Explainable AI(XAI)-based System for Early Diagnosis and Predicting Prognosis of Alzheimer's Disease Using VR Biomarkers, and Method thereof
An AI-based system integrates VR biomarkers with brain MRI to enhance understanding and management of Alzheimer's dementia by providing structural explanations, addressing limitations of existing methods.
Patent Information
- Application Number
- KR1020230105821
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2023-08-11
- Publication Date
- 2026-11-11
- Estimated Expiration
- 2043-08-11
AI Technical Summary
Current methods for diagnosing and predicting Alzheimer's dementia, such as neuropsychological assessments and brain MRI, face limitations in reproducibility, accuracy, and cost, while VR biomarkers lack sufficient explanatory power to interpret structural and neurological brain associations.
An explainable artificial intelligence-based system that integrates VR biomarkers with brain MRI, using VR data collection, feature extraction, and brain MRI data analysis to provide a brain structural explanation through explainable AI, enhancing subject understanding and facilitating lifestyle changes.
Enhances subject understanding of dementia diagnosis and prognosis by linking VR biomarkers with existing biomarkers, improving diagnostic accuracy and enabling management through lifestyle adjustments.
Smart Images

Figure 112023088966433-PAT00002_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to an early diagnosis and prognosis prediction system for Alzheimer's dementia, and more specifically, to an explainable artificial intelligence-based system and method for early diagnosis and prognosis prediction of Alzheimer's dementia utilizing VR biomarkers, which can enhance the subject's understanding by linking the results derived from VR biomarkers with existing biomarkers through explainable artificial intelligence, and can also assist in the management of dementia by inducing changes in habits or lifestyle. Background Technology
[0003] Dementia is an irreversible disease, and to date, there is no treatment that can prevent its onset or completely halt its progression. In particular, mild cognitive impairment (MCI) serves as an intermediate stage between aging and dementia, making it highly likely to transition into dementia. According to the 2021 Dementia Status Report by the National Dementia Center, approximately 10.33% of the elderly aged 65 and older currently suffer from dementia, and results indicate that about 10–20% of MCI patients progress to dementia annually. Furthermore, the national cost required for dementia treatment and management is estimated to increase to approximately 56.9 trillion won by 2040. For these reasons, early detection at the MCI stage, along with consistent follow-up and prognosis prediction, is essential. Since MCI is a disease that requires patient effort and lifestyle changes for prevention and management, it is becoming increasingly important to present test results in a manner that is easy for the subject to understand and persuasive.
[0004] Currently, biomarkers such as neuropsychological assessments (e.g., SNSB) or brain magnetic resonance imaging (brain MRI) are primarily used for the early diagnosis of dementia. While neuropsychological assessments are mainly used in the early stages when mild cognitive impairment is suspected, there are concerns regarding the limited reproducibility and accuracy of cognitive ability evaluations because they are influenced by subject characteristics (e.g., age, gender, years of education) or subjective interventions by the evaluator. Furthermore, in the early stages of mild cognitive impairment, the degree of brain atrophy is minimal, making it difficult to accurately identify patients based solely on brain MRI. Additionally, since the results are influenced by characteristics such as head circumference, there is a risk of misdiagnosing normal individuals as having mild cognitive impairment if brain atrophy is present. Recently, methods utilizing Virtual Reality (VR) technology to simulate virtual daily life and diagnose mild cognitive impairment through the performance of daily activities have been gaining attention. While early diagnosis using VR technology demonstrates excellent performance and has the advantage of being more intuitively understandable to subjects than existing biomarkers, there are limitations to utilizing VR as a standalone biomarker due to a lack of connection to the structural or neurological changes in the brain that explain dementia.
