JP7285046B2 - Information processing device, information processing method and program - Google Patents

Description

The present invention relates to an information processing device, an information processing method, and a program for predicting dementia.

A device for assisting diagnosis of dementia has been proposed (see Patent Document 1).

JP 2017-217052 A

An object of the present invention is to provide an information processing device, an information processing method, and a program capable of easily and accurately predicting the onset of dementia in humans or animals.

1. Information processing device The information processing device of the present invention comprises:
Environmental data surrounding a person or animal, biometric data of the person or animal, behavior data of the person or animal, image or video data of the person or animal, and audio data of the person or animal a data acquisition unit that acquires one type;
a prediction unit that predicts the onset of behavioral/psychological symptoms in dementia of the human or animal or the onset time thereof based on the data acquired by the data acquisition unit.

In the present invention,
The prediction unit can use at least one of inference analysis, regression analysis, HotSpot analysis, proximity analysis, and spatio-temporal analysis to predict behavioral/psychological symptoms of dementia in the human or animal.

In the present invention,
including an indexing unit for indexing the data obtained by the data obtaining unit;
The prediction unit inputs the index value indexed by the indexing unit, and predicts using a function or classifier that outputs a predicted value related to behavioral and psychological symptoms of dementia in the human or animal. be able to.

In the present invention, the classifier may be SVM (Support Vector Machine), neural network or linear regression model.

In the present invention, the difference between the prediction of the human or animal behavioral/psychological symptoms in dementia of the prediction unit and the actual behavioral/psychological symptoms of the human or animal is evaluated, and based on the evaluation result, the function or A training unit for training the classifier can be included.

In the present invention, the learning unit evaluates the difference between the prediction of behavioral/psychological symptoms in dementia of the person or animal predicted by the prediction unit and the actual behavioral/psychological symptoms of the person or animal,
Information about the difference between the prediction of the behavioral/psychological symptoms of dementia of the person or animal by the prediction unit and the actual behavioral/psychological symptoms of the person or animal, and information about the data acquired by the data acquisition unit. In the context of , the function or classifier can be trained by deep learning.

In the present invention, the function includes parameters related to numerical values related to the data acquired by the data acquisition unit and coefficients,
The learning unit can adjust the coefficients by deep learning.

In the present invention,
The data acquisition unit acquires the environmental data and the biometric data,
The environmental data may include ambient temperature and humidity of the person or animal, and the biometric data may include pulse, respiration rate and body temperature of the person or animal.

In the present invention, the data storage unit stores the environmental data, the biological data, and information related to behavioral/psychological symptoms of dementia of the human or animal corresponding to the environmental data and the biological data in association with each other. be able to.

In the present invention, a correlation table is stored that is configured by associating the environmental data and the biological data with information on behavioral and psychological symptoms of dementia of the human or animal corresponding to the environmental data and the biological data. can include a storage unit that

The present invention may include a coping method derivation unit that generates or selects a coping method based on the predicted content of the behavioral/psychological symptoms of dementia of the person or animal predicted by the prediction unit.

In the present invention, the actual behavioral/psychological symptom of the person or animal is determined by the behavioral/psychological symptom determination unit according to a determination algorithm,
The learning unit evaluates the difference between the judgment result by the behavioral/psychological symptom judging unit and the judgment result of the behavioral/psychological symptom of the human or animal actually judged by a person, and the judgment algorithm based on the evaluation result. can be learned.

2. Information processing method The information processing method of the present invention comprises:
a group of environmental data around a person or animal, biometric data of the person or animal, action data of the person or animal, image or video data of the person or animal, and audio data of the person or animal a step of obtaining at least one selected from
and a step of predicting the onset of behavioral/psychological symptoms of dementia in the human or animal or the time of onset thereof, based on the data acquired by the data acquisition unit.

3. Program The program of the present invention is
A data acquisition unit in a computer is provided with environmental data around a person or animal, biometric data of the person or animal, behavior data of the person or animal, image or video data of the person or animal, and voice data of the person or animal. obtaining at least one selected from the group of
and a step of predicting the onset of behavioral/psychological symptoms in dementia of the human or animal or the time of onset thereof, based on the data acquired by the data acquisition unit.

By predicting the onset of BPSD or the time of onset, it is possible to prevent the onset of BPSD and greatly reduce the occurrence of BPSD itself by early treatment. Since BPSD can be predicted in advance, the burden on caregivers can be reduced.

