Dementia information calculation method and analysis device using 2D MRI

Abstract

The method for calculating dementia information using two-dimensional MRI includes a step of an analysis device receiving an input of a 2D (dimension) MRI (Magnetic Resonance Imaging) slice of a subject, a step of the analysis device inputting the 2D MRI slice into a segmentation model to extract a region of interest, a step of the analysis device inputting pixel information of the region of interest into a first learning model that has been pre-trained to predict the cerebral cortical thickness of at least one of the regions of interest and the volume of at least one of the regions of interest, and a step of the analysis device inputting the cerebral cortical thickness of the at least one region and the volume of the at least one region into a second learning model that has been pre-trained to calculate dementia information of the subject.

Description

[Technical field]

[0001] The technology described below is a technique for calculating dementia-related information using two-dimensional MRI. [Background technology]

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

[0003] Brain imaging such as magnetic resonance imaging (MRI) and positron emission tomography (PET) is used to diagnose dementia. Cortical thickness is typically used as a factor related to dementia. Summary of the Invention [Problem to be solved by the invention]

[0004] Previously, early diagnosis of dementia was performed using brain images such as 3D (dimension) MRI and PET-CT. However, 3D MRI requires high-spec MRI equipment, which is time-consuming and expensive. Furthermore, 2D MRI scanners used in general health screening centers and primary hospitals generate about 20 images, making accurate diagnosis difficult.

[0005] The technology described below seeks to provide a technique for calculating dementia-related information or predicting dementia using a small number of 2D MRI images. [Means for solving the problem]

[0006] The method for calculating dementia information using two-dimensional MRI includes a step of an analysis device receiving 2D MRI slices of a subject, a step of the analysis device inputting the 2D MRI slices into a segmentation model to extract regions of interest, a step of the analysis device inputting pixel information of the regions of interest into a first learning model that has been pre-trained to predict the cortical thickness of at least one of the regions of interest and the volume of at least one of the regions of interest, and a step of the analysis device inputting the cortical thickness of the at least one region and the volume of the at least one region into a second learning model that has been pre-trained to calculate dementia information of the subject.

[0007] The method for calculating dementia information using two-dimensional MRI includes a step of an analysis device receiving 2D MRI slices of a subject, a step of the analysis device inputting the 2D MRI slices into a segmentation model to extract regions of interest, a step of the analysis device inputting pixel information of the regions of interest into a first pre-trained learning model to predict the volume of at least one of the regions of interest, and a step of the analysis device inputting the volume of the at least one region into a second pre-trained learning model to calculate dementia information of the subject.

[0008] The region of interest includes at least one of frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricles, hippocampal formation, and extracerebral cerebrospinal fluid regions. Effect of the Invention

[0009] The technique described below uses a learning model that analyzes a small number of 2D MRI images to calculate dementia-related information. The technique described below can provide information useful for diagnosing dementia based on brain MRIs generated by low-end 2D MRI scanners. [Brief description of the drawings]

[0010] [Figure 1] This is an example of the entire process of calculating a subject's dementia information using 2D brain MRI.

[0011] [Diagram 2] This is an example of the process of training a segmentation model to extract regions of interest in brain images.

[0012] [Diagram 3] 1 is an example of the learning process of a learning model that predicts the thickness of the cerebral cortex and the volume of a brain region based on a region of interest.

[0013] [Figure 4] 13 is an example of a learning process of a learning model that calculates dementia information based on the volume of a brain region.

[0014] [Diagram 5] 1 is an example of the learning process of a learning model that calculates dementia information using the thickness of the cerebral cortex and the volume of a brain region.

[0015] [Figure 6] 1 is an example of an analysis device that calculates dementia information of a subject using brain images. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] The technology described below can be modified in various ways and can have various embodiments, and a specific embodiment will be described in detail with reference to the drawings. However, this is not intended to limit the technology described below to a specific embodiment, and it should be understood that the technology described below includes all modifications, equivalents, and alternatives included in the spirit and technical scope of the technology described below.

[0017] Terms such as first, second, A, B, etc. may be used to describe various components, but the components are not limited by the terms and are used only to distinguish one component from another. For example, a first component may be named a second component and similarly a second component may be named a first component without departing from the scope of the technology described below. The term and / or includes a combination of multiple associated listed items or any item of multiple associated listed items.

[0018] In the terms used in this specification, the singular expression should be understood to include the plural expression unless the context clearly indicates otherwise, and the term "comprise" or the like should be understood to mean the presence of a stated feature, number, step, operation, component, part, or combination thereof, but not to exclude the possibility of the presence or addition of one or more other features, numbers, step operations, components, parts, or combinations thereof.

[0019] Before going into detailed description of the drawings, it should be made clear that the components in this specification are merely classified according to the main function of each component. That is, two or more components described below may be combined into one component, or one component may be divided into two or more components according to more specific functions. It goes without saying that each component described below may additionally perform some or all of the functions of other components in addition to its own main function, and some of the main functions of each component may be exclusively performed by a different component.

[0020] Furthermore, in performing a method or method of operation, the steps of the method may occur in a different order than that stated, unless the context clearly dictates a particular order, i.e., the steps may occur in the same order as stated, may be performed substantially simultaneously, or may be performed in the reverse order.

[0021] Hereinafter, dementia includes Alzheimer's disease.

