A digital biomarker for assessing cognitive impairment

Abstract

The present invention provides a digital biomarker for assessing cognitive impairment. The initial data for the digital biomarker is generated by collecting behavioral data from users using digital props, such as mahjong, playing cards, chess, or Go. This collected initial data is sent to an analysis unit and / or server. This data is then analyzed and calculated by an artificial intelligence program on the analysis unit and / or server to produce a series of digital biomarkers that characterize the user's cognitive abilities, such as perceptual cognition, sensory cognition, and thinking cognition. The resulting digital biomarkers can be used to label real-time data collected while the user is using the digital props, thereby assessing the user's cognitive abilities.

Description

Technical Field

[0001] The present invention relates to the field of digital therapy technology, and in particular to a digital biomarker for evaluating cognitive impairment. Background Art

[0002] Traditional biomarkers generally refer to indicators obtained through biochemical testing, used to indicate changes in the structure or function of organs and other tissues. Digital biomarkers are objective characteristic data obtained by collecting physiological indicators, movement status, and other types of data through the Internet of Things and other means when users use various smart devices such as mobile phones, computers, tablets, and smart wearables, and then analyzing and organizing them through artificial intelligence. Characteristic data can be used to predict and evaluate users' real-time physiological state or health status. On the one hand, the current development goal of digital biomarkers is to effectively supplement traditional biomarkers, rather than replace the latter; on the other hand, digital biomarkers can effectively promote the transformation of the health care model from passive response to active prevention. Digital biomarkers are expected to become an effective means of gaining a deeper understanding of human health and disease.

[0003] Prior art, such as embodiments disclosed in Australian Patent Publication No. AU2020310165A1, relates to systems and methods for detecting cognitive decline in a subject using passively acquired data from at least one mobile device. In an exemplary embodiment, a computer-implemented method includes receiving the passively acquired data from at least one mobile device. The method further includes generating digital biomarker data from the passively acquired data. The method further includes analyzing the digital biomarker data to determine whether the subject exhibits signs of cognitive decline.

[0004] Prior art, such as U.S. Patent Publication No. US20190200915A1, discloses a mobile device comprising a processor, at least one sensor, a database, and software tangibly embedded in the device and configured to execute a disclosed method when executed on the device. The invention provides for determining cognitive and / or fine motor activity parameters from an activity measurement dataset obtained from a subject using the mobile device. The determined activity parameters are compared with a reference, and cognitive and motor diseases or disorders are assessed.

[0005] Existing technologies all collect and obtain initial data through mobile devices, and this type of initial data only includes interaction data and / or language data generated when the user uses the mobile device. When judging whether a person has cognitive impairment, combining the facial data of the examinee can effectively improve the accuracy of the judgment. The present invention combines a wide-ranging entertainment activity to collect user interaction data, language data, eye movement data, etc. as initial data, and extracts accurate and effective digital biomarkers through multi-dimensional and comprehensive initial data.

[0006] In addition, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the applicant studied a large number of documents and patents when making the present invention, but due to space limitations, not all details and contents are listed in detail. However, this does not mean that the present invention does not have the characteristics of these prior arts. On the contrary, the present invention already has all the characteristics of the prior art, and the applicant reserves the right to add relevant prior art to the background technology. Summary of the Invention

[0007] In response to the shortcomings of the existing technology, the present invention provides a digital biomarker for assessing cognitive impairment. The initial data of the digital biomarker is formed by collecting at least behavioral data of various operations performed by users using digital props such as mahjong, playing cards, chess, go, etc.; the collected initial data will be sent to an analysis unit and / or a server. The initial data will be analyzed and calculated by an artificial intelligence program on the analysis unit and / or server to obtain a series of digital biomarkers that can characterize the user's cognitive abilities, such as perceptual cognition, sensory cognition, and thinking cognition. The obtained digital biomarkers can be used to mark the real-time data collected when the user uses the digital props, and the marking can be used to assess the user's cognitive abilities. According to the present invention, when using a smart wearable device, multiple users wear their own smart wearable devices.

[0008] When extracting digital biomarkers according to the present invention, the behavioral data of each user wearing a smart wearable device is analyzed. Specifically, the initial data collected and analyzed according to the present invention includes user behavioral data, which is collected by the digital prop during the user's use of the digital prop. Preferably, a set of digital props can correspond to multiple sets of smart wearable devices, wherein at least one of the digital props can have at least one sensor, and each smart wearable device can pre-establish a digital link with the prop with the sensor to collect behavioral data.

[0009] Through this configuration, the method of screening / evaluating the user's health status (such as cognitive ability) based on digital props is more easily accepted by ordinary users (especially middle-aged and elderly people). That is, on the one hand, the long-term use of this digital prop will not affect the user's normal life. On the other hand, because this digital prop can arouse the sustained interest of ordinary users (especially middle-aged and elderly people), it can enable users to actively and long-term use this digital prop. The user's behavioral data collected by this digital prop can be used as one of the initial data and sent to the server, and then processed and analyzed by the server to calculate the digital biomarker that can characterize the user's cognitive ability; in addition, it can also significantly reduce the economic cost and social cost of screening / evaluating the user's health status (such as cognitive ability) and avoid the waste of medical resources.

[0010] When the digital props are used in a community chess and card entertainment venue, there are likely to be multiple users using the digital props in the venue, and the digital props are used at the same time (for example, multiple users of the digital props are arranging the physical chess and card at the same time, turning the physical chess and card in their palms at the same time, etc., or the same digital mahjong can be moved or picked up by multiple different users at the same time). Whenever the digital props of the system are used / moved by a user, the digital props can at least send the motion data of the digital props at that time to the analysis unit and / or the server. Since each digital prop can be used by multiple different users in a time period, The track sensor or server is unable to accurately and quickly assign the motion data of each digital prop in use / movement to the action database of the user using the digital prop. This means that the ownership of the motion trajectory and other data recorded by the digital prop remains unresolved. Therefore, users of this system are equipped with a portable smart device. This wearable device is capable of at least transmitting an identifier that identifies the user's identity information to the track sensor within the digital prop, thereby resolving the ownership issue of the motion trajectory and other data recorded by the digital prop. In other words, a pre-bound data link relationship can exist between the wearable device and at least one prop, which can be achieved through protocols such as Near Field Communication, ZigBee, and Bluetooth.

[0011] Preferably, the track sensor is capable of acquiring an identifier of the portable smart device. Preferably, the track sensor is capable of measuring the distance between itself and the portable smart device. Particularly preferably, the track sensor within each digital prop measures the distance to the portable smart devices surrounding the digital prop at the second moment of being moved by the user, and sends a data attribution request to the portable smart device closest to the digital prop. Preferably, upon receiving the data attribution request, the portable smart device closest to the digital prop sends its identifier to the track sensor to identify the identity information of the user using the portable smart device.

[0012] Preferably, the portable smart device is capable of manually entering the user's identity information. Preferably, the user's identity information may include, but is not limited to, the user's name, age, gender, etc. At the same time, the portable smart device is capable of at least measuring the distance between the moving digital prop and the portable smart device.

[0013] Preferably, the track sensor can automatically time the movement or vibration of the track sensor or the chess and card entity. Preferably, the motion data collected by the track sensor can be accompanied by a timestamp corresponding to the motion data.

[0014] Preferably, the first moment can be determined by a track sensor. Preferably, when the vibration value of the track sensor and / or the chess and card entity measured by the track sensor changes from a value of zero to a non-zero value, the track sensor determines that this time is the first moment.

[0015] Particularly preferably, when the vibration value of the track sensor and / or the chess and card entity measured by the track sensor changes from zero to a non-zero value and the height of the track sensor deviates from a first height, the track sensor determines this moment as the first moment. Preferably, the first height is the height of the digital prop usage platform. Preferably, the digital prop usage platform can be a mahjong table, etc. Through this configuration, the track sensor disposed within the digital prop can be moved by the user, and only when the track sensor is moved outside the digital prop usage platform, and the track sensor begins recording the track sensor's spatial coordinate information, etc., thereby enabling the track sensor to record the user's complete card-drawing action (particularly, the pause and thought process from drawing a card on the table to placing the chess and card entity back on the table, and the complete card-drawing trajectory). This avoids the situation where the track sensor is paused in mid-air, breaking down a complete card-drawing action into multiple sub-parts for recording, and thus easily losing potentially important digital biometric markers such as the mid-air pause or dwell time and the complete card-drawing trajectory.

[0016] Preferably, the second moment can be determined by a track sensor. Preferably, when the vibration value of the track sensor and / or the chess and card entity measured by the track sensor changes from a non-zero value to a value of zero, the track sensor determines that this time is the second moment.

[0017] Particularly preferably, when and only when the track sensor in the digital prop starts searching for the portable smart device around the track sensor at the second moment, the track sensor obtains the identifier of the portable smart device closest to the track sensor.