It is a figure which shows the whole structure of the flow of information processing. It is a figure which shows the structural example of an information processing apparatus. It is a figure which shows the structural example of an information processing system. It is a figure explaining the processing flow of information processing. It is a figure which shows the data example of five senses data. It is a figure which shows the data example of environmental data. FIG. 4 is a diagram showing an example of biometric data; It is a figure which shows the data example of a care record. It is a figure which shows a neural network typically. FIG. 4 is a diagram showing the correlation between behavior and BPSD; It is the figure which arranged the output content of each step in data processing. It is a figure which shows the correspondence table|surface of BPSD and a coping method. FIG. 3 is a flow schematic diagram for BPSD decision and decision algorithm learning;

Preferred embodiments of the present invention will be described below with reference to the drawings.

An embodiment will be described with an example of predicting behavioral/psychological symptoms in human dementia. Below, the behavioral/psychological symptoms in human dementia are simply referred to as “BPSD”.
1. Outline of Information Processing Apparatus (1) Information Processing Apparatus As shown in FIG. Function F3 for predicting the possibility of onset of behavioral and psychological symptoms of dementia, the content and timing of onset, function F4 for outputting prediction results regarding BPSD, and generation and output of predicted BPSD coping methods. It includes a function F5, a function F6 for determining whether or not BPSD has developed, a function F7 for verifying symptoms after confirmation and execution, and a function F7 for learning a prediction algorithm based on the verification results.

(2) BPSD (behavioral and psychological symptoms of dementia)
Behavioral symptoms are symptoms that are revealed by observation. Evaluation items for behavioral symptoms include, for example, wandering, dangerous manipulation, concealment, pseudo-sexual work, trouble, collecting, and shouting/excitement. Psychological symptoms are symptoms that are revealed by an interview. Evaluation items for psychological symptoms include, for example, depression and delusions. Behavioral and psychological symptoms may or may not appear depending on the person.

2. Configuration of Information Processing Apparatus (1) Functional Configuration of Information Processing Apparatus As shown in FIG. there is As devices, for example, various IoT sensors can be applied. A urine volume sensor 60 g and the like can be mentioned.

The information processing apparatus 10 includes a data acquisition unit 40, a processing unit 30, a data acquisition unit 40, and a storage unit 20, as shown in FIG.

The data acquisition unit 40 acquires at least one selected from the group of environmental data around a person, biometric data of a person, action data of a person, image or video data of a person, and voice data of a person. The data acquisition unit 40, for example, acquires environmental data and biometric data, the environmental data including ambient temperature and humidity of the person, and the biometric data including the pulse, respiration rate and body temperature of the person.

The processing unit 30 includes an indexing unit 30a, a prediction unit 30b, a learning unit 30c, a coping method derivation unit 30d, and a BPSD determination unit 30e.

The indexing unit 30a is for indexing the data acquired by the data acquiring unit 40 .

The prediction unit 30b predicts the onset of BPSD or its onset time based on the data acquired by the data acquisition unit 40 . The prediction unit 30b can predict the BPSD using at least one of inference analysis, regression analysis, HotSpot analysis, proximity analysis, and spatio-temporal analysis.

The prediction unit 30b can receive the index value indexed by the indexing unit 30a and perform prediction using at least one of the function 20a and the classifier 20b that output a prediction value for BPSD.

The classifier 20b can be an SVM (Support Vector Machine), a neural network or a linear regression model. The function 20a can be configured to include parameters related to numerical values related to the data acquired by the data acquisition unit 40, and coefficients. The neural network can be, for example, a three-layer feedforward neural network consisting of an input layer, an intermediate layer and an output layer, as shown in FIG.

The learning unit 30c evaluates the difference between the BPSD prediction of the prediction unit 30b and the actual human behavioral/psychological symptoms, and learns at least one of the function 20a and the classifier 20b based on the evaluation result. The learning unit 30c can feed a machine learning algorithm to train the function 20a or the classifier 20b. Specifically, the learning unit 30c evaluates the difference between the BPSD prediction predicted by the prediction unit 30b and the actual human behavior/psychological symptoms, and compares the BPSD prediction of the prediction unit 30b with the actual human behavior. - The function 20a or the classifier 20b can be learned by deep learning in relation to the evaluation content of the difference from the psychological symptoms and the information on the data acquired by the data acquisition unit 40 . The learning unit 30c may adjust the coefficients by deep learning.

The coping method deriving unit 30d derives a coping method based on the predicted possibility of onset of BPSD, the details of the onset, and the onset time.

The BPSD determination unit 30e determines whether or not it is BPSD based on the input data.

In realizing each function of the processing unit 30, it is possible to extract the feature amount of the input data and perform each function processing.

The storage unit 20 can store, for example, a function 20a, a classifier 20b, a correlation table 20c, a correspondence table 20d, and a data storage unit 20e.