[0005] Recently, research utilizing Explainable AI (XAI) for existing biomarkers has been actively conducted, particularly those using brain MRI as input. For instance, IBM's "Watson OpenScale" is a technology that analyzes a subject's brain MRI images by region to determine the presence of mild cognitive impairment and provides supplementary explanations to help the subject understand their results. Additionally, Class Activation Maps (CAM) are a method that visualizes areas suspected of dementia in brain MRIs as heat maps, presenting visual evidence to the subject. However, while these methodologies present the rationale derived from the analysis and judgment of test results by XAI, they lack sufficient consideration regarding whether such explanations are helpful for the subject to understand their own condition. Furthermore, existing biomarkers, such as neuropsychological tests and brain MRI, suffer from the drawbacks of being time-consuming and costly. For these reasons, there is a demand for testing methods that can provide explanations to the subject while resolving the spatiotemporal and economic issues associated with conventional biomarkers.
[0006] Recently, VR technology has garnered attention as a method for evaluating daily living activities, a key indicator distinguishing between mild cognitive impairment and normal aging. Various daily living tasks (e.g., using an ATM, taking a bus, shopping) are implemented in a VR environment, and behavioral data (e.g., hand movements, eye movements) and performance metrics (e.g., number of errors, execution time) collected while subjects perform the VR experience are recorded. VR biomarkers extracted based on this data not only resolve the spatiotemporal and economic limitations of conventional biomarkers but also demonstrate superior performance in the early diagnosis of mild cognitive impairment. Furthermore, they offer the significant advantage of allowing subjects to intuitively understand their own condition. For these reasons, VR biomarkers are expected to play a pivotal role in future biomarker research and clinical applications related to dementia diagnosis. However, current VR biomarker methodologies have limitations in that they lack sufficient explanatory power to adequately interpret the structural and neurological brain associations between the collected data and mild cognitive impairment.
[0007] Meanwhile, Korean Published Patent Application No. 10-2022-0104706 (Patent Document 1) discloses "a biomarker for diagnosing the silent stage of Alzheimer's disease and its use," wherein a method for predicting the progression of the silent stage of Alzheimer's disease in a subject according to the above is characterized by comprising the steps of determining a molecular signature by measuring the level, amount, or concentration of at least five biomarkers selected from a group of biomarkers in a sample obtained from the subject, comparing the obtained molecular signature with a reference signature, and predicting the progression of Alzheimer's disease based on the correlation between the molecular signature and the reference signature.
[0008] Although Patent Document 1 described above has the advantage of being able to predict AD in the silent phase and further stratify AD into different grades, the group of biomarkers used here consists mainly of various protein-based substances or cell regulatory substances for the generation and maintenance of cells in vivo, and thus has the disadvantage of lacking a connection to structural or neurological changes in the brain that can explain dementia. Prior art literature
[0009] Korean Published Patent Application No. 10-2022-0104706 (Published July 26, 2022) The problem to be solved
[0010] The present invention was created by comprehensively considering the above-mentioned matters, and aims to provide an explainable artificial intelligence-based system and method for the early diagnosis and prognosis prediction of Alzheimer's dementia utilizing VR biomarkers, which evaluates the ability to perform daily activities in a virtual environment through VR, provides evidence regarding dementia to the subject by linking it with existing biomarkers such as brain MRI, and enhances the subject's understanding by explaining the results derived from VR biomarkers in conjunction with existing biomarkers through explainable artificial intelligence, thereby inducing actual changes in habits or lifestyle to improve the condition and assisting in the management of dementia. means of solving the problem
[0012] To achieve the above objective, the explainable artificial intelligence-based early diagnosis and prognosis prediction system for Alzheimer's dementia utilizing VR biomarkers according to the present invention is,
[0013] VR data collection unit for collecting VR biomarkers from subjects;
[0014] A feature extraction unit that extracts features from VR biomarkers collected by the VR data collection unit through an artificial intelligence model;
[0015] A brain MRI data collection unit that receives a subject's brain MRI image, extracts features through an artificial intelligence model, analyzes each brain MRI region, and generates a single MRI biomarker; and
[0016] The invention is characterized by including a control unit that controls the status check and operation of the VR data collection unit, the feature extraction unit, and the brain MRI data collection unit, receives VR biomarkers from the VR data collection unit and MRI biomarkers from the brain MRI data collection unit, respectively, analyzes the correlation between the VR biomarkers and MRI biomarkers through explainable artificial intelligence, and provides a brain structural explanation regarding the VR biomarkers.