[0022] The techniques described below are techniques for calculating dementia information using 2D brain MRI. The techniques described below include techniques for diagnosing or predicting dementia using dementia information extracted from 2D brain MRI.

[0023] Dementia information refers to information for diagnosing or predicting dementia. For example, dementia information includes whether or not a subject has dementia, whether or not they are at high risk for dementia (possibility of developing dementia in the future), dementia-related indicators, etc. Dementia-related indicators are those that pay attention to diagnosing or predicting dementia, and may include various research scores, the degree of brain atrophy, etc.

[0024] Hereinafter, a device that analyzes brain MRI and calculates dementia information for a subject is referred to 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, a chipset dedicated to data processing, etc.

[0025] The analysis device can predict the presence or absence of dementia based on brain images using a number of learning models. The learning models are machine learning models. Therefore, the classification model can be any one of various types of models. For example, the learning model can be a model implemented in any one of the following ways: decision tree, random forest (RF), K-nearest neighbor (KNN), Naive Bayes, support vector machine (SVM), deep neural network (DNN), regression model, etc. There are various types of DNN models. For example, DNN includes a segmentation model that extracts regions of interest from an image, and a classification model that performs classification based on image features.

[0026] FIG. 1 is an example of an overall process 100 for calculating dementia information of a subject using 2D brain MRI.

[0027] The analysis device can calculate dementia information for a subject based on a small number of 2D MRI slices. In this case, the number of 2D MRI slices can be the number calculated by a general 2D MRI scanner. The researchers built a learning model using 20 2D MRI slices. Therefore, for the convenience of the following explanation, the number of 2D MRI slices is assumed to be 20.

[0028] The analyzer receives a 2D brain MRI of the subject 110. For example, the analyzer can receive 20 slices of the 2D MRI of the subject.

[0029] The analyzer extracts regions of interest (ROIs) from the input 2D MRI slices (120). The analyzer can extract multiple ROIs that are set using a pre-trained segmentation model. The ROIs can be regions for calculating the volume of a particular region. Moreover, the ROIs can include regions for calculating the thickness of the cerebral cortex.

[0030] The region of interest may include at least one of the following: frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricle, hippocampus, and extracerebral cerebrospinal fluid.

[0031] The analyzer can use the extracted regions of interest to calculate certain brain imaging indices (130).

[0032] The brain image index may include the number of pixels in the extracted region of interest. The brain image index may also include total brain size, which may be calculated from the size of the region of the brain image. Brain size may be calculated by adding up the size of the extracerebral cerebrospinal fluid, the seven regions of interest, and the white matter of the regions of interest. Meanwhile, various methodologies for determining brain size have been researched, and the analysis device may calculate the subject's brain size using one of various image processing techniques.

[0033] Meanwhile, the analysis device may normalize the region of interest to a preset size. For example, the analysis device may calculate the size of the entire brain and normalize the size of the region of interest to a constant size. The analysis device may calculate brain imaging indices such as the number of pixels based on the normalized region of interest.

[0034] The analyzer can utilize the pre-trained model to estimate the volume of the brain region (140), or the analyzer can utilize the pre-trained model to estimate the cortical thickness and the volume of the brain region (140).

[0035] The analysis device can input the brain imaging indices into the learning model to predict the thickness of the cerebral cortex in the region of interest. The analysis device can predict the thickness of the cerebral cortex in each region of interest using multiple learning models. For example, the analysis device can calculate the thickness of the cerebral cortex in each region using separate learning models for the frontal, temporal, parietal, and occipital regions.

[0036] The analysis device may also predict the volume of a brain region using a separate learning model. The analysis device may predict the volume of a brain region by inputting brain image indices into the learning model. The analysis device may predict the volume of each region of interest using multiple learning models. For example, the analysis device may calculate the volume of each region by individually using a learning model for the lateral ventricles, a learning model for the hippocampus, and a learning model for the extracerebral cerebrospinal fluid region.

[0037] Meanwhile, the analysis device can calculate the thickness of the cerebral cortex or the volume of the brain region by using the subject's clinical information (gender, age, etc.) in addition to the brain image. In this case, each learning model must be trained in advance to calculate the thickness of the cerebral cortex or the volume of the brain region using clinical information in addition to the brain MRI.

[0038] The analysis device may calculate dementia information for the subject based on the cortical thickness and volume of the region of interest (150), or the analysis device may calculate dementia information for the subject based only on the volume of the region of interest (150).

[0039] The dementia information calculation process in Figure 1 uses multiple learning models to calculate the dementia information of the subject. The researchers constructed multiple learning models for the above process and verified the performance of each individual model and the dementia information calculation performance. Below, we explain the construction process of the learning models that the researchers used to calculate the dementia information of the subject.

[0040] [Table 1]

[0041] The researchers used data from 980 individuals from the population to build a segmentation model. The 980 individuals' data were 3D MRI data generated using a 3.0 T MRI scanner (Philips 3.0T Achieva). The researchers used the 3D segmentation masks in the CIVET pipeline to perform structural image analysis. In CIVET, the thickness of the cerebral cortex is calculated as the Euclidean distance between the inner and outer surfaces of the cerebral cortex. The regions of interest included (i) the frontal, temporal, parietal and occipital grey matter, (ii) the lateral ventricles, (iii) the hippocampal formation and (iv) the extracerebral cerebrospinal fluid regions. The researchers used the regions of interest extracted from the data from the 980 individuals as ground truth for training the segmentation model.