[0018] This configuration avoids the following situation: when a user moves a digital prop or a digital prop moves, the user's track sensor searches for multiple other users or other users' portable smart devices during the first moment or between the first moment and the second moment. This is because during the user's use of the digital prop, especially at the moment when a digital prop is used or moved (for example, the second moment), the digital prop is located close to the user using the digital prop, and thus can maintain a relatively long distance from the portable smart devices of other users, thereby avoiding the situation where the track sensor recognizes the same minimum distance to the portable smart devices of multiple users, that is, avoiding the situation where the user's track recording sensor searches for multiple other users or other users' portable smart wearable devices during the user's movement of the digital prop or the movement of the digital prop. In the present technical solution, the track sensor chooses to measure the distance between the track sensor and the portable smart devices around it at the second moment. That is, when a digital prop is used or moved, the digital prop is already located on the side close to the user using the digital prop, and can thus maintain a relatively long distance from the portable smart devices of other users. As a result, the distances between the user using the digital prop and the user not using the digital prop and the digital prop are significantly different, thereby improving the accuracy of the attribution of the spatial coordinate data collected each time for the digital prop.

[0019] For example, the track sensor begins measuring the distance between itself and surrounding portable smart devices and sends an identity request to the portable smart device closest to the digital prop. The portable smart device closest to the digital prop then sends an identifier to the track sensor inside the digital prop that identifies the identity of the user wearing the portable smart device.

[0020] When the track sensor receives the identifier of the portable smart device that is closest to the digital prop, the track sensor inserts the identifier into the motion data sent by the track sensor to the analysis unit and / or server at that time to identify the user from whom the motion data collected at that time originated.

[0021] Particularly preferably, the track sensor is only capable of measuring the distance to portable smart devices within a distance threshold range. Preferably, the distance threshold can be manually set based on the actual scenario, for example, the distance threshold can be 15 centimeters. This configuration significantly reduces the distance measurement requirements of the track sensor, thereby significantly reducing the amount of data collected by the track sensor (e.g., the distance to several nearby portable smart devices). This in turn enables the track sensor to more quickly identify the portable smart device closest to the current track sensor and ultimately obtain the identity information of the user associated with the portable smart device.

[0022] Preferably, the identifier can be composed of characters, numbers, etc. For example, the identifiers can be A1, A2, A3, A4, where a group of users uses the same set of digital props, the letter A can represent which group of users they belong to, and the number can identify the first user belonging to the corresponding user group.

[0023] Preferably, after the track sensor completes transmitting the current motion data to the analysis unit and / or cloud server through the wireless gateway, the track sensor automatically clears the recorded identifier so that the track sensor can receive the identifier from the portable smart device for identifying the identity information of the next user when it is used / moved by the user next time.

[0024] The track sensor sends the current movement data to the analysis unit and / or the server.

[0025] When the analysis unit and / or server receives the motion data, it first identifies the identifier in the motion data to identify the data source of the motion data, and then enters the motion data into the database of the user corresponding to the identifier in the motion data, thereby solving the data ownership problem of each digital prop collected each time.

[0026] Preferably, the motion data collected by the track sensor disposed in the digital prop can also be sent to an analysis unit and / or a server via the portable smart device.

[0027] Particularly preferably, the track sensor is also capable of measuring its own vibration and / or the vibration of the chess and card game entity. Preferably, the track sensor is capable of starting to record spatial coordinate data of the track sensor or chess and card game entity when the track sensor or chess and card game entity vibrates / moves. Preferably, the track sensor is capable of stopping recording spatial coordinate data of the track sensor or chess and card game entity when the track sensor or chess and card game entity stops vibrating / moving. Particularly preferably, the track sensor is capable of starting to record spatial coordinate data of the track sensor or chess and card game entity when the track sensor or chess and card game entity vibrates / moves and the track sensor is not at a first height, where the first height is the height of the chess and card game platform on which the user is using the digital prop. For example, the chess and card game platform may be a mahjong table. This configuration prevents the track sensor within the digital prop from collecting data during the digital prop's movement when the digital prop is being shuffled, for example, when the digital prop is not being used independently by the user. Such data (such as when the digital prop is shuffled by an automatic mahjong machine or when the digital mahjong game is manually shuffled by multiple users) cannot be used as initial data for individual user usage, thereby further reducing the collection of invalid initial data, thereby improving the quality of the initial data and reducing the data processing load on the server. Preferably, the spatial coordinate data collected by the track sensor can have a maximum accuracy of at least five millimeters. Preferably, the spatial coordinate data collected by the track sensor can be recorded for a maximum of ten seconds.

[0028] According to a preferred embodiment, the initial data can further include one or more of the following: interaction data, language data, eye movement data, pupil data, and daily data. The interaction data, language data, eye movement data, and pupil data are collected by the analysis unit during the user's interaction with the analysis unit, and the daily data is collected by the intelligent monitoring unit during the user's daily life activities.

[0029] According to a preferred embodiment, the digital prop may include: a chess and card entity and a track sensor. The track sensor is capable of at least acquiring spatial coordinate data of the chess and card entity during use by the user and a timestamp corresponding to the spatial coordinate data.

[0030] According to a preferred embodiment, the user is capable of wearing a portable smart device. The portable smart device is configured to at least send an identification code capable of identifying the user's identity information to the analysis unit and / or the server. The analysis unit and / or the server is capable of obtaining the identification code capable of identifying the user's identity information carried by the portable smart device.

[0031] According to a preferred embodiment, the portable smart device can also collect or be input with the user's physiological data. The physiological data may include one or more of heart rate, blood pressure, pulse oximetry, and body temperature. The portable smart device can send the individual characteristic data and / or physiological data to the analysis unit and / or the server. Preferably, the above physiological data can also be used as initial data. The individual characteristic data may include familiarity with the chess and card entity, education level, and medical history.

[0032] According to a preferred embodiment, the behavior data collected by the track sensor disposed in the digital prop can also be sent to the analysis unit and / or the server via the portable smart device.

[0033] According to a preferred embodiment, the analysis unit and / or the method for processing and analyzing to calculate the digital biomarker capable of characterizing the user's cognitive ability is:

[0034] extracting feature data from the initial data and further screening candidate digital biomarkers from the feature data;

[0035] Performing characteristic causal inference on the candidate digital biomarkers to obtain the digital biomarkers through computational analysis.

[0036] According to a preferred embodiment, the method for performing feature causal inference on the candidate digital biomarkers to infer effective digital biomarkers is:

[0037] A causal analysis knowledge base is constructed based on the characteristic data, and causal inference is performed on the candidate digital biomarkers through the established causal analysis knowledge base based on causal analysis theory, so as to obtain effective digital biomarkers through computational analysis.

[0038] Preferably, the method for performing causal inference on the candidate digital biomarkers based on the causal analysis theory through the established causal analysis knowledge base is:

[0039] S1: Literature unit builds the original literature library;

[0040] S2: Data unit constructs the dataset;

[0041] S3: Causal units construct causal relationships between symptoms;

[0042] S4: The knowledge unit stores the original document library, the data set and / or the average causal effect to construct the causal analysis knowledge base that can be read and / or displayed.

[0043] Preferably, the document unit is capable of acquiring numerous relevant documents containing various cognitive abilities and classifying them into a number of document units to construct the original document library, so that the data unit can acquire the main characteristic parameters based on the document units and construct a data set based on the main characteristic parameters. Through this configuration, the server can also generate a causal network model based on the characteristics of the relevant variables to perform causal reasoning, so as to identify the causal relationship between the relevant characteristic variables related to the changes in the user's cognitive abilities and the changes in cognitive abilities, as well as the strength of the causal relationship, thereby providing medical personnel or medical researchers with an effective way to identify effective digital biomarkers or pathological causes that cause changes in human individual cognitive abilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a simplified schematic diagram of module connection relationships of a preferred embodiment of the digital props and the server provided by the present invention;

[0045] Figure 2 This is a model display diagram of the digital prop of the present invention.

[0046] Reference Signs List

[0047] 1: Digital props; 2: Server; 3: Analysis unit;

[0048] 4: Intelligent monitoring unit; 5: Portable intelligent device; 101: Chess and card game entity;

[0049] 102: Track sensor. DETAILED DESCRIPTION

[0050] The following is a detailed description with reference to the accompanying drawings. The present invention provides a digital biomarker for assessing cognitive impairment. The initial data of the digital biomarker is formed by collecting behavioral data of various operations performed by a user using a digital prop 1, such as mahjong, playing cards, chess, or go. The collected initial data is sent to an analysis unit 3 and / or a server 2. The initial data is analyzed and calculated by an artificial intelligence program on the analysis unit 3 and / or server 2 to obtain a series of digital biomarkers that can characterize the user's cognitive abilities, such as perceptual cognition, sensory cognition, and thinking cognition. The obtained digital biomarkers can be used to label the real-time data collected when the user uses the digital prop 1, and the labeling can be used to assess the user's cognitive abilities.