The correlation table 20c can consist of a database configured by associating environmental data, biometric data, and information on BPSD corresponding to the environmental data and biometric data. The correspondence table 20d can be made up of a database configured by associating the presence or absence of predicted occurrence of BPSD, the time of occurrence, and coping methods. The data storage unit 20e stores data obtained by the data obtaining unit 40, data obtained by processing the data by the processing unit 30, data input by the input unit 52, and reference for various judgments by the processing unit 30 in advance. data to be processed, etc. are stored.

The data acquisition unit 40 may acquire processed data, or may acquire raw data (measurement data or input data) and process it in the processing unit 30 . Examples of data acquired by the data acquisition unit 40 include the following.

(a) Five Senses Data The five senses data detect, for example, sound source detection, sound type identification, olfactory perception, face recognition, motion detection, side distance function, stereoscopic detection, and the like, as shown in FIG. Five senses data can be measured with cameras, microphones, etc. Note that the device is not particularly limited as long as it can obtain these data.

(b) Environmental Data Environmental data, for example, as shown in FIG. 6, measures temperature, humidity, illuminance, amount of water, and air pressure, and detects movement distance. Environmental data can be measured from a temperature sensor, a humidity sensor, an illuminance sensor, an atmospheric pressure sensor, and the like. Note that the device is not particularly limited as long as it can obtain these data.

(c) Biological Data As shown in FIG. 7, biological data includes, for example, heartbeat, respiration, getting out of bed, sleep (detection of non-REM sleep and REM sleep), wakefulness, excretion, amount of exercise, body temperature, and the like. A sensor for detecting biometric data is not particularly limited as long as it can perform these detections, and can be composed of one or a plurality of sensors. Taking sleep as an example, a Doppler sensor can be used to ascertain a non-REM sleep state and a REM sleep state. Regarding excretion, the timing of excretion can be detected by combining a toilet door open/close sensor and a human sensor in the toilet, or by analyzing tags with communication functions (BLE tags) and camera images. can be done. Regarding excretion, the content of excretion can be detected in consideration of the timing of nocturnal urination and the time required for excretion. The amount of exercise can be calculated using software that calculates calorie consumption, calculated from the distance traveled by a tag with a communication function (BLE tag), or based on caregiver records such as exercise, cleaning, washing, etc. You can calculate the amount of exercise and calorie consumption by calculating the time you moved the . Note that the device is not particularly limited as long as it can obtain these data.

(d) Nursing care record The items of the nursing care record are, for example, as shown in FIG. Examples include what the caregiver observes, the care recipient's condition assessment, the care recipient's or caregiver's impressions, the caregiver's response, the caregiver's voice, and what the caregiver did.

In the care record, it is preferable to have the following records in order to improve the prediction accuracy of BPSD.
1) The occurrence of BPSD 2) The time (time zone) and location (role of place, location relationship, etc.) of occurrence of BPSD must be clearly recorded 3) Environmental factors at the time of occurrence of BPSD (Temperature, humidity, illuminance, air pressure, noise, odor, presence of people, content of conversation, subject's behavior and attitude) 4) Subject's behavior and attitude before and after the occurrence of BPSD be recorded. In addition, the same environmental factors shall be recorded.

Combining these, the role of places in the occurrence of BPSD, the distance between places, the time zone in the life rhythm, the environmental factors for place and time, the behavior before and after, the relationship with people before and after, etc. From this point of view, it is possible to make a highly accurate analogy of a situation in which the possibility of occurrence of BPSD increases.

(e) Data Acquisition Method The data acquired by the acquisition unit may be acquired, for example, as follows.
1) Periodic acquisition of environmental data (temperature, humidity, air pressure, illuminance) 2) Periodic acquisition of biometric data (heart rate, breathing rate, sleep state, bed leaving state) 3) Target user's position by BLE tag Acquire information and calculate the distance traveled 4) In conjunction with the AI camera, obtain the location information of the target user and calculate the distance traveled 5) Link the above obtained data with the user ID and send it to the information processing device Cycle sent to

In the above 1) periodic acquisition of temperature and humidity, the temperature/humidity sensor can transmit acquired data in a period of 10 minutes from real time by default. Each gateway can acquire target sensor data in association with a user ID. Analysis is possible by acquiring multiple sensor data with one gateway and supporting multiple users at the same time.

In the above 2) periodical acquisition of atmospheric pressure and illuminance, an expansion module equipped with an atmospheric pressure sensor and illuminance sensor can be connected to the IoT gateway to update the atmospheric pressure and illuminance information in 10-minute cycles from real time. A single gateway can support multiple users' data and can handle each analysis.