[0042] 2 shows an example of a process 200 for training a segmentation model for extracting a region of interest in a brain image. Hereinafter, the model training process will be described as being performed by a training device. The training device refers to a computer device capable of processing image data and performing the training process of a machine learning model.

[0043] The learning device learns a segmentation model using MRI images of multiple subjects. However, Figure 2 shows an example of the process of learning a segmentation model using the MRI of a single subject.

[0044] The learning device receives as input (210) a 3D MRI of a particular subject from the population described above.

[0045] The learning device prepares a learning dataset (220). The learning dataset is composed of input slices (original slices) for each of a number of slices and correct answer images for the corresponding slices. The learning device selects 20 slices from the 480 slices of the 3D MRI. The learning device can select slices from the 480 slices that match the 20 slices acquired from the 2D MRI scanner. The learning device also receives input of correct answer images for the selected 20 slices. The correct answer image is an image in which the region of interest has been segmented for the input slice. The correct answer image can be an image in which a medical imaging expert has segmented and displayed the region of interest (annotated).

[0046] The learning device performs learning of a segmentation model using the learning data (230). The learning device can learn the segmentation model using input slices and a ground truth image for 20 slices. The segmentation model is trained to extract (segment) a region of interest in the ground truth image when the input slices are input.

[0047] The segmentation model can be a semantic segmentation model. For example, the segmentation model can be a fully convolutional network (FCN)-based model such as U-net. Researchers have used a U-net-based segmentation model to extract regions of interest from 2D MRI slices.

[0048] The segmentation model may be predefined for each region of interest. As described above, the region of interest may include at least one of frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricles, hippocampal formation, and extracerebral cerebrospinal fluid region.

[0049] Alternatively, the segmentation model may be trained to extract multiple regions of interest. To accurately extract multiple regions of interest with different features, researchers performed curriculum learning. Curriculum learning is a learning strategy that mimics the human learning process by first learning easy data and then learning increasingly difficult data.

[0050] The researchers built a segmentation model using the following five learning sequences and verified its performance:

[0051] [Table 2]

[0052] The performance of the segmentation model was measured using IoU (intersection over union) and DSC (dice similarity coefficient), which are calculated based on the correct answer and the region of interest extracted by the segmentation model. IoU and DSC can be expressed by the following Equation 1 and Equation 2, respectively.

[0053] [Mathematical formula 1]

[0054] IoU=A overlap / (A GT +A pred -A overlap )

[0055] [Mathematical formula 2]

[0056] DSC=2*A overlap / (A GT +A pred )

[0057] In the above mathematical formula, A GT is the correct answer, A pred is the region of interest (prediction region) divided by the segmentation model, A overlap A GT and A pred This is an overlapping area.

[0058] The segmentation model showed the IoU and DSC for the following six regions of interest as shown in Table 3. The segmentation model showed the highest performance in extracting the lateral ventricle.

[0059] [Table 3]

[0060] Table 4 below shows the extraction performance of the segmentation model for the extracerebral cerebrospinal fluid region. The segmentation model showed the highest performance in the lateral ventricle region.

[0061] [Table 4]

[0062] Table 5 below shows the results of analyzing the performance for the learning order in Table 2. The learning method in which the lateral ventricle is learned first showed slightly better performance than the other learning orders. Therefore, it can be seen that the technique of segmenting the region of interest and performing curriculum learning is meaningful for the performance of the segmentation model.

[0063] [Table 5]

[0064] FIG. 3 is an example of a learning process 300 of a learning model that predicts the thickness of the cerebral cortex and the volume of a brain region based on a region of interest. A learning model that predicts the thickness of the cerebral cortex may be prepared in advance for each region of interest. In addition, a learning model that predicts the volume of a brain region may also be prepared in advance for each region of interest. In FIG. 3, the learning model can predict the thickness of the cerebral cortex or the volume of a brain region based on each region of interest. FIG. 3 illustrates an integrated learning process for each region of interest.

[0065] The learning in Fig. 3 is premised on the fact that the aforementioned region of interest has been extracted from the brain MRI of a subject belonging to the population. In other words, the learning device can use the region of interest calculated using the segmentation model in Fig. 2.

[0066] The learning device extracts input data based on the regions of interest extracted from the brain MRI (310). The input data can include brain imaging indices and clinical information extracted from the regions of interest extracted by the segmentation model.

[0067] Brain image indices include size information for each region of interest (frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricles, hippocampus, and extracerebral cerebrospinal fluid region). The size of a region of interest may be determined by the number of pixels in the image of the region of interest. That is, the size of a region of interest may be the total number of pixels belonging to the region. Meanwhile, the size of a region of interest may be normalized based on the size of the entire brain. Normalization is the process of adjusting the brain size, which varies from subject to subject, to a uniform size.

[0068] In addition, the brain imaging indices may further include brain size, which may be calculated by summing the sizes of the extracerebral cerebrospinal fluid region, the seven regions of interest, and the white matter of the regions of interest.