[0051] Preferably, at least the motion data of the user moving the digital prop 1 and the touch data of the user contacting the digital prop 1 can be used as the user's behavior data on the digital prop 1. Preferably, the motion data of the user using the digital prop 1 can include the user's operations of moving the digital prop 1 and rotating the digital prop 1. Preferably, the motion data of the user using the digital prop 1 can also include at least the user's operations of lifting the digital prop 1, dropping the digital prop 1, touching the digital prop 1 to cause the digital prop 1 to fall, and the user's operations of righting the fallen digital prop 1, for example Figure 2 The digital prop 1 shown. Preferably, the touch data of the user contacting the digital prop 1 may include the magnitude, direction, and frequency of the force applied by the user to the digital prop 1, including the magnitude, direction, and frequency of the force generated by the user using the digital prop 1. Preferably, the digital biomarkers obtained through analysis can represent user attributes, which may include but are not limited to: cognitive state / ability, behavioral state / ability. Preferably, user attributes may also include but are not limited to learning and analytical ability, expression ability, memory ability, and psychological endurance.

[0052] Preferably, ordinary physical servers and / or cloud servers can serve as Server 2. Server 2 can analyze the collected initial data, including but not limited to behavioral data, and extract digital biomarkers, and use these digital biomarkers to assess the user's health status. Preferably, health status can include cognitive abilities, such as observation, attention, and imagination. Preferably, health status can also include learning ability, analytical ability, expression ability, memory ability, and psychological endurance. Preferably, health status can also include whether the user suffers from other diseases.

[0053] According to a preferred embodiment, Figure 1-2 As shown, the digital prop 1 can be a physical entity of various chess and card games, such as mahjong, chess, Chinese chess, go, playing cards, etc. The digital prop 1 can include a chess and card entity 101 and a track sensor 102. The number of digital props 1 is determined by the rules of the various chess and card games. The track sensor 102 is installed within the chess and card entity 101 of the digital prop 1, or the track sensor 102 is integrated into the chess and card entity 101, so that the chess and card entity 101 of the digital prop 1 is indistinguishable from an ordinary chess and card in appearance. The track sensor 102 can at least be used to obtain the spatial coordinates of the chess and card entity 101 during the user's use and the timestamp associated with the spatial coordinates, and transmit this information to the analysis unit 3 and the server 2.

[0054] Preferably, the process of a user using the chess and card entity 101 may include: moving the chess and card entity 101, rotating the chess and card entity 101, lifting the chess and card entity 101, dropping the chess and card entity 101, causing the chess and card entity 101 to fall over, and the user righting the chess and card entity 101. Preferably, the process of a user using the chess and card entity 101 may also include: detecting the magnitude, direction, and frequency of the force exerted on the chess and card entity 101 when touching the chess and card entity 101, and detecting the magnitude, direction, and frequency of the force exerted on other objects, such as a chess platform, by using the chess and card entity 101. Particularly preferably, the chess and card entity 101 may be a chess set, particularly a chess piece with a three-dimensional volume. Preferably, the chess and card entity 101 may also be other types of chess and cards, such as mahjong, go, xiangqi, or playing cards. Preferably, the track sensor 102 integrated or disposed within the chess and card entity 101 may be used to acquire three-dimensional spatial coordinate data of the digital prop 1 and transmit the acquired three-dimensional spatial data to the server 2 and / or the analysis unit 3. Preferably, the trajectory sensor 102 integrated or disposed within the chess and card entity 101 can capture spatial trajectory data of the user moving the digital prop 1 in three-dimensional space, such as above the chess and card platform, and / or in a two-dimensional plane, such as the chess and card platform, as well as spatial trajectory data of the user touching the digital prop 1 to cause it to fall and then righting the fallen digital prop 1. Preferably, the spatial trajectory data to be captured by the trajectory sensor 102 includes the duration of the user's grasping of the chess and card entity 101, the duration of the user's pause during the grasping of the chess and card entity 101, the distance the user moves the chess and card entity 101 over a corresponding period of time, the angular change caused by the user to cause the chess and card entity 101 relative to a normally placed chess and card entity 101, and the initial and final positions of the chess and card entity 101 relative to the user's palm when grasped.

[0055] According to a preferred embodiment, the user can wear a portable smart device 5 that matches the digital prop 1 and is equipped with an identifier that can send user identity information to the track sensor 102. Preferably, the track sensor 102 in the digital prop 1 can be connected to the portable smart device 5 wirelessly, such as via Bluetooth or a local area network, and the track sensor 102 obtains the identity information identifier configured on the portable smart device 5 through this connection.

[0056] Preferably, the user can choose to wear the portable smart device 5 24 hours a day or for a specific period of time. The portable smart wearable device 5 may include: a smart watch, a smart bracelet, a smart ring, a smart necklace, a smart helmet, etc. Preferably, the portable smart device 5 can be worn on the user's wrist, such as a smart watch or smart bracelet. Preferably, the portable smart device 5 can also be worn on other parts of the user's body, such as a smart ring on the user's finger, a smart necklace on the user's neck, and a helmet on the user's head.

[0057] Preferably, the user can wear a portable smart device 5 that matches the digital prop 1 and is equipped with an identifier that can send user identity information to the track sensor 102. When, and only when, the track sensor 102 in the digital prop 1 is at the second moment when the track sensor 102 and / or the chess and card entity 101 is stationary relative to the portable smart device 5, the track sensor 102 can perform a blanket search for surrounding portable smart devices 5, establish a connection with the portable smart device 5 closest to the track sensor 102 in a straight line, and obtain the identity information identifier configured on the portable smart device 5.

[0058] According to a preferred embodiment, the portable smart device 5 can autonomously collect and summarize the user's physiological data and / or the user's individual characteristic data, or can manually input and have the portable smart device 5 summarize and analyze the user's physiological data and / or the user's individual characteristic data. The user's physiological data may include heart rate, blood pressure, blood oxygen saturation, body temperature, and timestamps corresponding to each characteristic. After being collected and analyzed by the portable smart device 5, the physiological data is sent to the server 2. The server 2 determines the user's current physiological state based on the physiological data uploaded by the portable smart device 5. The user's individual characteristic data may include the user's familiarity with the chess and card entity 101 currently being used, the user's ability to think independently (IQ), and medical history information that may have an impact on cognitive ability.

[0059] Preferably, the portable smart device 5 is internally configured with an acceleration measurement device, which enables the portable smart device 5 to obtain information related to the user's activities, such as daily motion status. Preferably, the motion status may include the user's daily walking steps, the user's running distance, the user's jumping frequency, and the corresponding timestamps. Preferably, the motion status may also include the movement of the user's body parts, such as the shaking frequency, shaking amplitude, and corresponding timestamps of the user's body parts. Preferably, the user's body parts may include hands, legs, waist, shoulders, and neck. Preferably, the user activity-related information obtained by the acceleration measurement device configured in the portable smart device 5 can be autonomously collected by the portable smart device 5 and incorporated into physiological data.

[0060] According to a preferred embodiment, the system further includes an intelligent monitoring unit 4 capable of acquiring daily data of the user in daily life and sending the daily data to a server 2 connected to the intelligent monitoring unit 4 .

[0061] Preferably, the intelligent monitoring unit 4 can be used by the user to autonomously select and collect user behavior information throughout the day or during specific time periods as daily data. Preferably, the daily data can be sent by the intelligent monitoring unit 4 to the server 2, and the server 2 determines the user's real-time status based on the user's daily data uploaded by the intelligent monitoring unit 4. Preferably, the real-time status can include the current state of the user's cognitive abilities, such as observation, attention, and imagination. Preferably, the real-time status can also include the user's current physical health status. Preferably, the intelligent monitoring unit 4 can include, but is not limited to, the following devices: water immersion sensors, ultrasonic sensors, microwave sensors, voice recorders, etc.

[0062] For example, daily data can include user behavior information collected by water sensors installed in the user's toilet, kitchen, and / or other areas with access to water sources, such as the frequency of forgetting to turn off the faucet and the duration between opening and closing the faucet. This type of daily data can serve as a basis for assessing the user's memory status. Ultrasonic sensors can be used to determine the user's location coordinates, and microwave sensors can be used to determine the user's movement. Ultrasonic and microwave sensors installed within the user's daily activity range can thus obtain information such as the direction, speed, and distance of the user's movement, which can be used to assess the user's observation, attention, and mobility. Voice recorders installed within the user's daily activity range can be used to record user voice information and determine the user's attention based on its frequency, speed, and voice information. The user's instantaneous or delayed memory for speech can also be assessed based on information such as repeated sentences and keywords. Preferably, daily data can also include risk factors and physiological indicators. Preferably, risk factors can include, but are not limited to, smoking frequency and drinking frequency. Preferably, physiological indicators can include, but are not limited to, heart health and brain health. Preferably, daily data can also include other motion data required for cognitive ability assessment. According to a preferred embodiment, an analysis unit 3 is further provided or integrated within a mobile intelligent platform, such as a smartphone or tablet computer. The analysis unit 3 can record and analyze the interaction process between the user and the analysis unit 3, and organize and summarize the interaction data between the user and the analysis unit 3, the user's language data, the user's eye movement data, and the user's pupil data. The above data is sent to the server 2 through the analysis unit 3. The server 2 analyzes and evaluates the interaction process between the user and the analysis unit 3, and stores and backs up the evaluation data obtained from the analysis.

