CN111419249B - Depression prediction model generation method and prediction system - Google Patents
Depression prediction model generation method and prediction system Download PDFInfo
- Publication number
- CN111419249B CN111419249B CN202010220741.6A CN202010220741A CN111419249B CN 111419249 B CN111419249 B CN 111419249B CN 202010220741 A CN202010220741 A CN 202010220741A CN 111419249 B CN111419249 B CN 111419249B
- Authority
- CN
- China
- Prior art keywords
- depression
- data
- prediction model
- user
- time sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
- A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The application provides a generation method and a prediction system of a depression prediction model, wherein the generation method of the model comprises the following steps: step 1: collecting heart rate wave band time sequence data and depression measurement values of a user; step 2: extracting time sequence characteristic data of a user based on the heart rate wave band time sequence data, and constructing a training sample and a test sample; step 3: generating a depression prediction model based on the time sequence characteristic data of the user and the depression measurement value of the user; step 4: and generating an optimal depression prediction model. According to the invention, through heart rate data acquired by the wearable equipment, the automatic identification of the depression state of the user is realized by using a machine learning algorithm, the intrusion on the user is less, the natural behavior of the user is recorded conveniently, and the technical support is provided for the longitudinal timely tracking and early warning of the depression state of the user.
Description
Technical Field
The invention relates to the field of psychology and artificial intelligence, in particular to a method and a system for generating a depression prediction model.
Background
Currently, society is increasingly competitive, almost everyone is running in overload, and depression moods with different degrees are easily generated, which is a common emotional component. When people suffer from mental stress, life frustration, pain and death, natural disasters and accidents, the depressed emotion can be generated. Almost all people feel low mood at some point, often because of something that is not desirable in life. But persistent depression, major depression, is another concern. Worldwide, the number of people affected by some form of depression is 25% of all women, 10% of all men, and 5% of all teenagers. This is the most common psychological problem in the united states, and is therefore annoying for approximately seven hundred and sixty thousand people each year.
Depression, also known as depressive disorder, is a major type of mood disorder with marked and persistent depression as a major clinical feature. Clinically, the mood is dissatisfied with the mood, the mood can be subsided from smoldering to sad and absolute, the user can feel depressed, even pessimistic and aversive, and suicide attempts or behaviors can be realized; even wood stiffness occurs; some cases have significant anxiety and motor agitation; in severe cases, psychotic symptoms such as hallucinations and delusions may occur. Each episode lasts at least 2 weeks, longer or even years, most cases have a tendency to recur, most of each episode can be alleviated, some can have residual symptoms or be converted to chronic.
Depression is the fourth disease in the world, and the prevalence rate of depression in China is reported to be 3% -5%, but the prevalence rate is reported to be 6.1%. At present, the medical treatment and prevention of the depression in China is still in the situation of low recognition rate, the recognition rate of hospitals above the ground level market is less than 20%, and only less than 10% of patients receive relevant drug treatment; moreover, the onset of depression (and suicidal events) has begun to appear as a trend to low age (university, or even primary and secondary school student population). Therefore, attention is paid to the science popularization, prevention and treatment of depression, and the prevention and treatment of depression are put into the key point of mental health work nationwide.
The recurrence risk of the depression is large, and people trace the depression patients for 10 years to find that 75-80% of the patients relapse repeatedly, so that the depression patients need preventive treatment. More than 3 times of attack should be treated for a long time, even if the medicine is taken for life. The dose of the maintenance therapeutic agent should be the same as the therapeutic dose as well as be observed by periodic outpatient follow-up. Psychological treatment and social support systems are also very important in preventing recurrence of this disease, and should relieve or alleviate as much as possible the psychological burden and stress that the patient is overweight, help the patient solve the practical difficulties and problems in life and work, improve the patient's coping ability, and actively create a good environment for it, in order to prevent recurrence.
To date, there is no specific examination for depressive disorders. The depression self-rating scale (Se lf-rat ing depress ion sca l e, SDS) is a clinically common depression rating scale. The characteristics are that the subjective feeling of the depressed patient and the change of the subjective feeling in the treatment can be reflected quite intuitively. Is mainly suitable for adults with depression symptoms, including outpatients and inpatients. In addition, diagnosis of depression is also based mainly on medical history, clinical symptoms, course and physical and laboratory examinations, and diagnosis of typical cases is generally not difficult. International general diagnostic criteria are typically I CD-10 and DSM-V. The existing assessment method has some defects, has difficulty in assessing patients with severe slow symptoms of depression, has poor use effect for people with low cultural level or poor intelligence level, and takes longer time for assessment.
Disclosure of Invention
In order to solve the problem of measurement accuracy deviation in the aspect of measuring depression, which is influenced by subjective consciousness of a tested person in the prior art, the invention provides a generation method and a generation system of a depression prediction model, thereby providing basis for early diagnosis and treatment of psychological depression.