[0069] The total number of pixels in a particular region of interest may be the total number of pixels in the region of interest for each of the 20 slices. With respect to the frontal gray matter, the total number of pixels in the frontal gray matter may be the total number of pixels in the frontal gray matter region for each of the 20 slices. Assume that the indices of the 20 slices are 0 to 19. The total number of pixels in the frontal gray matter may be the total number of pixels in the frontal gray matter region of slice 0, the number of pixels in the frontal gray matter region of slice 1, ..., the number of pixels in the frontal gray matter region of slice 19. Furthermore, the number of pixels in the region of interest may be the average number of pixels in all slices.

[0070] The clinical information may include the subject's age and sex, and may include the APOE4 genotype.

[0071] The researchers used combined pixel counts of regions of interest, brain size and clinical information as all input data.

[0072] The learning device performs a process of learning a learning model using the extracted input data (320). The learning device repeats the learning process of the learning model using input data extracted from the brain MRIs of multiple subjects.

[0073] The learning device performs learning by inputting the extracted input data into the learning model and comparing the value (predicted value) output by the learning model with the correct answer. The learning model is trained to predict the thickness of the cerebral cortex of a specific region of interest or the volume of a specific region of interest. The correct answer is information calculated from MRI and PET-CT of subjects belonging to the population (see Table 1).

[0074] The learning model can be constructed as an individual model depending on the information to be predicted. The learning model is a machine learning model and can be embodied as one of various types of models. The researchers used a regression model. The researchers individually constructed a model to predict the thickness of the frontal cerebral cortex (frontal model), a model to predict the thickness of the temporal cerebral cortex (temporal model), a model to predict the thickness of the parietal cerebral cortex (parietal model), a model to predict the thickness of the occipital cerebral cortex (occipital model), a model to predict the volume of the lateral ventricles (lateral ventricle model), a model to predict the volume of the hippocampal formation (hippocampal formation model), and a model to predict the volume of the extracerebral cerebrospinal fluid area (extracerebral cerebrospinal fluid model).

[0075] For example, the frontal model can predict the thickness of the cerebral cortex in the frontal region based on the input of the sum of the pixels of the frontal gray matter and clinical information, and the frontal model can predict the thickness of the cerebral cortex in the frontal region based on the input of the sum of the pixels of the frontal gray matter, brain size, and clinical information.

[0076] For example, the extracerebral cerebrospinal fluid model can predict the volume of the extracerebral cerebrospinal fluid region by inputting the total number of pixels of the extracerebral cerebrospinal fluid region and clinical information. Also, the extracerebral cerebrospinal fluid model can predict the volume of the extracerebral cerebrospinal fluid region by inputting the total number of pixels of the extracerebral cerebrospinal fluid region, brain size, and clinical information.

[0077] The researchers constructed models that use brain size as input data and models that do not, and verified the performance of the models. Table 6 below shows the models that evaluated the performance of the learning models constructed by the researchers. The performance indicator is the Pearson Correlation Coefficient. In Table 6, Model 1 is a model that does not use brain size, and Model 2 is a model that uses brain size. It can be seen that the performance of Model 1 or Model 2 is slightly higher depending on the region of interest. Therefore, models that predict the thickness of the cerebral cortex and the volume of brain regions can selectively use Model 1 or Model 2 depending on the region of interest. However, there was not much difference in performance between Model 1 and Model 2.

[0078] [Table 6]

[0079] The researchers verified the performance of the learning model to predict the cortical thickness or volume of the region of interest. The researchers used the learning model to compare the predicted values ​​based on 2D MRI with the correct values ​​measured with 3D MRI. The results of the comparison are shown in Table 7 below. The cortical thickness and volume predicted by the learning model showed a significant correlation with the actual correct values. In particular, the volume showed a high correlation.

[0080] The researchers verified the performance of the learning model to predict the cortical thickness or volume of the region of interest. The researchers used the learning model to compare the predicted values ​​based on 2D MRI with the correct values ​​measured with 3D MRI. The results of the comparison are shown in Table 7 below. The cortical thickness and volume predicted by the learning model showed a significant correlation with the actual correct values. In particular, the volume showed a high correlation.

[0081] [Table 7]

[0082] FIG. 4 shows an example of a learning process 400 of a learning model that calculates dementia information based on the volume of a brain region. The learning model that finally calculates dementia information can be embodied as one of various machine learning model types. Researchers have embodied a model that calculates dementia information using a deep learning model.

[0083] The learning in Fig. 4 is premised on the acquisition of the thickness of the cerebral cortex, the volume of a specific brain region, and clinical information for subjects belonging to a population. For example, the learning device can use the volume of a specific brain region by region of a subject calculated using the learning model in Fig. 3.

[0084] The learning device can extract input data required for learning (410). The learning device extracts the volume of a specific brain region extracted from a 2D brain MRI of a subject belonging to a population. The volume of the brain region can include the volume of the lateral ventricle, the hippocampus, and the extracerebral cerebrospinal fluid region. The learning device also acquires clinical information of the subject. The clinical information can include at least one of age, sex, and APOE4 genotype. The learning data can include the volume of the specific brain region calculated using the learning model of FIG. 3. Alternatively, the learning data can include the thickness of the cerebral cortex and the volume of the specific brain region calculated from the brain image of an actual subject. The researchers used the thickness of the cerebral cortex and the volume of the specific brain region calculated from the results of 3D MRI analysis of the actual subject as learning data.