[0063] Preferably, the interaction data is data generated when the user manipulates the chess and card entity 101 of the digital prop 1 or interacts with the analysis unit 3, such as click data, frequency data, action continuity data, trajectory data, information content selection data, sliding selection data, and content change logical relationship data. Preferably, the analysis unit 3 provided or integrated within the mobile smart device can identify, analyze, and extract language data for analyzing and determining the user's emotional state, emotional information, and other information. Preferably, the language data can include voice information and graphic information generated by the user's regular interactions with others or with the smart device through the mobile smart device.

[0064] Preferably, the user-generated graphic and text information collected by the analysis unit 3 within the mobile smart device may include graphic and text information used for communication, spoken and written language, and / or images or text that express the language, such as commonly used emoticons and emojis that can more vividly express the user's emotions. By being disposed within or integrated within the mobile smart device, the analysis unit 3 can connect to the network to identify and analyze the graphic and text information, thereby directly or indirectly obtaining the user's real-time emotional state and emotional information, thereby facilitating the determination of the user's real-time cognitive ability.

[0065] Preferably, the voice information is the voice generated when a user interacts with an analysis unit 3 disposed in or integrated within a mobile smart device. Users typically communicate with others or directly with the mobile smart device through the mobile smart device. The voice information generated by the communication includes voice commands, single words, and short phrases. Preferably, the analysis unit 3 is capable of performing low-level processing on the voice and extracting different phonemes and repeated combinations of phonemes in the voice to identify and extract the characteristics of each voice, including volume, pitch conversion, and audio pitch. The analysis unit 3 can send such collected and analyzed data to the server 2.

[0066] Preferably, eye movement data can be acquired by an electrooculogram sensor or other related eye movement sensor provided in the analysis unit 3. The electrooculogram sensor or other related eye movement sensor in the analysis unit 3 can capture the user's eye movement status via the camera of the mobile smart device. Preferably, the eye movement data can include, but is not limited to, data related to the user's eye gaze, eye saccades, and eye tracking. Preferably, the eye movement data can also include the coordinates of the user's gaze point, the duration of the gaze point, pupil data at the gaze point, etc. The eye movement data acquired by the analysis unit 3 via the electrooculogram sensor or other related eye movement sensor can be sent to the server 2 via the analysis unit 3.

[0067] Preferably, the pupil data of the user captured by the electrooculogram sensor or other related eye movement sensor in the analysis unit 3 through the camera of the mobile smart device includes but is not limited to a series of changes in the pupils when the user touches the chess and card entity 101 during the operation of the digital prop 1, hears or sees the chess and card entity 101 played by others, such as changes in pupil size, change speed and focusing frequency.

[0068] Preferably, the analysis unit 3, located or integrated within the mobile smart device, can autonomously collect user interaction data, language data, eye movement data, and pupil data through the mobile smart device. Alternatively, the analysis unit 3 can manually input data from the input port of the mobile smart device and organize and summarize the data. Preferably, the analysis unit 3 can be located or integrated within a mobile smart device such as a mobile phone or tablet. Preferably, the analysis unit 3, located or integrated within the mobile smart device, can autonomously analyze the operating status of the electrooculogram sensor or other related eye movement sensors within the unit 3 on the mobile smart device based on the user's actual needs to obtain the corresponding interaction data, language data, and / or eye movement data required by the user.

[0069] For example, the analysis unit 3 disposed within or integrated within a mobile smart device can assess the user's cognitive ability by acquiring input information generated when interacting with the mobile smart device carrying the analysis unit 3 and comparing the input information with information input standards established by the analysis unit 3 and / or the server 2. The information input standards established by the analysis unit 3 and / or the server 2 may include: semantic clarity, whether the trajectory memory is clear, the length of the test time, whether the input information exceeds a specified range (e.g., a number between 1 and 9 should be entered, but the actual input exceeds this range), etc. For another example, by disposing the analysis unit 3 within or integrating it into a leisure app on a mobile smart device (e.g., a mobile phone or tablet), the mobile smart device can acquire interaction data and language data generated by the user with the relevant app during leisure time. The analysis unit 3 disposed within or integrated within the mobile smart device can also capture the action information and trajectory information generated by the user when interacting with the chess and card entity 101 by interacting with the trajectory sensor 102 within the digital prop 1. For example, a user may be asked to draw a clock or the shape of a current chess and card entity 101 on a leisure app installed or integrated within a mobile smart device and having an analysis unit 3 installed or integrated therein. This method can obtain more evaluation information than traditional scale assessments of clock drawing. User operation data is captured by sensors connected to the analysis unit 3 and provided on the screen of the mobile smart device installed or integrated therein and on an electronic stylus that interacts with the mobile smart device. The user operation data on the app may include, but is not limited to: the force with which the user uses the electronic stylus to contact the touch screen of the analysis unit 3 when drawing on the touch screen, the speed and / or pause time of the tip of the electronic stylus moving across the touch screen when drawing, the shape of the handwriting formed by the user using the electronic stylus, the pressure applied by the user to the stylus body, and the humidity of the stylus body when the user holds it. The above operation data is stored in the analysis unit 3 itself or sent to the server 2 via the mobile smart device, so that the analysis unit 3 itself or the server 2 can conduct a preliminary assessment of the digitization of the user's cognitive ability and health status.

[0070] Preferably, the evaluation data generated by the analysis unit 3 and / or server 2 for the preliminary evaluation of the user's cognitive ability and health status includes, but is not limited to: the force with which the user uses the electronic stylus to contact the touch screen of the analysis unit 3 when drawing with the electronic stylus, the speed and / or pause time of the tip of the electronic stylus moving across the touch screen while drawing, the shape of the handwriting formed by the user using the electronic stylus, the pressure applied by the user to the stylus body, the humidity of the stylus body when the user grips the stylus, the clarity of the user's voice when interacting with the analysis unit 3 during the evaluation process, the smoothness of the trajectory formed by the movement of the digital prop 1, the time required to complete an evaluation, etc. The analysis unit 3 or server 2 compares the acquired evaluation data with normal data generated by a normal person performing the same operation under normal conditions collected from big data to preliminarily determine the user's cognitive ability and health status.

[0071] Through this configuration, not only can the analysis unit 3 or the server 2 compare the data information such as the gesture trajectory formed on the screen of the mobile smart device when the user uses the mobile smart device with an internal setting or integrated analysis unit 3, the strength of the electronic stylus corresponding to the mobile smart device, and other data information, and compare such information horizontally with the data information of other people with normal cognitive abilities who perform the same operations collected by the analysis unit 3 or the server 2 set up or integrated inside the mobile smart device, but also the physical health status can be preliminarily judged based on the information such as the pressure applied by the user to the stylus body and whether the palm of the hand sweats when holding the stylus. What is more important is that the analysis unit 3 or the server 2 can monitor and evaluate the changing trends of the user's cognitive ability and physical condition at different periods, thereby avoiding evaluating the test subjects with originally low cognitive ability as having insufficient cognitive ability and simply predicting the user's physical condition. For example, on a mobile smart device equipped with or integrated with an analysis unit 3, a leisure app can be used to randomly cut a chess and card entity 101, such as a mahjong tile, into several fragments, and then shuffle the order. The user is then asked to determine which mahjong tile these fragmented images are cut from, and then the corresponding mahjong tile is selected from a number of mahjong tiles to obtain evaluation data corresponding to the user's visual-spatial ability and thinking ability. For another example, the user is asked to completely piece together several simple randomly cut fragments to analyze the user's memory and visual-spatial ability. Through this configuration, the user's operation, voice, and other data can be obtained through interaction with the mobile smart device equipped with or integrated with the analysis unit 3, or the leisure app on the mobile smart device, or other evaluation device, so as to score or evaluate the user's cognitive ability, such as visual-spatial ability, thinking ability, or memory.

[0072] Both the analysis unit 3 and the server 2 may have data processing functions. The data processing function of the analysis unit may be set to the first data processing stage, and the data processing function of the server 2 may be set to the second data processing stage.

[0073] The first data processing stage of the analysis unit 3 set inside the mobile smart device is generally divided into the following steps: questions raised by the user or user feedback can be submitted to the analysis unit 3 inside the mobile smart device. The analysis unit 3 will make a preliminary judgment on the information submitted by the user through the input end of the mobile smart device. The preliminary judgment may include whether the date of birth corresponds to the ID card, whether the scores of each cognitive assessment ability are within the specified numerical range, whether the input data is clear, timeout, etc. The numerical range interval judgment; after completing the preliminary judgment, the analysis unit 3 will perform a logical check to determine whether the information submitted by the user is contradictory; after passing the logical check, the analysis unit 3 sends the information submitted by the user and the cognitive ability score of the user's corresponding cognitive domain to the server 2. The server 2 sums the scores of each analysis unit 3, and compares the total score with the cognitive assessment level corresponding to the set score interval to achieve an assessment of the subject's cognition.