[0017] Here, the VR biomarkers collected by the VR data collection unit from the subject may include the subject's eye movement data, hand movement data, and performance data.
[0018] In addition, the MRI biomarker generated by the brain MRI data acquisition unit may include frontal lobe MRI data and temporal lobe MRI data of the brain.
[0019] In addition, to achieve the above objective, the explainable artificial intelligence-based method for early diagnosis and prognosis prediction of Alzheimer's dementia using VR biomarkers according to the present invention is,
[0020] a) A step in which a VR data collection unit collects VR biomarkers from a subject;
[0021] b) A step in which a feature extraction unit extracts features from VR biomarkers collected by the VR data collection unit through an artificial intelligence model;
[0022] c) A step in which a brain MRI data collection unit receives a subject's brain MRI image, extracts features through an artificial intelligence model, analyzes them by brain MRI region, and generates MRI biomarkers;
[0023] d) a control unit receiving VR biomarkers from the VR data acquisition unit and MRI biomarkers from the brain MRI data acquisition unit, respectively; and
[0024] e) The method is characterized by including a step in which the control unit analyzes the correlation between the VR biomarker and the MRI biomarker through explainable artificial intelligence and provides a brain structural explanation for the VR biomarker.
[0025] Here, the VR biomarker collected by the VR data collection unit from the subject in step a) above may include the subject's eye movement data, hand movement data, and performance data.
[0026] In addition, the MRI biomarker generated by the brain MRI data acquisition unit in step c) above may include frontal lobe MRI data and temporal lobe MRI data of the brain. Effects of the invention
[0028] According to the present invention, the ability to perform daily activities is evaluated in a virtual environment through VR, and evidence regarding dementia is provided to the subject in conjunction with existing biomarkers such as brain MRI. Furthermore, by explaining the results derived from VR biomarkers in conjunction with existing biomarkers through explainable artificial intelligence, the subject's understanding can be enhanced, and consequently, changes in habits or lifestyle can be induced to improve the condition, thereby providing assistance in the management of dementia. Brief explanation of the drawing
[0030] FIG. 1 is a schematic diagram showing the configuration of an explainable artificial intelligence-based early diagnosis and prognosis prediction system for Alzheimer's dementia using VR biomarkers according to the present invention. FIG. 2 is a flowchart illustrating the execution process of an explainable artificial intelligence-based method for early diagnosis and prognosis prediction of Alzheimer's dementia using VR biomarkers according to the present invention. FIG. 3 is a diagram showing an overall overview of an explainable artificial intelligence-based method for early diagnosis and prognosis prediction of Alzheimer's dementia using VR biomarkers according to the present invention. FIG. 4 is a diagram showing an experimental set for a virtual kiosk test introduced to implement the method according to the present invention. FIG. 5 is a diagram showing an overview of extracting four VR biomarkers from behavioral data in a virtual kiosk test employed for the implementation of the method according to the present invention. Figure 6 is a diagram showing an overview of extracting eight MRI biomarkers from each hemisphere in T1-weighted MRI using a split-attention U-net architecture. Specific details for implementing the invention
[0031] Terms and words used in this specification and claims should not be interpreted as being limited to their ordinary or dictionary meanings, and should be interpreted in a meaning and concept consistent with the technical spirit of the invention, based on the principle that the inventor can appropriately define the concept of the terms to best describe his invention.
[0032] Throughout the specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "…part," "…unit," "module," and "device" as used in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.
[0033] Embodiments of the present invention will be described in detail below with reference to the attached drawings.
[0034] FIG. 1 is a schematic diagram showing the configuration of an explainable artificial intelligence-based early diagnosis and prognosis prediction system for Alzheimer's dementia using VR biomarkers according to an embodiment of the present invention.