[0085] The learning device performs a process of learning a learning model using the extracted input data (420). The learning device repeats the learning process of the learning model using input data extracted from a number of subjects.

[0086] The learning device performs learning by inputting the extracted input data into a learning model and comparing the value (predicted value) output by the learning model with the correct answer. The learning model is trained to calculate dementia information of the subject. For example, the learning model can output a binary classification result on whether or not the subject has dementia. In this case, the correct answer is information on whether or not the subjects belonging to the population have dementia (see Table 1).

[0087] FIG. 5 shows an example of a learning process 500 of a learning model that calculates dementia information using the thickness of the cerebral cortex and the volume of a brain region. The learning model that finally calculates dementia information can be embodied as one of various machine learning model types. Researchers have embodied a model that calculates dementia information using a deep learning model.

[0088] The learning in Fig. 5 is premised on the acquisition of the cortical thickness, the volume of a specific brain region, and clinical information for subjects belonging to a population. For example, the learning device can use the cortical thickness by region and the volume of a specific brain region of a subject calculated using the learning model in Fig. 3.

[0089] The learning device can extract input data required for learning (510). The learning device extracts the thickness of the cerebral cortex and the volume of a specific brain region extracted from the brain MRI of the subjects belonging to the population. The thickness of the cerebral cortex can include at least one of the thicknesses of the frontal, temporal, parietal, and occipital cerebral cortex. The volume of the brain region can include at least one of the volumes of the lateral ventricle, the hippocampal formation, and the extracerebral cerebrospinal fluid region.

[0090] The learning device also acquires clinical information of the subject. The clinical information may include at least one of age, sex, and APOE4 genotype. The learning data may include the thickness of the cerebral cortex and the volume of a specific brain region calculated using the learning model of FIG. 3. Alternatively, the learning data may include the thickness of the cerebral cortex and the volume of a specific brain region calculated from brain images of an actual subject. The researchers used the thickness of the cerebral cortex and the volume of a specific brain region calculated from the results of 3D MRI analysis of an actual subject as learning data.

[0091] The learning device performs a process of learning a learning model using the extracted input data (520). The learning device repeats the learning process of the learning model using input data extracted from a number of subjects.

[0092] The learning device performs learning by inputting the extracted input data into a learning model and comparing the value (predicted value) output by the learning model with the correct answer. The learning model is trained to calculate dementia information of the subject. For example, the learning model can output a binary classification result on whether or not the subject has dementia. In this case, the correct answer is information on whether or not the subjects belonging to the population have dementia (see Table 1).

[0093] The researchers built multiple learning models using different input data sets. The researchers used data from 924 people (80%) of the population as learning data and 196 people (20%) as validation data. The researchers divided the model groups into groups that used significantly different ROIs, and within each group, they built models using various input data separately. The multiple learning models built by the researchers are shown in Table 8 below. Table 8 shows the model groups that used four different ROIs and the learning data used to build the different models in each group. The models in Table 8 below are models that predict whether a subject has dementia (dementia or normal). In other words, in Figure 4, the learning model corresponds to a model that binary classifies the subject as having dementia or being normal.

[0094] [Table 8]

[0095] In Table 8, ROIs are indicated by abbreviations. F stands for frontal, T stands for temporal, P stands for parietal, O stands for occipital, L stands for lateral ventricle, H stands for hippocampus, and E stands for extracerebral cerebrospinal fluid. Meanwhile, the learning model may be a model that uses the thickness and / or volume of a part (at least one) of the above-mentioned regions of interest as input data, unlike Table 8. In other words, various types of models can be constructed depending on which region of interest is used.

[0096] The researchers verified the constructed learning model. During the verification process, the researchers used the cortical thickness and brain region volume predicted by the learning model in Figure 3 as input data for the learning model in Figure 4. The researchers performed 10-fold cross-validation. The verification results are shown in Table 9 below. The verification was performed by comparing the results predicted by the learning model with the correct answer. The performance indicators used were AUC (area under the receiver operating characteristic curve) and AUPRC (area under the precision-recall curve).

[0097] The researchers compared the performance of a conventional 3D MRI-based classification model with the aforementioned 2D MRI-based learning model. The conventional 3D MRI-based classification model used a previously studied model (Rebsamen, M., et al., Direct cortical thickness estimation using deep learning-based anatomy segmentation and cortex parcellation. Human Brain Mapping, 2020. 41(17): p. 4804-4814.).

[0098] Looking closely at the results in Table 9, there is not much difference in performance between the 3D MRI-based model and the 2D MRI-based model. Therefore, it can be said that the 2D MRI-based model developed by the researchers is also quite useful for predicting dementia. In addition, the models that use only the cortical thickness of the region of interest extracted from 2D MRI (learning models 10-12), the models that use only the volume of the region of interest (learning models 7-9), and the models that use the cortical thickness and volume of the region of interest (learning models 1-6) all showed a reasonable level of performance. The other models showed slightly better performance than the model that uses only the cortical thickness (learning models 10-12).

[0099] [Table 9]

[0100] The researchers also constructed a model to calculate indices related to dementia in previous research. The researchers used data from 924 people (80%) of the population as learning data, and data from 196 people (20%) as validation data. The researchers constructed a model that calculates a certain index by inputting the thickness and / or volume of the cerebral cortex for the region of interest calculated using the learning model in Figure 3 described above. In Figure 4, the learning model corresponds to the model that calculates a certain index (score).