In the above 3) periodic acquisition of heart rate, respiratory rate, sleep state, and bed leaving state, the Doppler sensor unit was installed on the wall around the bed, and heart rate, breathing rate, sleep state and bed leaving / landing event notification can be performed. Heart rate, respiration rate, and sleep state are measured every 10 minutes from real time, and analysis is possible by notifying the IoT gateway. The leaving/landing state may be immediately notified when an event occurs.

(2) Physical Configuration of Information Processing Apparatus The information processing apparatus 10 is composed of a computer, for example, a large computer such as a supercomputer, a computer equipped with a GPU (Graphics Processing Unit), or a plurality of personal computers. , quantum computers, etc.

The processing unit 30 can be realized by an arithmetic device such as a CPU. The storage unit 20 can be stored in a known storage device such as a ROM, a hard disk, or an external storage device (CD, DVD, etc.), for example. The storage unit 20 may be configured separately from the computing device, or may be configured within the computing device. The information processing apparatus 10 may be composed of one computer or may be composed of a plurality of computers.

A program that causes information processing apparatus 10 to execute information processing can be stored in a storage device (eg, ROM, hard disk) or the like included in information processing apparatus 10 .

The data acquisition unit 40 can be composed of, for example, a reception unit capable of receiving data from a communication network.

The input unit can consist of, for example, a known input device such as a keyboard, mouse, or touch panel. A known display such as a liquid crystal display or an organic EL display can be applied to the display unit 54 . The transmission unit 56 transmits information to terminals connected via a communication network, and a known transmission device can be applied.

3. Information Processing Method (1) Overview of Information Processing Method An information processing method will be described with reference to FIG.

The data acquisition unit 40 acquires five senses data, environmental data, biometric data, behavior data, and care records of the person receiving care from an IoT device or the like (S1).

The data acquired by the indexing unit 30a is indexed to calculate evaluation index parameters (S2).

The prediction unit 30b analyzes the evaluation index parameters and predicts the BPSD (S3).

A coping method is generated based on the presence/absence of the occurrence of BPSD predicted by the coping method derivation unit 30d and the time of occurrence (S4).

The data acquisition unit 40 acquires behavior data and the like, and the BPSD determination unit 30e evaluates whether or not a BPSD symptom has actually occurred (S5).

Based on the difference between the BPSD predicted by the learning unit 30c and the actual symptom, learning (updating) of the function 20a or classifier (neural network) is performed based on deep learning (S6).

FIG. 11 shows measurement data and processing data obtained at each stage. Each data of FIG. 11 can be stored in the data storage section 20e of the storage section 20 in association with each other.

(2) Example of indexing process (a) Environmental index As for the environmental index, the degree of discomfort calculated for each temperature + humidity, atmospheric pressure, and illuminance can be used as an index. First, a graph indicating the well-known comfortable range of temperature + humidity is used, and whether it is in the comfortable range or not is judged whether it is uncomfortable or not.

As a process for this, first, it is determined whether the humidity is between 40% and 70%, and if it is outside the range, it is considered uncomfortable.

この数式の範囲に気温と湿度が存在する場合には快適とし、この範囲を逸脱する場合には不快であるとする。不快であると判断した場合の不快度は簡易的に快適範囲の中心点から現在の気温湿度の直線と快適範囲の交点からの距離を不快度の指標として用いる。Also, if the humidity is between 40% and 70%, the following formula is used to determine whether the temperature is appropriate.

If the temperature and humidity are within the range of this formula, it is considered comfortable, and if it deviates from this range, it is considered uncomfortable. The degree of discomfort when it is determined to be uncomfortable is simply obtained by using the distance from the center point of the comfort range to the intersection of the current temperature/humidity line and the comfort range as an indicator of the degree of discomfort.

ここで気圧をPとしたときに不快指標はP’となる。Next, for the atmospheric pressure discomfort index, the following formula is applied using the standard atmospheric pressure of 1013 hPa as a standard value.

Here, when the air pressure is P, the discomfort index is P'.

照度Iとして、照度の不快指標であるＩ’を算出する。また、基準照度は活動時２００ｌｕｘ、安静時５０ｌｕｘ、睡眠時２０ｌｕｘとして計算する。Next, regarding the illuminance discomfort index, the reference illuminance differs depending on the user's condition. Using the reference illuminance as _I0 , the following formula is used.

As the illuminance I, an illuminance discomfort index I' is calculated. Also, the reference illuminance is calculated as 200 lux during activity, 50 lux during rest, and 20 lux during sleep.

(b) Bioindex A bioindex P' is obtained by the following formula, where P is the heart rate per minute and _P0 is the average heart rate.

As for the respiratory rate, B is the respiratory rate per minute, and _B0 is the average respiratory rate.