[0035] Referring to FIG. 1, the explainable artificial intelligence-based early diagnosis and prognosis prediction system for Alzheimer's dementia using VR biomarkers according to the present invention (100) comprises a VR data collection unit (110), a feature extraction unit (120), a brain MRI data collection unit (130), and a control unit (140).
[0036] The VR data collection unit (110) collects VR biomarkers from a subject. Here, the VR biomarkers collected by the VR data collection unit (110) from the subject may include the subject's eye movement data, hand movement data, and performance data. That is, as can be seen in FIG. 4, the VR data collection unit (110) can collect the subject's eye movement data and hand movement data using a hand controller from a subject wearing a head-mounted display (HMD). At this time, the eye and hand movement data may be collected at intervals of 100ms. In addition, the VR data collection unit (110) may also collect performance data including the number of errors and execution time while the subject performs a given task. In FIG. 4, the base station is a device for measuring the patient's hand movements, and the laptop computer serves to display a VR environment on the screen that allows the subject to perform the task, and also visualizes and displays the collected VR biomarkers on the screen. At this time, the laptop computer may be equipped with a specific software program for VR testing of the subject (which is a type of VR testing application pre-written by the inventor).
[0037] The feature extraction unit (120) extracts features from VR biomarkers collected by the VR data collection unit (110) through an artificial intelligence model (120a) (a type of software program for feature extraction that is pre-trained to extract features from given data). This feature extraction unit (120) extracts features by analyzing the time-series data of collected hand movements and eye movements through the artificial intelligence model (120a) based on features that were identified as significant for cognitive function evaluation in previous research (previous artificial intelligence model learning process).
[0038] The brain MRI data collection unit (130) receives a brain MRI image of a subject, extracts features through an artificial intelligence model (130a), and analyzes the brain MRI region to generate a single MRI biomarker. Here, the MRI biomarker generated by the brain MRI data collection unit (130) may include frontal lobe MRI data and temporal lobe MRI data of the brain, as shown in FIG. 3.
[0039] The control unit (140) controls the status check and operation of the VR data collection unit (110), the feature extraction unit (120), and the brain MRI data collection unit (130), and receives VR biomarkers from the VR data collection unit (110) and MRI biomarkers from the brain MRI data collection unit (130), respectively, and analyzes the correlation between the VR biomarkers and MRI biomarkers through explainable artificial intelligence (140a) to provide a brain structural explanation regarding the VR biomarkers. Such a control unit (140) may be equipped with a database (DB) (140d). This database (DB) (140d) may store the subject's eye movement data, hand movement data, performance data, and frontal lobe MRI data and temporal lobe MRI data related to the subject's brain MRI images.
[0040] The explainable artificial intelligence-based Alzheimer's dementia early diagnosis and prognosis prediction system (100) utilizing VR biomarkers according to the present invention having the above configuration may preferably further include a communication unit (150) for transmitting / receiving data (information) with an external device (e.g., a medical staff terminal or a service provider server).
[0041] In addition, depending on the case, the explainable artificial intelligence-based early diagnosis and prognosis prediction system (100) for Alzheimer's dementia utilizing the VR biomarker of the present invention as described above may be composed of a single computer system in which each component is integrated as a whole.
[0042] Then, below, we will describe a method for early diagnosis and prognosis prediction of Alzheimer's dementia based on an explainable artificial intelligence-based system for early diagnosis and prognosis prediction of Alzheimer's dementia utilizing a VR biomarker according to the present invention having the configuration as described above.
[0043] FIG. 2 is a flowchart showing the execution process of an explainable artificial intelligence-based method for early diagnosis and prognosis prediction of Alzheimer's dementia using VR biomarkers according to an embodiment of the present invention, and FIG. 3 is a diagram showing an overall overview of an explainable artificial intelligence-based method for early diagnosis and prognosis prediction of Alzheimer's dementia using VR biomarkers according to the present invention.