[0074] The second data processing stage of the server 2 which is always connected to the portable smart device 5, the track sensor 2 inside the chess and card entity 101 of the digital prop 1, the analysis unit 3 set or integrated inside the mobile smart device, and the smart monitoring unit 4 distributed in the user's daily life space may include the following steps: based on multimodal data fusion technology, the server 2 processes the user's motion data captured by the track sensor 102 inside the chess and card entity 101 of the digital prop 1, the interaction data, language data, eye movement data, pupil data sent by the analysis unit 3, the sensor data cognitive ability score data collected by the sensor of the smart monitoring unit 4, and the portable smart device 5 The captured user body motion data is fused; after the server 2 completes the data fusion, a feature extraction method based on machine learning (such as partial least squares, autoencoder algorithm and its derivative algorithm, adversarial network learning algorithm and its derivative algorithm, etc.) is used to extract digital biomarkers that may be used to characterize the user's cognitive state; after the extraction of digital biomarkers is completed, the extracted digital biomarkers are associated with the cognitive domain to obtain relevant digital biomarkers that can characterize the user's cognitive state, and the importance of each digital biomarker is sorted according to the weight of each digital biomarker in the feature extraction algorithm, so as to achieve comprehensive and reliable cognitive screening and help early detection of cognitive impairment. Through this configuration, the analysis unit 3 is set or integrated into the leisure APP on the mobile smart device. Not only can the question and answer information between the user and the analysis unit 3 be obtained through voice communication between the mobile smart device and the user, but also the action information when the user interacts with the mobile smart device or the leisure APP on the mobile smart device, the strength of the electronic stylus corresponding to the mobile smart device, and other data information can be captured through the screen of the mobile smart device. For example, a user can draw a clock or the shape of a current chess and card entity 101 on a leisure app installed or integrated within a mobile smart device and having an analysis unit 3 built into it. This method can obtain more evaluation information than traditional scale evaluation of drawing a clock. The user's operation data can be captured by sensors connected to the analysis unit 3 and provided on the screen of the mobile smart device installed or integrated with the analysis unit 3 and on the electronic stylus that interacts with the mobile smart device. The user's operation data on the app may include, but is not limited to: the force with which the electronic stylus touches the touch screen of the mobile smart device with the built-in analysis unit 3 when the user uses the electronic stylus to draw on the touch screen, the speed and / or pause time of the tip of the electronic stylus across the touch screen when drawing, the shape of the handwriting formed by the user using the electronic stylus, the pressure applied by the user to the electronic stylus body, and the humidity of the electronic stylus body when the user holds the electronic stylus. The above operation data is stored in the analysis unit 3 itself or sent to the server 2 via the mobile smart device. The acquisition of the above operation data is more comprehensive and multidimensional, and the extracted feature data is also more comprehensive and multidimensional.The analysis unit 3 itself or the server 2 can also make a more comprehensive and objective assessment of the user's cognitive ability and health status through more comprehensive and multi-dimensional feature data.

[0075] Particularly preferably, the analysis unit 3 integrated within the mobile smart device can establish interaction with the user through voice, image, and physical contact, such as touch, to obtain language data, image data, and the like, when the user grasps the digital prop 1 (particularly, when the user grasps the chess and card entity 101 equipped with a track sensor 102). For example, the analysis unit 3 integrated within the mobile smart device can ask the user questions related to the rules of the chess and card entity 101 of the current digital prop 1. For example, using chess as an example, the analysis unit 3 integrated within the mobile smart device can issue a voice question or display a text message to the user through the mobile smart device: "Please explain the next possible moves for this chess piece." If the user responds verbally, the analysis unit 3 integrated within the mobile smart device receives the voice through the mobile smart device's earpiece and performs semantic conversion on the voice, converting it into information recognizable by the analysis unit 3. If the user types an answer through the mobile smart device, the analysis unit 3 integrated within the mobile smart device receives the answer from the mobile smart device input terminal and converts it into information recognizable by the analysis unit 3. For the answer information given by the user, the analysis unit 3 inside the mobile smart device judges the answer information and scores the user's answer based on the judgment result, and stores the result locally and uploads it to server 2. The typed answer information is also judged and scored, stored locally and uploaded to server 2. Server 2 can determine the score through machine learning algorithm.

[0076] According to a preferred embodiment, the motion data of the user when using the digital prop 1, the interaction data generated when the user interacts with the analysis unit 3, the language data, the eye movement data, the pupil data, and the daily data of the user obtained by the intelligent monitoring unit 4 can be obtained by the analysis unit 3 and the server 2 respectively. The various types of data obtained above are stored in the analysis unit 3 and / or the server 2, and are used for subsequent data analysis by the server 2 and / or the analysis unit 3 through artificial intelligence algorithms to calculate digital biomarkers that can characterize the user's cognitive impairment and other diseases.

[0077] According to a preferred embodiment, after the analysis unit 3 and / or the server 2 completes data processing, the analysis unit 3 set or integrated in the mobile intelligent platform or the server 2 storing data from each component analyzes the data after the second stage of data processing by the server 2, and the process is as follows: the spatial trajectory data of the user moving the digital prop 1 in three-dimensional space, such as above the chess and card platform and / or in a two-dimensional plane, such as the chess and card platform, captured by the trajectory sensor 102 integrated or set inside the chess and card entity 101, the spatial trajectory data of the user touching the digital prop 1 to make the digital prop 1 fall over, and the spatial trajectory data of the fallen digital prop 1 being righted, and the motion data formed by the corresponding timestamps are input into the deep learning network to extract feature data related to characterizing the user's cognitive ability; the user's language data obtained by the user using the analysis unit 3 is converted into feature data through technical means such as Fourier transform and frequency domain analysis, and the feature data is further extracted from the above data through stack autoencoding; the daily data of the user obtained by the intelligent monitoring unit 4 is extracted from possible feature data through ontology learning. When the server 2 or analysis unit 3 obtains the user's motion data when using the digital prop 1, the user's interaction data, language data, eye movement data, pupil data recorded when using the analysis unit 3, and the user's daily data obtained by the intelligent monitoring unit 4 for data analysis (i.e., extraction of digital biomarkers), the data analysis is completed by artificial intelligence algorithms. The main steps are as follows: the spatial trajectory data captured by the user using the digital prop 1 in three-dimensional space, such as above the chess and card platform, and / or in a two-dimensional plane, such as the chess and card platform, captured by the trajectory sensor 102 integrated or installed in the chess and card entity 101, the spatial trajectory data of the user touching the digital prop 1 to cause the digital prop 1 to fall, and the spatial trajectory data of the fallen digital prop 1 being righted, and the corresponding timestamps are input into the deep learning network to extract feature data related to characterizing the user's cognitive ability or motion state; the daily data of the user obtained by the intelligent monitoring unit 4 wirelessly connected to the server 2 (such as water immersion, ultrasound, and pressure data obtained when the user uses the toilet) is extracted from it through ontology learning to characterize the user's attributes. Preferably, user attributes may include but are not limited to: cognitive state / ability, action state / ability. Preferably, user attributes may also include but are not limited to learning and analytical ability, expression ability, memory ability, and psychological endurance.

[0078] According to a preferred real-time method, the above-mentioned feature data may not have the characteristics of becoming a digital biomarker due to some reasons, so the analysis unit 3 and / or server 2 are required to screen the above-mentioned extracted features to generate digital biomarkers that can accurately characterize the user's cognitive ability or health status. After obtaining accurate digital biomarkers, it is still necessary to further sort the importance of the digital biomarkers, and the sorting method can optionally rely on marginal contribution analysis to complete. Preferably, the feature selection method of effective digital biomarkers can include one or more of partial least squares, variational autoencoders, and adversarial network learning. Preferably, the feature selection method of effective digital biomarkers can also adopt other categories of methods, such as wrapper methods (Wrapper) and embedded methods (Embedding). After the server 2 or the analysis unit 3 preliminarily extracts the above-mentioned feature data, the server 2 or the analysis unit 3 can perform feature selection on all the above extracted features by partial least squares:

[0079] Y=X*(X T S(T T X T S) -1 T T Y+R e (1), Where S is the vector mapped by the independent variable X, T is the vector mapped by the dependent variable Y, and R eThe analysis unit 3 and / or server 2 receives various data transmitted from the track sensor 102 within the chess and card entity 101 of the digital prop 1, the intelligent monitoring unit 4 distributed within the user's activity space, the portable intelligent device 5 that captures the user's body movement status, and the data collected by the analysis unit 3 set or integrated within the mobile intelligent device. The analysis unit 3 and / or server 2 use the least squares method to accurately and effectively screen the extracted feature data that can be used to roughly characterize the user's cognitive ability and physical health status, thereby generating digital biomarkers that can accurately and effectively characterize the user's cognitive ability or health status. After obtaining accurate and effective digital biomarkers, the digital biomarkers still need to be further ranked by importance, and the ranking method can optionally be completed by marginal contribution analysis. At the same time, the server 2, which is constantly connected to the portable smart device 5, the track sensor 102 within the chess and card entity 101 of the digital prop 1, the analysis unit 3 installed or integrated within the mobile smart device, and the intelligent monitoring units 4 distributed throughout the user's daily living space, can obtain in real time the preliminary scores of the user's cognitive ability from each analysis unit 3 installed or integrated on the mobile smart device. The preliminary scores of the analysis units 3 can assist in identifying or mining digital biomarkers that can accurately represent changes (including declines and improvements) in the human user's cognitive ability. This setting can improve the success rate of screening and extracting digital biomarkers that may accurately represent changes (including declines and improvements) in the human user's cognitive ability or other diseases from the user's motion data when using the digital prop 1, interaction data when interacting with the analysis unit 3, language data, eye movement data, pupil data, and the user's daily data obtained by the intelligent monitoring unit 4.