ここでｐ１～ｐ５は各パラメータ、α～ε’は認知症の行動・心理症状（ＢＰＳＤ）に発生時に関連する項目の度合いを表現する。(3) Prediction Processing (a) Prediction Logic When P is the prediction logic, P can be calculated by performing matrix calculation based on the following formula.

Here, p1 to p5 represent the respective parameters, and α to ε′ represent the degrees of items related to behavioral/psychological symptoms of dementia (BPSD) at the time of occurrence.

(b) Prediction of Occurrence Time of BPSD For prediction of the occurrence time of BPSD, it is possible to derive the relevance of time zones based on several reference points. As an indicator, the following times are listed as candidates for reference points.
1) 0:00 (date change point)
2) sunrise (sunset)
3) Wake up 4) Meal

For the occurrence of BPSD, the elapsed time from each reference point can be calculated, and the correlation can be derived from the bias. Regression analysis can be performed on the timing at which BPSD occurs and the elapsed time from the reference point to derive the relationship. At this time, waking up may be classified according to nighttime sleep or nap, and meals may be classified according to types such as breakfast, lunch, dinner, and snacks. When the time when BPSD occurs is defined as point A, 0 o'clock, the time of waking up from nighttime sleep, the elapsed time of breakfast, lunch, and dinner, and the elapsed time from the nap if it is after a nap, or after a snack. The elapsed time from the snack may be additionally used as an indicator.

When a new event occurs, people tend to be more likely to have memory deficits in which they forget events that happened shortly before. When depressed, the person is lethargic and does not know what to do, he/she tends to repeat the same behavior in search of something, and to wander aimlessly. Due to excessive involvement by instructions, there is a tendency to make annoying and annoying kanji, and to suddenly raise a loud voice and get angry. These flows have characteristic time lags, and predictions can be made considering the time series.

In predicting the BPSD, it is preferable to use the function 20a or the classifier 20b that considers the influence of time series, unlike image analysis of photographs. Learning of this function 20a or classifier 20b is similar to language/speech/video analysis among deep learning, so it is preferable to learn using LSTM (Longshort-term memory).

(c) Prediction of BPSD from environmental factors Since the surrounding environment has a large impact on the occurrence of BPSD, it is possible to analyze environmental factors, especially temperature, humidity, air pressure, illuminance, odor, and noise when BPSD occurs. can. Regarding temperature and humidity, the degree of comfort can be diagnosed based on the comfort index graph for the relationship between temperature and humidity, and can be used as an index. For air pressure, illuminance, odor, and noise, it is possible to calculate the influence and degree of each index on BPSD by analyzing the relationship with the occurrence of BPSD as a single index, and performing regression analysis for each.

(d) Prediction from Natural Language Analysis BPSD can also be analyzed by natural language analysis. After identifying the part of speech included in the question or inquiry by natural language processing, a hypothesis is generated, and then the hypothesis is supported or evidence is sought. A treatment method is assigned based on BPSD onset status according to a statistical model approach of evidence-weighted scores.

(e) Prediction from Behavior As shown in FIG. 10, the correlation of BPSD can be found from the behavior of the dementia patient. This makes it possible to predict the onset of BPSD from the behavior of dementia patients.

(4) Coping method Based on the symptoms of BPSD, a coping method can be generated based on the function 20a or the classifier 20b. Alternatively, a BPSD coping method may be selected based on a correspondence table 20d between BPSD symptoms and coping methods. A correspondence table 20d as shown in FIG. 12 may be stored in the storage unit 20. FIG.

In addition, in the coping method, when it is necessary to prepare the contents of the meal menu, the nutrition calories may be calculated based on the amount of exercise and the calorie consumption, and may be used for the menu.

(5) Verification (a) Evaluation of Occurrence Verification of the onset of BPSD can be performed using a BPSD troublesome behavior scale (TBS: Troublesome Behavior Scale). TBS describes disruptive behaviors and burdens in elderly people with dementia.

TBS defines 15 items describing the disruptive behavior and burdensome behavior of elderly dementia patients and their frequency, and the applicant has reliability and validity as a scale for evaluating the behavioral transfer of dementia patients. Confirmed that there is. In addition, the problem behaviors that are relatively often observed in people with dementia are evaluated (e.g., 5-grade evaluation) based on the frequency observed by caregivers during a predetermined period of time in the past (e.g., the past month), and the prediction results are verified. can do. The frequency can be rated as "once a day or more", "several times a week", "several times a month", and "never".

BPSD analysis can be performed in speech pathology analysis. Using emotion recognition technology (Sensibility Technology), it is possible to analyze the ratio of the four emotions of ``angry'', ``joy'', ``sadness'', and ``normal'' and the degree of ``excitement'' during speech, and display the degree. Verification of onset of BPSD may be performed by imaging diagnosis of the brain.