[0044] Referring to FIGS. 2 and 3, the method for early diagnosis and prognosis prediction of Alzheimer's dementia based on explainable artificial intelligence using VR biomarkers according to the present invention first involves a VR data collection unit (110) collecting VR biomarkers from a subject (step S201). Here, the VR biomarkers collected by the VR data collection unit (110) from the subject may include the subject's eye movement data, hand movement data, and performance data. That is, as previously explained with reference to FIG. 4, the VR data collection unit (110) can collect the subject's eye movement data and hand movement data using a hand controller from a subject wearing an HMD (head mounted display). At this time, the eye and hand movement data may be collected at intervals of 100ms. In addition, the VR data collection unit (110) may also collect performance data including the number of errors, execution time, etc., while the subject performs a given task (work).
[0045] When VR biomarkers are collected from the subject in this way, the feature extraction unit (120) extracts features from the VR biomarkers collected by the VR data collection unit (110) through an artificial intelligence model (120a) (step S202). That is, the feature extraction unit (120) extracts features by analyzing the time-series data of the collected hand movements and eye movements through the artificial intelligence model (120a) based on features that were identified as significant for cognitive function evaluation in previous research (previous artificial intelligence model learning process).
[0046] Meanwhile, the brain MRI data collection unit (130) receives the subject's brain MRI image, extracts features through an artificial intelligence model (130a), and analyzes them by brain MRI region to generate MRI biomarkers (step S203). The MRI biomarkers generated by the brain MRI data collection unit (130) may include frontal lobe MRI data and temporal lobe MRI data of the brain. Here, we will add a little more explanation regarding these MRI biomarkers.
[0047] In the present invention, "T1-weighted MRI" was used to generate MRI biomarkers. "T1-weighted MRI" is an image specialized for detailed representation of brain structures, characterized by its ability to clearly show anatomical details, and is useful for studying brain volume, morphological changes, and structural changes. MRI scan data is reconstructed in 3D through a deep learning model (artificial intelligence model), and the volume of each region is used as an MRI biomarker.
[0048] Eight regions in the temporal lobe related to memory and learning were selected as MRI biomarkers, and a total of 16 regions—eight from each hemisphere—were used for the analysis. Among these, six regions satisfying p<0.006 (i.e., 0.05 / 8=0.00625) and p<0.001 (i.e., 0.01 / 8=0.00125) after applying Bonferroni correction were selected as the final MRI biomarkers. The core regions among the temporal lobe regions are shown in Table 1 below.
[0049] Biomarkers Description Amygdala It is involved in regulating emotional processing and memory connections. Damage to the amygdala makes it difficult to process situations involving high stress or emotional instability, which can be confirmed by symptoms such as reduced performance of VR biomarkers. Hippocampus It is associated with memory formation, spatial memory, etc. Damage to the hippocampus may result in difficulty remembering the target itself, and even if the target is remembered, difficulties may be experienced, such as exhibiting complex patterns in hand or eye movements to reach the target. Entorhinal cortex It is an area that performs a role similar to the hippocampus and is associated with spatial memory and perception. Acting as an intermediate pathway before memories enter the hippocampus, this can be confirmed by the similarly observed low performance in hand and eye movements.
[0050] Afterwards, the control unit (140) receives the VR biomarker from the VR data collection unit (110) and the MRI biomarker from the brain MRI data collection unit (130), respectively (step S204).
[0051] Then, the control unit (140) analyzes the correlation between the VR biomarker and the MRI biomarker through explainable artificial intelligence (140a) and provides a brain structural explanation for the VR biomarker (step S205).
[0052] Below, we will provide further explanation regarding the explainable artificial intelligence-based method for early diagnosis and prognosis prediction of Alzheimer's dementia utilizing VR biomarkers according to the present invention as described above.
[0053] FIG. 5 is a diagram showing an overview of extracting four VR biomarkers from behavioral data in a virtual kiosk test employed for the implementation of the method according to the present invention.