[0080] According to a preferred embodiment, a user causal analysis knowledge base is constructed based on all feature data proposed, including user operation motion data provided by the track sensor 102 inside the chess and card entity 101 when the user operates the chess and card entity 101 of the digital prop 1, interaction data generated by the interaction between the user and the mobile smart device provided by the analysis unit 3 set or integrated inside the mobile smart device, language data, eye movement data, pupil data, daily data provided by the intelligent monitoring unit 4 connected to the server 2 in real time and distributed in the user's activity space, and the user's own motion data provided by the portable smart device 5 worn on the user's body. The establishment of this causal analysis knowledge base helps to perform causal analysis on digital biomarkers based on causal analysis theory, thereby effectively judging the accuracy of the digital biomarkers in representing changes (including decline and improvement) in the cognitive ability of human users, and helps to discover digital biomarkers that can effectively represent changes (including decline and improvement) in the user's cognitive ability and physical health status.

[0081] The behavioral data of the user's operation of the digital prop 1 can also be used as a basis for evaluating the user's cognitive ability. This is referred to as a cognitive ability evaluation method based on the digital prop 1. The method uses the motion data generated by the user's operation of the digital prop 1 to obtain the interaction data, language data, eye movement data, and pupil data generated when the user interacts with the analysis unit 3 through the analysis unit 3, and obtains the user's daily data within the scope of daily activities through the intelligent monitoring unit 4. The server 2 obtains the above-mentioned motion data through the trajectory sensor 102, the interaction data, language data, eye movement data, pupil data through the analysis unit 3, and the user's daily data through the intelligent detection unit 4 to extract possible digital biomarkers. Based on all the acquired feature data, a user causal analysis knowledge base is constructed. The establishment of this causal analysis knowledge base facilitates causal analysis of digital biomarkers based on causal analysis theory, thereby effectively determining the accuracy of the digital biomarkers in representing changes (including declines and improvements) in human users' cognitive abilities, and helps to discover digital biomarkers that can effectively represent changes (including declines and improvements) in users' cognitive abilities and physical health status.

[0082] The present invention also provides a method for performing causal inference from candidate digital biomarkers based on causal learning. The method comprises:

[0083] S1: Literature unit builds the original literature library;

[0084] S2: Data unit constructs the dataset;

[0085] S3: Causal units construct causal relationships between symptoms;

[0086] S4: The knowledge unit stores original literature, data sets, and / or average causal effects to construct a knowledge base that can be read and / or displayed. Thus, the information provided by the knowledge base can be provided to medical professionals in a quantitative data format for reference, learning, and / or decision-making.

[0087] Preferably, the document unit can retrieve and collect a large number of relevant documents containing multiple cognitive abilities and classify them through a machine learning algorithm to form a number of document units to construct an original document library, so that the data unit can obtain data provided by the track sensor 102 in the chess and card entity 101 of the digital prop 1, the portable smart device 5, the analysis unit 3 set or integrated in the mobile smart device, and the intelligent monitoring unit 4 based on the document unit, and the server 2 connected to each component at all times analyzes the provided data to extract the main feature parameters and construct a data set based on the main feature parameters, thereby reducing the interference of the huge feature parameters formed by the massive relevant documents on the causal relationship between cognitive ability and disease and improving the utilization value and utilization efficiency of the original document library.

[0088] Preferably, the data provided by the track sensor 102 within the chess and card entity 101 of the digital prop 1, the portable smart device 5, the analysis unit 3 provided or integrated within the mobile smart device, and the intelligent monitoring unit 4 are analyzed by the server 2, which is constantly connected to each component, to extract the main characteristic parameters and data sets. The causal unit constructs a Bayesian network capable of deriving the average causal effect between cognitive abilities through data pattern analysis, thereby enabling the knowledge unit to construct a knowledge base based on relevant literature by forming a corresponding relationship between cognitive abilities and the average causal effect between cognitive abilities. For example, the average causal effect between cognitive abilities can reflect whether the cognitive abilities constitute complications and comorbidities.

[0089] Preferably, the main characteristic parameters can be data provided by the track sensor 102 within the chess and card entity 101 of the digital prop 1, the portable smart device 5, the analysis unit 3 set or integrated in the mobile smart device, and the intelligent monitoring unit 4, and the server 2, which is constantly connected to each component, analyzes and extracts the provided data to form digital biomarkers that can characterize changes in the human user's cognitive ability and physical health status. Preferably, the data set used to construct the knowledge base can include but is not limited to voice feature data obtained by the server 2 through the analysis unit 3, behavioral feature data and physiological indicator data obtained by the track sensor 102 and the portable smart device 5, and risk factor data obtained by the intelligent monitoring unit 4. Preferably, the behavioral feature data can include motion data of the digital prop 1 and human motion data. Preferably, the behavioral feature data can also include other user behavior information collected by the portable smart device 5, such as the number of steps the user takes per day, sleep time, and other data information.

[0090] Preferably, the causal unit can also analyze the direct causal effects between cognitive abilities through data patterns, so that the knowledge unit can construct a knowledge base based on relevant literature in a way that forms a correspondence between cognitive abilities and the direct causal effects between cognitive abilities.

[0091] Preferably, the primary characteristic parameters used by the causal unit may include, but are not limited to, motion data from the user using the digital prop 1, the user's own motion data, interaction data generated when the user interacts with the analysis unit 3, language data, eye movement data, pupil data, and daily data within the user's daily activities provided by the intelligent monitoring unit 4. Preferably, the document unit is capable of retrieving and collecting numerous relevant documents covering various cognitive abilities and classifying them using a machine learning algorithm to form a number of document units to construct an original document library. This allows the data unit to obtain primary characteristic parameters based on the document units and construct a data set based on the primary characteristic parameters, thereby reducing the interference of the large number of characteristic parameters generated by the vast amount of relevant documents on the causal relationship between cognitive abilities and symptoms and improving the utilization value and efficiency of the original document library. The relevant documents include medical records, research reports, conference papers, journal articles, books, academic papers, and patents. The documents required to construct the original document library are quite large. In order to better manage the documents, more effectively observe the relationship between cognitive abilities, and reduce the operating load of each component, it is necessary to classify the vast amount of documents required to construct the original document library. There are various classification criteria, such as disease categories such as digestive tract diseases, cardiovascular diseases, and neurological diseases, or academic fields such as rehabilitation and psychology. Literature classification is also a crucial process, as it can directly impact the distinction between complications and comorbidities. Therefore, accurately and efficiently classifying the vast amount of literature is a key issue that needs to be addressed. Bayesian, SVM, and k-NN methods are preferred for literature classification.

[0092] Preferably, the relevant documents are classified as follows: S11: The document unit counts the frequency of words / phrases in each document and obtains the joint occurrence probability of the words / phrases according to the independence assumption. For example, for a specific document, its joint occurrence probability distribution can be calculated using the Naive Bayes method.

[0093] S12: The document unit calculates the strength of the association of words / phrases. The calculation of the strength of association can reflect the association of words / phrases, which is suitable for document classification. Preferably, when classifying, define N as the set of document samples, V as the set of document types, and Vi as the subset of the i-th document type. W is the set of words / phrases, and Wi is the subset of the i-th word / phrase. Vi contains Sj samples, where the association reduced coordinate Tp of the p-th sample is an n-dimensional array:

[0094]

[0095] Among them, the number of occurrences of the i-th word in ki (i = 1, 2, 3, ... n), Normalization coefficient.