As shown in FIG. 13, in the evaluation of whether BPSD has developed, and if it has developed, the content of the onset and the time of onset, the data acquired by the data acquisition unit 40 or whether the dementia patient has developed BPSD or Based on the data input by the input unit 52, the BPSD determination unit 30e can determine. This BPSD judgment algorithm can be learned by a learning unit. The method of learning the BPSD determination algorithm can be performed in the same manner as the learning of the prediction algorithm of the prediction unit 30d.

Specifically, the data acquisition unit 40 acquires data (F11), the indexing unit 30a indexes the data, the BPSD determination unit 30e determines whether it is BPSD (F13), and the determination result of whether it is BPSD to output The status of whether the input is actual BPSD and the result determined by the BPSD determination unit 30e are verified (F15), and the learning unit 30c learns the determination algorithm used when the BPSD determination unit 30e determines whether it is BPSD. (F16).
(b) Evaluation of Coping Method Through the evaluation of BPSD coping, the derivation of the care method evaluates its reliability according to how much evaluation it has obtained. By running analytics on the large amounts of data collected from care settings, insights can be gathered and converted into inspiration so that appropriate care methods can be derived. In the evaluation of the coping method for BPSD, based on the data acquired by the data acquisition unit 40 or the data entered by the input unit 52 whether the dementia patient developed BPSD, the processing unit 30 verifies whether the coping method is appropriate. may Based on this verification result, the learning unit 30d can learn the derivation algorithm of the coping method derivation unit 30d and update the derivation algorithm. The method of learning the derivation algorithm of the coping method derivation unit 30d can be performed in the same manner as the learning of the prediction algorithm of the prediction unit 30d.

3. Effect By predicting the onset of BPSD or the time of onset, it is possible to prevent the onset of BPSD and greatly reduce the occurrence of BPSD itself by early treatment. Since BPSD can be predicted in advance, the burden on caregivers can be reduced.

Various modifications can be made to the above embodiment within the scope of the present invention. In the above embodiments, the behavioral/psychological symptoms of dementia in humans have been described, but the present invention is not limited to this and can be widely applied to the behavioral/psychological symptoms of dementia in animals.

4. Application Example (1) AI Artificial Intelligence System The information processing apparatus according to the present embodiment can be applied to the following AI (Artificial Intelligence) artificial intelligence system. In other words, the AI artificial intelligence system consists of an information processing device that consists of a data storage unit, analysis unit, sensitivity processing unit, and planning unit based on the knowledge expression method, and data collected from the IoT gateway through automatic identification, automatic handling, and automatic notification. It can also consist of a data mining unit that integrates nursing care record data collected by human interface, natural language analysis, and voice recognition, and further statistical analysis.

Based on knowledge data that can be used for care, it is possible to continuously grasp the condition of the elderly with dementia, predict the onset of BPSD specific to the change in condition, and suggest an appropriate response method stored in advance. . It has the following effects:
(a) focus on people with dementia as well as caregivers; (b) be multidimensional and flexible, tailored to the needs of caregivers and people with dementia; In (d) information sharing, it is possible to correlate and analyze data obtained from IoT information while paying attention to the following information (a) to (c).
1) Specific information on coping with BPSD 2) Information on ensuring the physical safety and well-being of people with dementia 3) Information on dealing with difficult ADL (activities of daily living) information

Fundamental research on AI can be applied and utilized to deal with behavioral and psychological symptoms of dementia (BPSD) based on reasoning and learning.

(2) Functions of the AI artificial intelligence system The AI artificial intelligence system can have the following functions.
(a) Expert system A system that accumulates expert knowledge as rules and solves problems using inference techniques.
(b) Speech recognition Speak to a smartphone or tablet. The computer identifies who is speaking and what is said is understood and documented.
(c) Natural language processing The information is sorted and recorded so that the computer can understand the meaning and content of the written information so that it can be searched for according to the F-SOAIP Life Support Recording Method.
(d) Sensitivity processing Based on the knowledge of cognitive science and ergonomics, sensations such as warmth or coldness are received from environmental sensors and implemented on a computer.
(e) Image Recognition A computer is made to understand the content captured by a camera or the like, and the brightness and color tone of the living room can be sorted into comfort and discomfort.
(f) machine learning
It is a system that finds consistent rules (patterns) from data collected from IoT sensors and nursing care records. It is strongly related to the statistical branch of mathematics and is analyzed and organized using, for example, the following statistical techniques.