[0054] Referring to Figure 5, the behavioral data collected during the test was converted into four VR biomarkers: hand movement speed, scan path length, completion time, and number of errors. The first VR biomarker, hand movement speed, was determined by dividing the total distance of hand movements by the total test time, which correlates with cognitive abilities such as recognition and processing speed. The second VR biomarker, scan path length, represents overall cognitive ability, information processing efficiency, and level of comprehension, reflecting the distance the participant's gaze traveled during the test. The third VR biomarker, completion time, indicates the time required for the participant to complete all six stages of the test. The final VR biomarker, number of errors, records the total number of incorrect actions during the test, such as incorrect choices regarding dining location, burger, drink, side menu, payment method, or password entry. In Figure 5, hand movement speed is calculated using hand movement data collected from the virtual kiosk test. Scan path length is derived from eye movement data. Completion time and the number of errors are calculated based on task performance data.
[0055] Figure 6 is a diagram showing an overview of extracting eight MRI biomarkers from each hemisphere in T1-weighted MRI using a split-attention U-net architecture.
[0056] Referring to Fig. 6, MRI scans were performed at Hanyang University Hospital in Guri and Seoul using a "PHILIPS Ingenia CX 3T" scanner. The scans utilized 3D T1-weighted magnetization-prepared rapid gradient echo (MP-RAGE) technology with specific protocols set for each hospital.
[0057] The inventors of this study used AQUA 3.0 software (Neurophet Inc., South Korea) to process acquired T1-weighted MRI. This method helps to accurately identify and contour regions of interest in the brain, as shown in Fig. 6. AQUA 3.0 software is based on a Split-Attention U-Net deep learning architecture that fuses ResNeSt and U-Net elements. The architecture of Split-Attention U-Net demonstrates robustness against neuroanatomical variability through encoder / decoder structures, skip paths, and split-attention modules. Furthermore, by utilizing EvoNorm-based convolutional layers and 3D ResNeSt blocks, it improves segmentation accuracy, particularly for small subcortical regions. This methodological approach provides valuable information regarding brain volume in MRI that can be utilized as MRI biomarkers for neuroimaging analysis. In addition, this software analyzes the left and right hemispheres separately to explain cerebral hemisphere asymmetry, allowing for a more comprehensive evaluation. As shown in Figure 6, each hemisphere has eight temporal lobe biomarkers, including the amygdala, hippocampus, entorhinal cortex, hippocampal gyrus, fusiform gyrus, superior temporal gyrus, middle temporal gyrus, and inferior temporal gyrus.
[0058] The inventors of this invention selected MRI biomarkers associated with the temporal lobe, a region that exhibits early atrophy in AD and plays a significant role in multiple cognitive functions. These MRI biomarkers include the amygdala, hippocampus, entorhinal cortex, hippocampal gyrus, fusiform gyrus, superior temporal gyrus, middle temporal gyrus, and inferior temporal gyrus. The inventors derived these MRI biomarkers from both hemispheres to obtain a total of 16 MRI biomarkers. Additionally, to account for overall brain size, intracranial volume (ICV), a measure that encapsulates total gray matter, white matter, and cerebrospinal fluid, was incorporated. Then, all MRI biomarkers were normalized to ICV to mitigate differences in brain volume caused by factors such as age, sex, and head circumference.
[0059] Meanwhile, the inventors of the present invention investigated the relationships between statistically significant biomarkers through Pearson correlation analysis. These biomarkers include VR biomarkers of hand movement speed, scan path length, completion time, and number of errors, in addition to MRI biomarkers of the amygdala, hippocampus, and entorhinal cortex. The goal was to identify significant associations between these biomarkers and to reveal their potential significance within the context of the present invention. The analysis results revealed a significant correlation between hand movement speed and the left hippocampus (r=0.425, p=0.004). Additionally, the number of errors showed a significant correlation with the right amygdala (r=-0.434, p=0.003). Bonferroni correction was performed for multiple comparisons, and significance thresholds were set at p<0.005 (i.e., 0.05 / 10=0.005) and p<0.001 (i.e., 0.01 / 10=0.001).