[0101] In previous studies, the W-score is an index for discriminating information on brain atrophy (Renaud La Joie et al., Region-specific hierarchy between atrophy, hypometabolism, and β-amyloid (Aβ) load in Alzheimer's disease dementia, J Neurosci. 2012 Nov 14; 32(46): 16265-73. Reference). Subjects with a W-score of 1.65 or higher can be judged to have brain atrophy. The researchers constructed a learning model to calculate the W-score. Table 10 below shows the correlation between the W-score of the learning model constructed by the researchers and the W-score calculated using the 3D MRI of the same subject. Table 10 compares the results of calculating scores by region of interest. In Table 10, Model 1 uses age and sex as score adjustment variables, and Model 2 uses age, sex, and education level as score adjustment variables.

[0102] [Table 10]

[0103] In previous studies, the AD score has been used as an indicator to determine whether or not dementia is present (Jin San Lee et al., Machine Learning-based Individual Assessment of Cortical Atrophy Pattern in Alzheimer's Disease Spectrum: Development of the Classifier and Longitudinal Evaluation, J Neurosci. SCIENTIFIC REPORTS, Vol.8(1):4161, 2018. Reference).

[0104] The researchers constructed a learning model to calculate AD-scores using the data from the aforementioned population. In other words, this corresponds to the case in Figure 4 where the learning model calculates AD-scores. The researchers verified the constructed model using test data of 196 faces. As a result of the verification, the learning model to calculate AD-scores showed high performance with AUC=0.92 and Accuracy=0.854.

[0105] Furthermore, the researchers calculated the AD-scores using the learning model that calculates the W-scores mentioned above. The researchers used the learning model to derive AD-scores by inputting the W-scores for the seven areas of interest mentioned above. The researchers analyzed the correlation between the AD-scores calculated using the learning model and the AD-scores obtained using previous research methods. Table 11 below shows the performance of the learning model that calculates the W-scores. Model 1 uses age and gender as score adjustment variables, while Model 2 uses age, gender and education level as score adjustment variables. A closer look at Table 11 reveals that there is a considerable correlation with the correct value even when using the model that calculates the W-scores.

[0106] [Table 11]

[0107] 6 shows an example of an analysis device 600 that calculates dementia information of a subject using a brain image. The analysis device 600 may be physically embodied in various forms. For example, the analysis device 600 may have the form of a computer device such as a PC, a smart device, a network server, a chip set dedicated to data processing, etc. Meanwhile, the analysis device 600 may be connected to a brain image device or may be an integrated device.

[0108] The analysis device 600 may include a storage device 610 , a memory 620 , a processing device 630 , an interface device 640 , a communication device 650 , and an output device 660 .

[0109] The storage device 610 can store 2D MRIs of a subject generated by medical imaging equipment.

[0110] The storage device 610 can store clinical information of the subject. The clinical information can include at least one of age, sex, educational level, and APOE4 genotype.

[0111] The storage unit 610 can store a segmentation model for extracting ROIs in brain images (2D MRI slices). The segmentation model is a pre-trained model.

[0112] The storage device 610 may store a learning model for predicting the thickness of the cerebral cortex and / or the volume of a specific brain region based on the information of the region of interest and clinical information (age, sex). The learning model for predicting the thickness of the cerebral cortex and / or the volume of a specific brain region is named a first learning model. As described above, the first learning model uses the total number of pixels of the region of interest (at least one of frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricle, hippocampal formation, and extracerebral cerebrospinal fluid region) and clinical information as input data. Furthermore, the first learning model may further use brain size as input data. Therefore, the first learning model may be divided into a model that does not use brain size and a model that uses brain size and may be prepared in advance. Also, the first learning model may be prepared in advance for each region of interest.

[0113] The storage device 610 may store a learning model that calculates dementia information based on the thickness of the cerebral cortex and / or the volume of a specific brain region. The learning model that finally predicts dementia is named the second learning model. As described above, the second learning model may be embodied in various models according to the type of input data. The second learning model may be any one of (i) a model that uses only the thickness of the cerebral cortex and the volume of a specific brain region as input data, (ii) a model that uses the thickness of the cerebral cortex, the volume of a specific brain region, and clinical information (age, sex, and APEO4) as input data, (iii) a model that uses only the thickness of the cerebral cortex of a specific region of interest as input data, (iv) a model that uses the thickness of the cerebral cortex of a specific region of interest and clinical information (age, sex, and APEO4) as input data, (v) a model that uses only the volume of the specific region of interest as input data, and (vi) a model that uses the volume of the specific region of interest and clinical information (age, sex, and APEO4) as input data. The second learning model may be any one of the models described in Table 7.

[0114] The dementia information may be any one of information such as a prediction result of whether or not a subject has dementia, a prediction result of whether or not a subject is at high risk of dementia, or a dementia-related index (brain atrophy information, W-score, AD-score, etc.).

[0115] The memory 620 can store data and information generated in the process in which the analysis device 600 calculates dementia information from brain images.

[0116] The interface device 640 is a device that receives certain commands and data input from the outside. The interface device 640 can receive the subject's 2D MRI and clinical information from a physically connected input device or an external storage device. The number of input 2D MRI slices is 20 or less. The interface device 640 can transmit dementia information predicted based on the 2D MRI to an external object.