1) FTA (Fault Tree Analysis) Analysis This is an analysis method that seeks the cause from the results, and analyzes the occurrence route, cause of occurrence, and occurrence probability of an event that should not occur. In order to analyze the frequency of occurrence of BPSD, logically trace the potential danger (fault) of the cause ("fault" here refers to events such as environmental and human errors), Add the probabilities of occurrence to calculate the probability that the basic event will occur.
2) ETA (Event tree analysis) Analysis This is also called danger prediction analysis, and analyzes the process leading up to the occurrence of BPSD based on the probability of occurrence and the success or failure of responses to countermeasures (care).
3) HAZOP (Hazard and Operability Study) analysis A method of analysis based on empirical rules. It is a method of analyzing whether it was born as a result of the influence of not doing.
The parameters here are data obtained from the IoT gateway such as body temperature, respiration, pulse, temperature, humidity, air pressure, sleep time, etc., and guide words such as whether the event occurs repeatedly.
4) Information retrieval This is a system that finds what is necessary for dementia care from the accumulated data.

(g) Inference This is a system for deriving a consistent answer by integrating a certain rule from various patterns.
1) Hot spot analysis
This is an analysis method in which the space (place) where behavioral/psychological symptoms have occurred in the past is regarded as the space (place) where the behavioral/psychological symptoms are highly likely to occur.
2) Regression Analysis
In addition to past BPSD, other variables related to BPSD such as environment and human relations are used as independent variables to predict future BPSD through regression analysis.
3) Near-Repeat Methods
Predict future BPSD based on the spatio-temporal proximity of one BPSD to the next BPSD.
4) Spatiotemporal Analysis
Behavioral and psychological symptoms are predicted from the movement patterns that accompany changes in the occurrence of BPSD over time and various factors that affect them.
5) Risk Terrain Analysis
Risk planes are created from proximity to spatial factors that affect BPSD to predict future BPSD occurrences.

(h) Data mining A system that combines database technology and machine learning to find potentially useful information from large amounts of unorganized data. to derive
(i) Human interface Smartphones and tablet devices that enable caregivers to operate devices such as computers more easily can be applied.
(j) Planning In deriving appropriate care responses for BPSD, a system can be applied to determine the order in which care should be provided.
(k) Multi-agent This system uses F-SOAIP to re-investigate the occurrence of BPSD and propose care methods when caregivers who solve BPSD gather and solve complex problems at the nursing care site.

By utilizing the above contents, it is possible to apply and utilize inference and learning as a basis for dealing with BPSD in dementia.

(3) Care methods Dementia requires care methods tailored to different behavioral and psychological symptoms. This is because different care methods have different effects on people with dementia. Information processing equipment can store various literature and historical information covering information on standard care methods and best personal assistance methods. All of this allows us to identify which care modalities offer caregivers the best options to adopt when caring for people with dementia. With the guidance of care professionals, the dementia-adaptive IoT service may collect the knowledge necessary to acquire competence in the care setting. We define this as a ``corpus of knowledge in dementia care''. Corpus creation can begin by loading a number of relevant documents into the dementia-enabled AI.

The corpus creation can also involve expert staff to select information or eliminate all information that is outdated, considered inferior, or irrelevant to the subject area. This is defined as “curation of dementia-friendly content”.

Indexes and other metadata that are pre-processed by curation and allow content to be linked more efficiently can be built from the field with the Assisted Living Record Method (F-SOAIP). A dementia-aware AI trained on question-answer pairs can continue to learn by continuously interacting with the bot. In addition, as new information is released, the dementia-aware AI will also be updated to constantly adapt to changes in knowledge and linguistic interpretations in a given area, identifying new insights and patterns hidden in the information. can be prepared to It is a method of identifying the part of speech included in a question or inquiry by natural language processing, generating a hypothesis, and then looking for supporting or evidence for the hypothesis.

Coping methods can be assigned based on the onset of behavioral and psychological symptoms according to a statistical model approach of evidence-weighted scores. Dementia-responsive AI can evaluate its reliability based on how much evaluation the derivation of care methods has obtained through successful examples of behavioral and psychological symptom management. In short, dementia-aware AI performs analytics on the large amounts of data collected from care settings, collects insights and transforms them into inspiration so that appropriate care methods can always be derived.

Various modifications can be made to the above embodiment within the scope of the present invention.

INDUSTRIAL APPLICABILITY The present invention is applicable as a management system in caring for dementia patients and dementia-affected animals.