[0096] The association vector in Vi is the average of the reduced coordinates of all sample associations in Vi. This value reflects the strength of the association between words / phrases in the document, namely:

[0097]

[0098] S13: Document unit 1 obtains the associated reduced coordinates of the document, and classifies the related documents in an iterative algorithm using a classification function constructed based on the associated reduced coordinates of all related documents to form several document units. Preferably, for any document, its associated reduced coordinates are:

[0099]

[0100] Where qi is the number of occurrences of the i-th word in the document. When classifying, the distance between the document to be classified and the supporting points (b1, b2, ..., bn) of each category of document Vi is recorded as:

[0101]

[0102] According to the strength of correlation, construct a document classification function:

[0103]

[0104] Where γi is related to the strength of the correlation.

[0105] Preferably, the iterative algorithm can adopt a minimization iterative algorithm, a minimum optimization iterative algorithm and an expected maximum iterative algorithm. Preferably, the classification function can perform deep learning based on the sample size of relevant documents, thereby enhancing the accuracy of the document unit.

[0106] Preferably, the document unit is capable of retrieving and collecting a large number of relevant documents containing various cognitive abilities and classifying them through a machine learning algorithm to form a number of document units to construct an original document library, so that the data unit can obtain data provided by the track sensor 102 in the chess and card entity 101 of the digital prop 1, the portable smart device 5, the analysis unit 3 set or integrated in the mobile smart device, and the intelligent monitoring unit 4 based on the document unit, and the server 2 connected to each component at all times analyzes the provided data to extract the main feature parameters and construct a data set based on the main feature parameters, thereby reducing the interference of the huge feature parameters formed by the massive amount of relevant documents on the causal relationship between cognitive abilities and symptoms and improving the utilization value and utilization efficiency of the original document library. Preferably, when the data unit obtains the document unit, the data unit obtains the data set in a pair-wise manner based on cognitive ability pairs. The data unit extracts the relationship between the cognitive ability pairs in each relevant document using a natural language processing syntactic analysis method to establish a relationship knowledge base of cognitive ability pairs. The relationship between cognitive ability pairs includes positive relationships, negative relationships, and vertical relationships. Furthermore, the data unit searches within the document unit for documents containing cognitive ability pairs based on the relational knowledge table. This fusion process obtains the relational reliability values ​​for the cognitive ability pairs and establishes a relational reliability database for the cognitive ability pairs. The relationships between cognitive ability pairs include positive, negative, and vertical relationship reliability values. Thus, the data unit constructs a dataset based on the relational knowledge database and relational reliability database established for all cognitive abilities in a pairwise manner. For example, within the relevant literature, a condition or behavior L1 and a cognitive state L2 are obtained. The relationship between condition or behavior L1 and cognitive state L2 may be a positive relationship, meaning that condition or behavior L1 affects cognitive state L2, denoted as L1→L2. The relationship between condition or behavior L1 and cognitive state L2 may also be an inverse relationship, meaning that cognitive state L2 affects condition or behavior L1, denoted as L2→L1. The relationship between condition or behavior L1 and cognitive state L2 may also be a vertical relationship, meaning that cognitive state L2 and condition or behavior L1 do not affect each other, denoted as L1⊥L2. Since complications or comorbidities are multiple, they may also include several cognitive abilities such as another symptom or behavior L3 and cognitive state L4. According to the above relationship between cognitive abilities, a relationship knowledge base between a certain symptom or behavior L1 and another symptom or behavior L3, a relationship knowledge base between cognitive state L2 and another symptom or behavior L3, and so on can be constructed. Then, within the unit document body, a relationship credibility value library is constructed based on the above relationship knowledge base according to the contents in different documents. Preferably, the sum of the positive relationship credibility value, the negative relationship credibility value and the vertical relationship credibility value is normalized. That is, within the unit document body, all documents are traversed and queried, and the positive relationship credibility value, the negative relationship credibility value and the vertical relationship credibility value are weighted according to frequency.The data unit constructs a data set using the above relationship knowledge base and relationship credibility value base and inputs it into the causal unit to proceed to the next step.

[0107] Preferably, for journal articles, the reliability value of the positive relationship L1→L2 can also be defined as follows:

[0108]

[0109] Here, C(Xi) is the credibility of document Xi, using the formula: C(Xi) = (IFi + 1) × (CIi + 1), where Xi represents the i-th document, IFi is the normalized impact factor of the journal in which document Xi is published, and CIi is the normalized citation count. If no document has a relationship between L1 and L2, then KL(L1 → L2) = 0, KL(L2 → L1) = 0, and KL(L1 ⊥ L2) = 1. Other types of documents can be defined in the same way. For example, medical records can be defined based on the authority of the doctor. Conference papers can be defined based on the authority of the conference, and so on.

[0110] Preferably, the causal unit constructs a Bayesian network based on the main characteristic parameters and the data set. Preferably, the main characteristic parameters include the forward relationship reliability value, the reverse relationship reliability value, and the vertical relationship reliability value. The causal unit constructs the Bayesian network in the following manner:

[0111] S31: Preferably, define the dataset D = (D1, D2 ... Di) as a set of cognitive abilities, and L = (L1, L2 ... Ln) as a specific set of cognitive abilities in a set of cognitive abilities. Construct a Bayesian network evaluation function based on the relational knowledge base:

[0112] logP(G,D,KL)=logP(G)+logP(D|G)+logP(KL|G)

[0113] Where G is a Bayesian grid, whose value includes a directed acyclic graph with a specific set of cognitive abilities of a certain group L = (L1, L2...Ln) as nodes. Among them, P(G) is the prior distribution. According to existing knowledge, the maximum value of logP(G) + logP(D|G) is equivalent to logP(G|D). LogP(G|D) can be scored according to the Bayesian Information Criterion BIC. Where,

[0114]

[0115] If any edge in structure G is represented as Lm→Ln, then KL(GLm, GLn) = KL(Lm→Ln). KL(Lm→Ln) is the relationship credibility value. The sum in the formula is the sum of the literature knowledge credibility of the positive relationships corresponding to all directed edges in structure G.

[0116] S32: Construct an undirected graph structure constraint based on a relational knowledge base. For a given dataset D, for any cognitive ability pair Lm and Ln in D, obtain the cognitive ability pair number of the attribute pair Lm and Ln by searching the cognitive ability pair relational knowledge base. Then, retrieve the Lm→Ln and Ln→Lm relationship reliability values ​​from the relational reliability value table of cognitive ability pairs Lm and Ln in the literature based on the cognitive ability pair number. If L1 affects L2, then the connection relationship is L1 connecting L2 and pointing to L2. A directed edge is constructed between L1 and L2, and a positive relationship reliability value is assigned. If L2 affects L1, then the connection relationship is L2 connecting L1 and pointing to L2. A directed edge is constructed between L2 and L1, and a negative relationship reliability value is assigned. If L2 does not affect L1, the two are not connected, and a vertical relationship reliability value is assigned.

[0117] S33: Construct a Bayesian network based on the Bayesian network evaluation function and undirected graph structure constraints. After determining the undirected graph structure constraints of the Bayesian network, a heuristic search algorithm, such as the K2 algorithm, can be executed to find the network structure with the optimal scoring function. The general steps are: start the search from the initial model. At each step of the search, first use the search operator to locally modify the current model to obtain a series of candidate models. Then calculate the score of each candidate model, and compare the optimal candidate model with the current model. If the score of the optimal candidate model is higher, it will be used as the next current model and the search will continue; otherwise, the search will be stopped and the current model will be returned. According to the Bayesian principle, the candidate model with the highest score is the Bayesian network. Preferably, a Bayesian network evaluation function is constructed based on the established Bayesian network and the Bayesian rule. The Bayesian network evaluation function can be constructed based on classic heuristic structure learning algorithms, such as the K2 algorithm, the Max-Min Parents and Children algorithm, and the Markov chain Monte Carlo search.

[0118] The causal unit is based on mining the average causal effect between cognitive abilities through data patterns. This allows the unit to determine whether the average causal effect between cognitive abilities constitutes a complication or comorbidity. When calculating the average causal effect, the causal unit uses the Pearl principle and Bayesian network structure to calculate the average causal effect between cognitive abilities. When Pearl explores whether event X is the cause of event Y, it is necessary to implement event X through intervention X and calculate E(Y|do(X)). Specifically, if the average change in event Y under intervention X exceeds the significance level, then X is considered the cause of Y. Specifically, within a given dataset D or Di, the cognitive abilities to be studied are first identified. These cognitive abilities include the target cognitive ability and other cognitive abilities that influence it. For example, to investigate whether a certain condition or behavior L1 is a complication of cognitive state L2, all edges pointing to cognitive abilities L1 are cut off. The average causal effect between condition L1 and cognitive state L2 is then observed. If this change exceeds a set causal effect threshold, the condition or behavior L1 is considered a complication with cognitive state L2; otherwise, it constitutes a comorbidity.