[0117] The communication device 650 refers to a configuration that receives and transmits certain information through a wired or wireless network. The communication device 650 can receive 2D MRI and clinical information of a subject from an external object. The number of 2D MRI slices is 20 or less. The communication device 650 can transmit dementia information predicted based on the 2D MRI to an external object such as a user terminal.

[0118] The interface device 640 may be a device that internally transmits data received by the communication device 650 .

[0119] The output device 660 is a device that outputs certain information. The output device 660 can output an interface required for the data processing process, a brain image, a region of interest divided in the brain image, dementia-related indices calculated based on the region of interest, dementia information, etc.

[0120] The computing device 630 may pre-process the subject's brain image (2D MRI) in a certain way. The computing device 630 may generate a mask for dividing the entire brain region through the data pre-processing process. The computing device 630 may divide the entire brain region in the brain image using the mask for the entire brain region.

[0121] The computing device 630 may normalize the size and / or resolution of the subject's 2D MRI to a uniformity.

[0122] The computing device 630 can input 2D MRI slices of the subject into the trained segmentation model to extract regions of interest, including at least one of frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricles, hippocampal formation, and extracerebral cerebrospinal fluid regions.

[0123] The calculation device 630 may extract size information for the region of interest. The calculation device 630 may calculate a total number of pixels for each region of interest. The calculation device 630 may calculate a total number of pixels by summing up the number of pixels identified in the entire 2D slice for each region of interest (e.g., the forehead).

[0124] The computing device 630 may calculate the total brain size based on the size of the region of interest and other regions in the brain MRI. The brain size may be determined by adding up the size of the seven regions of interest and the white matter in the regions of interest.

[0125] The computing device 630 can input the pixel sum and clinical information (at least one of age, sex, and APEO4 genotype) of the region of interest into the first learning model to predict the thickness and volume of the cerebral cortex of the region of interest. The computing device 630 can input the pixel sum and clinical information (at least one of age, sex, and APEO4 genotype) for each region of interest into an individual learning model to predict the thickness and volume of the cerebral cortex of the region of interest. In this case, the computing device 630 can input the thickness and volume of the cerebral cortex of the region of interest predicted in the first learning model into the second learning model to calculate the dementia information of the subject. For example, the computing device 630 can input the thickness of the cerebral cortex of the frontal region, the thickness of the cerebral cortex of the temporal region, the thickness of the cerebral cortex of the parietal region, the thickness of the cerebral cortex of the occipital region, the volume of the lateral ventricle, the volume of the hippocampus, and the volume of the extracerebral cerebrospinal fluid region into the second learning model to calculate the dementia information of the subject.

[0126] In addition, the computing device 630 may input the total number of pixels of the region of interest, the brain size, and clinical information (at least one of age, sex, and APEO4 genotype) into the first learning model to predict the volume of the region of interest. The computing device 630 may input the total number of pixels, the brain size, and clinical information (at least one of age, sex, and APEO4 genotype) for each region of interest into an individual learning model to predict the volume of the region of interest. In this case, the computing device 630 may input the volume of the region of interest predicted in the first learning model into the second learning model to calculate the dementia information of the subject. For example, the computing device 630 may input the volume of the lateral ventricle, the volume of the hippocampus, and the volume of the extracerebral cerebrospinal fluid region into the second learning model to calculate the dementia information of the subject.

[0127] The computing device 630 can further input clinical information (at least one of age, sex, and APOE4 genotype) in addition to the thickness or volume of the cerebral cortex into the second learning model to calculate the dementia information of the subject.

[0128] The computing device 630 may be a device such as a processor, AP, or a chip with an embedded program that processes data and performs certain operations.

[0129] In addition, the above-mentioned 2D MRI-based dementia information calculation method and 2D MRI-based dementia prediction method may be implemented as a program (or application) including an executable algorithm that can be executed by a computer. The program may be provided by being stored in a transitory or non-transitory computer readable medium.

[0130] A non-transitory readable medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short moment, such as a register, cache, memory, etc. Specifically, the various applications or programs described above can be stored and provided in a non-transitory readable medium, such as a CD, DVD, hard disk, Blu-ray disk, USB, memory card, ROM (read-only memory), PROM (programmable read only memory), EPROM (Erasable PROM, EPROM), EEPROM (Electrically EPROM), flash memory, etc.

[0131] Transient readable media refers to various types of RAM such as 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 (DRRAM).

[0132] The present embodiment and the drawings attached to this specification merely clearly show a portion of the technical ideas contained in the above-mentioned technology, and it is self-evident that all modified examples and specific embodiments that can be easily inferred by a person skilled in the art within the scope of the technical ideas contained in the specification and drawings of the above-mentioned technology are included in the scope of the rights of the above-mentioned technology.

Claims

1. receiving, by the analysis device, two-dimensional (2D) magnetic resonance imaging (MRI) slices of the subject; the analyzer inputting the 2D MRI slices into a segmentation model to extract a region of interest; The analysis device inputs pixel information of the regions of interest into a first pre-trained learning model to predict a cortical thickness of at least one of the regions of interest and a volume of at least one of the regions of interest; and The analysis device includes a step of inputting the thickness of the cerebral cortex of the at least one region and the volume of the at least one region into a pre-trained second learning model to calculate dementia information of the subject, A method for calculating dementia information using two-dimensional MRI, wherein the region of interest includes at least one of frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricles, hippocampal formation, and extracerebral cerebrospinal fluid region.