10 Information processing device 20 Storage unit 20a Function 20b Classifier 20c Correlation table 20d Correspondence table 20e Data storage unit 30 Processing unit 30a Indexing unit 30b Prediction unit 30c Learning unit 30d Coping method generation unit 30e BPSD determination unit 40 Data acquisition unit 52 Input Unit 54 Display unit 56 Transmission unit 70 Communication network

Claims (13)

a data acquisition unit that acquires at least environmental data about a person or animal ;
A prediction unit that predicts the onset of behavioral and psychological symptoms in dementia of the human or animal or the time of onset thereof based on the data acquired by the data acquisition unit;
an indexing unit for indexing the environmental data acquired by the data acquisition unit ;
the environmental data includes ambient temperature, humidity and illuminance of the person or animal;
The indexing unit includes a function of calculating a first degree of discomfort from temperature in relation to humidity and a function of calculating a second degree of discomfort from illuminance,
The prediction unit inputs index values including the first degree of discomfort and the second degree of discomfort indexed by the indexing unit, and predicts behavioral and psychological symptoms of dementia in the human or animal. An information processor in which predictions are made using functions or classifiers that output values . In claim 1,
The prediction unit uses at least one of inference analysis, regression analysis, HotSpot analysis, proximity analysis, and spatio-temporal analysis to predict behavioral/psychological symptoms of dementia in the human or animal. In claim 1,
The information processing device, wherein the classifier is an SVM (Support Vector Machine), a neural network, or a linear regression model. In claim 1,
Evaluating the difference between the prediction of the behavioral/psychological symptoms of dementia of the person or animal by the prediction unit and the actual behavioral/psychological symptoms of the person or animal, and learning the function or classifier based on the evaluation result Information processing apparatus including a learning unit that performs In claim 4,
The learning unit evaluates the difference between the prediction of behavioral/psychological symptoms in dementia of the person or animal predicted by the prediction unit and the actual behavioral/psychological symptoms of the person or animal,
Information about the difference between the prediction of the behavioral/psychological symptoms of dementia of the person or animal by the prediction unit and the actual behavioral/psychological symptoms of the person or animal, and information about the data acquired by the data acquisition unit. Information processing apparatus for learning the function or classifier by deep learning in the context of In claim 4,
The function includes parameters related to numerical values related to the data acquired by the data acquisition unit and coefficients,
The learning unit is an information processing device that adjusts the coefficient by deep learning. In claim 1,
The data acquisition unit acquires the biometric data,
The information processing device, wherein the biometric data includes the pulse, respiration rate and body temperature of the person or animal. In claim 7,
An information processing apparatus comprising a data storage unit in which the environmental data, the biological data, and information related to the behavioral/psychological symptoms of dementia of the human or animal corresponding to the environmental data and the biological data are stored in association with each other. In claim 7,
a storage unit storing a correlation table configured by associating the environmental data and the biological data with information on behavioral and psychological symptoms of dementia of the human or animal corresponding to the environmental data and the biological data; information processing equipment including; In claim 1,
An information processing apparatus including a coping method derivation unit that generates or selects a coping method based on the prediction content of the behavioral/psychological symptoms of dementia of the person or animal predicted by the prediction unit. In claim 4,
the actual behavioral/psychological symptom of the person or animal is determined by a behavioral/psychological symptom determination unit according to a determination algorithm;
The learning unit evaluates the difference between the judgment result by the behavioral/psychological symptom judging unit and the judgment result of the behavioral/psychological symptom of the human or animal actually judged by a person, and the judgment algorithm based on the evaluation result. Information processing device for learning. a data acquisition unit acquiring at least environmental data around a person or animal ;
a step of predicting the onset of behavioral/psychological symptoms in dementia of the human or animal or the time of onset thereof, based on the data acquired by the data acquisition unit;
an indexing step in which the indexing unit indexes the environmental data acquired by the data acquisition unit;
the environmental data includes ambient temperature, humidity and illuminance of the person or animal;
The indexing step includes a step of calculating a first degree of discomfort from temperature in relation to humidity, and a step of calculating a second degree of discomfort from illuminance,
The prediction unit inputs index values including the first degree of discomfort and the second degree of discomfort indexed by the indexing unit, and predicts behavioral and psychological symptoms of dementia in the human or animal. A method of processing information , including the step in which a prediction is made using a function or classifier that outputs a value . a data acquisition unit in a computer acquiring at least environmental data around a person or animal ;
A step in which the prediction unit predicts the onset of behavioral/psychological symptoms in dementia of the human or animal or the onset time thereof based on the data acquired by the data acquisition unit;
A program for causing an indexing unit to perform an indexing step of indexing the environmental data acquired by the data acquisition unit,
the environmental data includes ambient temperature, humidity and illuminance of the person or animal;
The indexing step includes a step of calculating a first degree of discomfort from temperature in relation to humidity, and a step of calculating a second degree of discomfort from illuminance,
The prediction unit inputs index values including the first degree of discomfort and the second degree of discomfort indexed by the indexing unit, and predicts behavioral and psychological symptoms of dementia in the human or animal. A program containing steps in which predictions are made using a function or classifier that outputs a value .

Images