluid sub-region, entire extracerebral cerebrospinal fluid region + lateral ventricle region, and extracerebral cerebrospinal fluid sub-region + lateral ventricle region.

[0113] The computing device 530 can also use a segmentation model to segment ROIs from the input brain images.

[0114] The computing device 530 can calculate the volume or area of the segmented ROIs. The computing device 530 can use any commercial program for calculating the volume of the brain region to calculate the volume or area of the ROIs. Further, the computing device 530 can normalize the initially calculated ROI volume based on the ICV.

[0115] The computing device 530 can estimate the degree of brain atrophy based on the final ROI volume or area. Alternatively, the computing device 530 can estimate the dementia degree of the subject based on the final ROI volume or area. Alternatively, the computing device 530 can estimate the dementia degree of the subject based on the degree of brain atrophy of the subject.

[0116] The computing device 530 can input the ROIs into a pre-constructed classification model to calculate dementia-related metrics. In addition, the computing device 530 can input the ROIs and patient information (age, gender, education level, etc.) into the classification model to generate dementia-related metrics. Further, the computing device 530 can input the ROIs, patient information (age, gender, education level, etc.), and clinical information (APOEe4, etc.) into the classification model to generate dementia-related metrics.

[0117] The computing device 530 can be a device such as a processor, AP, a chip embedded with a program, etc. that can process data and perform certain calculations.

[0118] In addition, the above-mentioned brain image processing method, dementia-related index calculation method, or dementia prediction method can be implemented as a program (or application program) including an executable algorithm executable on a computer. The program can be stored in a temporary or non-temporary readable medium (non-transitory computer readable medium).

[0119] Different from media such as registers, caches, and memories that store data in the short term, a non-temporary readable medium is a medium that stores data semi-permanently and can be read by a device. Specifically, the above various applications or programs can be stored in a non-temporary readable medium, such as a CD, DVD, hard disk, Blu-ray disc, USB, memory card, ROM (read-only memory), PROM (programmable read only memory), EPROM (Erasable PROM, EPROM), or EEPROM (Electrically EPROM), or flash memory, etc.

[0120] Temporary readable media refer to various RAMs, such as static RAM (Static RAM, SRAM), dynamic RAM (Dynamic RAM, DRAM), synchronous DRAM (Synchronous DRAM, SDRAM), double data rate SDRAM (Double Data Rate SDRAM, DDR SDRAM), enhanced SDRAM (Enhanced SDRAM, ESDRAM), synchronous DRAM (Synclink DRAM, SLDRAM), and direct rambus RAM (Direct Rambus RAM, DRRAM), etc.

[0121] This embodiment and the accompanying drawings only clearly illustrate some of the technical ideas contained in the foregoing technology. Obviously, all variations and specific embodiments that can be easily deduced by those skilled in the art within the scope of the technical ideas described in the foregoing technology and contained in the accompanying drawings are included within the scope of the claims of the foregoing technology.

Claims

1. A method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid regions, comprising the following steps: An analysis device receives an input of a subject's brain image; The analysis device differentiates regions of interest in the brain image; The analysis device calculates the volume or area of the regions of interest; and The analysis device calculates the dementia-related indicators of the subject based on the volume or area; Wherein, The regions of interest include extracerebral cerebrospinal fluid regions.

2. The method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid regions according to claim 1, wherein the dementia-related indicators include at least one of the degree of brain atrophy, dementia or not, and the degree of dementia progression.

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

4. The method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid regions according to claim 1, wherein the regions of interest further include ventricular regions; the ventricular regions include lateral ventricles.

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

6. A method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid regions, comprising the following steps: An analysis device receives an input of a subject's brain image; The analysis device differentiates regions of interest in the brain image; and The analysis device inputs the regions of interest into a pre-trained learning model to calculate the dementia-related indicators of the subject; Wherein, The regions of interest include extracerebral cerebrospinal fluid regions.

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 image into a pre-trained segmentation model to differentiate the regions of interest.

8. The method for calculating dementia-related indicators based on the volume of extracerebral cerebrospinal fluid regions according to claim 6, wherein the regions of interest include all of the extracerebral cerebrospinal fluid regions and ventricular regions.

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

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

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

12. An analysis device for calculating dementia-related indicators comprising: an input device that receives the input of the brain image of the subject; and a calculation device that differentiates the region of interest in the brain image and calculates dementia-related indicators based on the region of interest; the region of interest includes the extracerebral cerebrospinal fluid region.

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

14. The analysis device for calculating dementia-related indicators according to claim 12, wherein the calculation device inputs the brain image into a pre-trained segmentation model to differentiate the region of interest.

15. The analysis device for calculating dementia-related indicators according to claim 12, wherein the calculation device calculates the dementia-related indicators based on the volume or area of the region of interest.

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

17. The analysis 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 ventricle region.

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

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

Images