[0119] When the causal unit is based on mining the average causal effect between cognitive abilities through data patterns, the vast amount of literature, and thus the enormous size of the Bayesian grid, leads to the use of the backdoor principle to calculate the average causal effect. The backdoor principle states that the Bayesian grid G ​​is a directed acyclic graph, (Lm, Ln) is a pair of nodes in G, and the set of nodes Z is the backdoor of (Lm, Ln), where all nodes in Z are not descendants of Z and Z blocks all paths from Lm to Ln. Therefore, the backdoor principle can be used to infer the causal relationship between cognitive abilities and Lm and Ln.

[0120] In order to simplify the undirected graph constraints without affecting the causal relationship between cognitive ability pairs, the causal unit uses an independence test. For example, the independence test can use a chi-square independence test.

[0121] In the present invention, the independence test can also be performed in the following manner:

[0122] For cognitive ability Lm, the nodes connected to Lm are obtained through a compilation based on the constructed undirected graph to form its node set. The correlation between each node and cognitive ability Lm is calculated one by one, and the node with the largest correlation is selected to make the independence assumption. The nodes that are independent of Lm under the given subset D are deleted. In this invention, entropy is used to measure the uncertainty of the random variable with respect to Lm. Given the random variable Lm, the uncertainty of the random variable Ln can be measured using conditional entropy as follows:

[0123]

[0124] The degree of correlation between random variables Ln and Lm can be measured by mutual information:

[0125]

[0126] If the mutual information exceeds the mutual information threshold, Ln is considered to be correlated with Lm. If the mutual information does not exceed the mutual information threshold, Ln is considered to be uncorrelated with Lm. Through the configuration method, the motion data generated when the user uses the chess and card entity 101 of the digital prop 1 and captured by the track sensor 102 inside the chess and card entity 101, the interaction data generated when the user interacts with the analysis unit 3 set or integrated in the mobile smart device, the language data, the eye movement data, the pupil data, and the daily data of the user obtained by the intelligent monitoring unit 4 that is always connected to the server 2 can be obtained by the analysis unit 3 and the server 2 respectively. The various types of data obtained above are stored in the analysis unit 3 and / or the server 2 set or integrated in the mobile smart device, and are used for subsequent data analysis by the server 2 and / or the analysis unit 3 through artificial intelligence algorithms to calculate digital biomarkers that can characterize the user's cognitive impairment and other diseases. For example, using causal analysis theory based on the causal knowledge base, it may be inferred that: among depression and habitual low sleep efficiency, which are both effective / important digital biomarkers for characterizing the decline in human users' cognitive abilities, the causal relationship between habitual low sleep efficiency and cognitive decline is lower than the causal relationship between depression and the decline in human users' cognitive abilities.

[0127] For another example, using causal analysis theory based on the causal knowledge base, it may be inferred that when a user uses digital prop 1, especially a chess piece with a three-dimensional spatial volume, the frequency of the chess piece's tipping movement increases significantly in a short period of time or a short period of time, which can serve as an effective digital biomarker to characterize the user's cognitive ability decline.

[0128] For another example, using the causal relationship analysis theory based on the causal relationship knowledge base, it may be inferred that: when the interval between each movement of the chess and card entity 101 by the user is significantly reduced over a period of time or a short period of time, it can be used as an effective digital biomarker to characterize that the user's cognitive ability is in an improved state. For another example, using the causal relationship analysis theory based on the causal relationship knowledge base, it may be inferred that: when the activity frequency of the body part of the user wearing the portable smart device 5 is significantly reduced over a period of time or a short period of time, and a direct causal relationship is established between it and the user's symptoms such as reduced movement, muscle rigidity, tremor, and postural adjustment disorder, which can lead to paralysis angitas (Parkinson's disease), and may have a direct causal effect on the patient's mobility, attention, orientation, and visual-spatial ability, thus serving as an effective digital biomarker to characterize that the user's cognitive ability is in a decreased state.

[0129] Through this configuration, the server 2 that is always connected to the portable smart device 5, the track sensor 2 inside the chess and card entity 101 of the digital prop 1, the analysis unit 3 set or integrated inside the mobile smart device, and the smart monitoring unit 4 distributed in the user's daily living space can also construct a causal network model based on the characteristics of related variables, and autonomously perform causal reasoning through the causal network model, and can automatically identify the relevant characteristics of the user's cognitive ability changes or other diseases, as well as the causal relationship of cognitive ability changes or other disease quality inspections and the strength of the causal relationship, and send a signal, thereby providing medical personnel or medical researchers with an effective digital biomarker or an effective way to characterize the cause of the case that can characterize the changes in the cognitive ability of human individuals.

[0130] It should be noted that the above-mentioned specific embodiments are exemplary, and those skilled in the art can come up with various solutions inspired by the disclosure of the present invention, and these solutions also belong to the disclosure scope of the present invention and fall within the protection scope of the present invention. Those skilled in the art should understand that the present invention specification and its drawings are illustrative and do not constitute a limitation of the claims. The scope of protection of the present invention is defined by the claims and their equivalents. The present invention specification contains multiple inventive concepts, such as "preferably", "according to a preferred embodiment" or "optionally", all of which indicate that the corresponding paragraph discloses an independent concept, and the applicant reserves the right to file a divisional application based on each inventive concept. Throughout the text, the features guided by "preferably" are only an optional method and should not be understood as having to be set, so the applicant reserves the right to abandon or delete the relevant preferred features at any time.

Claims

1. A digital biomarker for assessing cognitive impairment in a multi-user game scenario, characterized by: The initial data of the digital biomarker is at least derived from behavioral data of the user when using the digital prop, wherein the behavioral data includes motion data of the user moving the digital prop and touch data of the user contacting the digital prop. The motion data of the user moving the digital prop includes operations of the user lifting the digital prop, dropping the digital prop, touching the digital prop to cause the digital prop to fall, and righting the fallen digital prop. The touch data of the user contacting the digital prop includes the magnitude, direction, and frequency of the force applied by the user to the digital prop. The initial data also includes interaction data, language data, eye movement data, and pupil data collected by the analysis unit during the user's interaction with the analysis unit, as well as daily data collected by the intelligent monitoring unit during the user's daily life behavior. The pupil data includes changes in pupil size, change speed, and focus frequency when the user touches a chess or card entity during the operation of a digital prop, or hears or sees a chess or card entity played by another person; The initial data can be sent to a server and / or an analysis unit, and processed and analyzed by the server and / or analysis unit to calculate the digital biomarker capable of characterizing the user's cognitive ability. The digital prop includes: a chess and card entity, and a track sensor arranged in or integrated into the chess and card entity, wherein the track sensor is capable of at least obtaining spatial coordinate data of the chess and card entity during use by the user and / or a timestamp corresponding to the spatial coordinate data.

2. The digital biomarker according to claim 1, characterized in that The user can wear a portable smart device, and the portable smart device is configured to at least send an identification code capable of identifying the user's identity information to the analysis unit and / or the server. The analysis unit and / or the server can obtain an identification code carried by the portable smart device that can identify the user identity information.

3. The digital biomarker according to claim 2, characterized in that The portable smart device can also collect or receive input of individual characteristic data and / or physiological data of the user, wherein the physiological data includes one or more of heart rate, blood pressure, pulse oximetry, and body temperature; Individual characteristic data includes familiarity with the chess and card entity, education level and medical history, The portable smart device is capable of sending the individual characteristic data and / or physiological data to the analysis unit and / or the server.

4. The digital biomarker according to claim 2, characterized in that The behavior data collected by the track sensor disposed in the digital prop can also be sent to the analysis unit and / or the server via a portable smart device.

5. The digital biomarker according to claim 4, characterized in that: The method by which the analysis unit and / or the server processes and analyzes to calculate the digital biomarker capable of characterizing the user's cognitive ability is: extracting feature data from the initial data and further screening candidate digital biomarkers from the feature data; Performing characteristic causal inference on the candidate digital biomarkers to obtain the digital biomarkers through computational analysis.

6. The digital biomarker according to claim 5, characterized in that The method for performing characteristic causal inference on the candidate digital biomarkers to infer effective digital biomarkers is as follows: A causal analysis knowledge base is constructed based on the characteristic data, and causal inference is performed on the candidate digital biomarkers through the established causal analysis knowledge base based on causal analysis theory, so as to obtain effective digital biomarkers through computational analysis.

7. The digital biomarker according to claim 6, characterized in that The method for performing causal inference on the candidate digital biomarkers based on the causal analysis theory and the established causal analysis knowledge base is as follows: S1: Literature unit builds the original literature library; S2: Data unit constructs the dataset; S3: Causal units construct causal relationships between symptoms; S4: The knowledge unit stores the original document library, the data set and / or the causal relationship between the diseases to construct the causal analysis knowledge base that can be read and / or displayed.

8. The digital biomarker according to claim 7, characterized in that: The document unit can obtain a large number of relevant documents containing multiple cognitive abilities and classify them into several document units to construct the original document library, so that the data unit can obtain main feature parameters based on the document unit and construct a data set based on the main feature parameters.

Images