[0061] As described above, the system and method for early diagnosis and prognosis prediction of Alzheimer's dementia based on explainable artificial intelligence utilizing VR biomarkers according to the present invention evaluates the ability to perform daily activities in a virtual environment through VR, provides evidence regarding dementia to the subject by linking it with existing biomarkers such as brain MRI, and enhances the subject's understanding by explaining the results derived from VR biomarkers in conjunction with existing biomarkers through explainable artificial intelligence. Consequently, it has the advantage of assisting in the management of dementia by inducing actual changes in habits or lifestyle to improve the condition.
[0062] Although the present invention has been described in detail through preferred embodiments, the invention is not limited thereto, and it is obvious to those skilled in the art that various modifications and applications can be made within the scope of the technical concept of the invention. Accordingly, the true scope of protection of the present invention should be interpreted by the following claims, and all technical concepts within an equivalent scope should be interpreted as being included within the scope of rights of the present invention. Explanation of the symbols
[0063] 100: (The present invention) Explainable AI-based system for early diagnosis and prognosis prediction of Alzheimer's dementia using VR biomarkers 110: VR Data Collection Unit 120: Feature Extraction Unit 130: Brain MRI data acquisition unit 140: Control unit 150: Communications Department 120a, 130a: Artificial Intelligence Model 140a: Artificial Intelligence 140d: Database (DB)
Claims
Claim 1 A VR data collection unit for collecting VR biomarkers from a subject; a feature extraction unit for extracting features from the VR biomarkers collected by the VR data collection unit through an artificial intelligence model; a brain MRI data collection unit for receiving a brain MRI image of a subject, extracting features through an artificial intelligence model, analyzing them by brain MRI region, and generating a single MRI biomarker; and a control unit for controlling the status check and operation of the VR data collection unit, the feature extraction unit, and the brain MRI data collection unit, receiving the VR biomarker from the VR data collection unit and the MRI biomarker from the brain MRI data collection unit, respectively, and providing a brain structural explanation for the VR biomarker by analyzing the correlation between the VR biomarker and the MRI biomarker through explainable artificial intelligence, wherein the MRI biomarker generated by the brain MRI data collection unit includes frontal lobe MRI data and temporal lobe MRI data of the brain. Claim 2 An explainable artificial intelligence-based system for early diagnosis and prognosis prediction of Alzheimer's dementia utilizing VR biomarkers, characterized in that, in claim 1, the VR biomarkers collected by the VR data collection unit from the subject include the subject's eye movement data, hand movement data, and performance data. Claim 3 a) a step in which a VR data collection unit collects VR biomarkers from a subject; b) a step in which a feature extraction unit extracts features from the VR biomarkers collected by the VR data collection unit through an artificial intelligence model; c) a step in which a brain MRI data collection unit receives a brain MRI image of a subject, extracts features through an artificial intelligence model, analyzes them by brain MRI region, and generates MRI biomarkers; d) a step in which a control unit receives the VR biomarkers from the VR data collection unit and the MRI biomarkers from the brain MRI data collection unit, respectively; and e) a step in which a control unit analyzes the correlation between the VR biomarkers and the MRI biomarkers through explainable artificial intelligence and provides a brain structural explanation for the VR biomarkers, wherein the MRI biomarkers generated by the brain MRI data collection unit in step c) include frontal lobe MRI data and temporal lobe MRI data of the brain. Claim 4 A method for early diagnosis and prognosis prediction of Alzheimer's dementia based on explainable artificial intelligence utilizing VR biomarkers, characterized in that, in step a) above, the VR biomarkers collected by the VR data collection unit from the subject include the subject's eye movement data, hand movement data, and performance data. Claim 5 delete
Citation Information
Patent Citations
Method for diagnosing cognitive impairment and predicting prognosis of cognitive impairment
KR1020230041129A
An explainable ai (XAI) platform for computational pathology
WO2020190851A1