2. The first learning model predicts the thickness and volume of the cerebral cortex based on the total number of pixels in the region of interest and clinical information of the subject, The dementia information calculation method using two-dimensional MRI according to claim 1 , wherein the clinical information includes at least one of the subject's age, sex, and APOE4 genotype.

3. The method of claim 1 , wherein the first learning model predicts the thickness and volume of the cerebral cortex by further receiving input of the brain size calculated from the 2D MRI slices.

4. The dementia information calculation method using two-dimensional MRI according to claim 1 , wherein the first learning model includes a plurality of learning models prepared in advance for each of the regions of interest.

5. The second learning model is 2. The method for calculating dementia information using two-dimensional MRI according to claim 1, further comprising receiving at least one input of a thickness of the frontal cerebral cortex, a thickness of the temporal cerebral cortex, a thickness of the parietal cerebral cortex, and a thickness of the occipital cerebral cortex, and receiving at least one input of a volume of the lateral ventricle, a volume of the hippocampal formation, and a volume of an extracerebral cerebrospinal fluid region to calculate the dementia information of the subject.

6. The method for calculating dementia information using two-dimensional MRI according to claim 5 , wherein the second learning model further receives input of at least one of the subject's age, the subject's gender, and the subject's APOE4 genotype.

7. receiving, by the analysis device, two-dimensional (2D) magnetic resonance imaging (MRI) slices of the subject; the analyzer inputting the 2D MRI slices into a segmentation model to extract a region of interest; The analysis device inputs pixel information of the regions of interest into a first pre-trained learning model to predict a volume of at least one of the regions of interest; and The analysis device includes a step of inputting the volume of the at least one region into a second learning model that has been trained in advance to calculate dementia information of the subject, A method for calculating dementia information using two-dimensional MRI, wherein the region of interest includes at least one of frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricles, hippocampal formation, and extracerebral cerebrospinal fluid region.

8. The first learning model predicts the volume based on the total number of pixels in the region of interest and clinical information of the subject, The dementia information calculation method using two-dimensional MRI according to claim 7 , wherein the clinical information includes at least one of the subject's age, sex, and APOE4 genotype.

9. The method of claim 7, wherein the first learning model predicts the volume by further receiving a brain size calculated from the 2D MRI slice.

10. The dementia information calculation method using two-dimensional MRI according to claim 7 , wherein the first learning model includes a plurality of learning models prepared in advance for each of the regions of interest.

11. The second learning model is 8. The method for calculating dementia information using two-dimensional MRI according to claim 7, further comprising receiving at least one input of a thickness of the frontal cerebral cortex, a thickness of the temporal cerebral cortex, a thickness of the parietal cerebral cortex, and a thickness of the occipital cerebral cortex, and receiving at least one input of a volume of the lateral ventricle, a volume of the hippocampal formation, and a volume of an extracerebral cerebrospinal fluid region to calculate dementia information of the subject.

12. The method for calculating dementia information using two-dimensional MRI according to claim 11, wherein the second learning model further receives input of at least one of the subject's age, the subject's gender, and the subject's APOE4 genotype.

13. an interface device that receives input of 2D (dimensional) Magnetic Resonance Imaging (MRI) slices of the subject; A storage device for storing a segmentation model for extracting a region of interest in a 2D MRI slice, a first learning model for predicting a volume in response to input of region of interest information, and a second learning model for calculating dementia information; and The present invention includes a computing device that inputs the received 2D MRI slices into the segmentation model to extract a region of interest, inputs pixel information of at least one first region among the extracted regions of interest into the first learning model to predict a volume of the at least one first region, and inputs the volume of the at least one first region into the second learning model to calculate dementia information of the subject, The region of interest includes at least one of frontal gray matter, temporal gray matter, parietal gray matter, occipital gray matter, lateral ventricles, hippocampal formation, and extracerebral cerebrospinal fluid region, and the analysis device calculates dementia information using two-dimensional MRI.

14. The analysis device for calculating dementia information using two-dimensional MRI according to claim 13 , wherein the first learning model includes a plurality of learning models prepared in advance for each of the regions of interest.

15. The analysis device for calculating dementia information using two-dimensional MRI according to claim 13, wherein the at least one first region includes a lateral ventricle, a hippocampal formation, and an extracerebral cerebrospinal fluid region.

16. The storage device further stores a third learning model for predicting a thickness of the cerebral cortex in response to input of region of interest information; The analysis device for calculating dementia information using two-dimensional MRI described in claim 13, wherein the calculation device inputs pixel information of at least one second region among the extracted regions of interest into the third learning model to predict the thickness of the cerebral cortex of the at least one second region, and further inputs the thickness of the cerebral cortex of the at least one second region into the second learning model to calculate the dementia information of the subject.

17. The analysis device for calculating dementia information using two-dimensional MRI according to claim 16, wherein the at least one second region includes at least one of a frontal region, a temporal region, a parietal region, and an occipital region.

18. The analysis device for calculating dementia information using two-dimensional MRI described in claim 13, wherein the calculation device further inputs at least one of the subject's age, the subject's gender, and the subject's APOE4 genotype into the second learning model to calculate the dementia information of the subject.

Images