In one aspect of the present invention, a method for generating a depression prediction model is provided, including:
step 1: collecting heart rate wave band time sequence data and depression measurement values of at least one user;
step 2: extracting time sequence characteristic data of a user based on the heart rate band time sequence data, and constructing training data and test data;
step 3: generating a depression prediction model based on the time sequence characteristic data of the user and the depression measurement value of the user;
step 4: and generating an optimal depression prediction model.
Preferably, in the step 2, the method includes the following steps:
s21: acquiring the minimum value min and the maximum value max of the number of the heart rate band time sequence data of the user;
s22: setting a sampling time window l, extracting time sequence characteristic data from the heart rate wave band time sequence data with the length of l, and constructing user training data and user test data under the time window l.
Preferably, extracting time sequence characteristic data from the intercepted heart rate wave band time sequence data with the final length of l; the time sequence characteristic data is a statistical characteristic calculated from a section of heart rate wave band time sequence data recorded in time sequence, and comprises the following steps: maximum, minimum, mean, standard deviation, dynamic range, kurtosis, skewness, slope, intercept, mean square error.
Preferably, in the step 3, the method includes the following steps:
s31: and using time sequence characteristic data of user training data under the time window l as input, using a user depression measurement value in the user training data as output, and training to obtain a prediction model under the time window l.
Preferably, in the step 4, the method includes the following steps:
s41: l is traversed [ min, max ], and the steps S2-S3 are repeated to generate max-min+1 prediction models;
s42: calculating an error between a prediction result obtained after time sequence characteristic data of user test data under a time window l are input into a prediction model under the time window l and a user depression measurement value in the user test data;
s43: calculating the average error of a depression prediction model under a time window l, and taking the average error as a performance evaluation index of the depression prediction model under the time window l;
s44: among the max-min+1 prediction models, a depression prediction model with the smallest average error is used as an optimal depression prediction model, and a sampling time window corresponding to the optimal depression prediction model is an optimal sampling time window.
According to another aspect of the present invention, there is provided a depression prediction system comprising: the system comprises a data acquisition module, a characteristic extraction module, a training sample construction module, a neural network training module, an optimal prediction model acquisition module and a prediction analysis module, wherein,
the data acquisition module is used for receiving the heart rate wave band time sequence data to be tested;
the characteristic extraction module is used for setting a sampling time window and generating time sequence characteristic data of heart rate wave band time sequence data under the time window; transmitting the time sequence characteristic data of the user to the training sample construction module, and transmitting the time sequence characteristic data to be tested to an optimal depression prediction model;
the training sample construction module is used for constructing training data and test data under a time window l for the time sequence characteristic data transmitted by the characteristic extraction module; transmitting the training data to the neural network training module, and transmitting the test data to the optimal prediction model acquisition module;
the neural network training module is used for training according to the training data under the time window l to obtain a prediction model under the time window l;
the optimal prediction model acquisition module is used for acquiring an optimal prediction model, and the optimal prediction model outputs a depression state score to be tested; and
the prediction analysis module is used for receiving the time sequence data of the heart rate wave band to be tested, transmitting the time sequence data of the heart rate wave band to the feature extraction module, transmitting the returned result to the optimal depression prediction model, and judging the depression state to be tested according to the returned depression state score to be tested.
Preferably, in the feature extraction module, generating time sequence feature data of heart rate band time sequence data with final length of l under a time window of l; the time sequence characteristic data is a statistical characteristic calculated from a section of heart rate wave band time sequence data recorded in time sequence, and comprises the following steps: maximum, minimum, mean, standard deviation, dynamic range, kurtosis, skewness, slope, intercept, mean square error; the time series characteristic data also includes depression measurements of the user.
Depression measurements may be obtained using a depression meter as described in the background or other methods, and later decisions are assessed accordingly to the criteria of the method used.
Preferably, in the training sample construction module, the time sequence feature data with a set proportion is randomly selected as training data, and the rest of the time sequence feature data is used as test data; the sampling time window l has the value of [ min, max ], and the max-min+1 group of training data and test data can be obtained.
Some numerical values can be set for the l manually, so that the calculation speed can be increased.
Preferably, in the neural network training module, time sequence characteristic data of a user in training data under a time window l is used as input of a depression prediction model, a depression measurement value corresponding to the user is used as output of the depression prediction model, and the depression prediction model under the time window l is obtained through training; and when the value of the sampling time window l is [ min, max ], obtaining max-min+1 depression prediction models, wherein the output of the depression prediction models is a depression state score.
Preferably, in the optimal prediction model obtaining module, the optimal prediction model is generated by using the following steps:
(1) Calculating an error between a prediction result obtained after time sequence characteristic data of user test data under a time window l are input into a prediction model under the time window l and a user depression measurement value in the user test data;
(2) Calculating the average error of a depression prediction model under a time window l, and taking the average error as a performance evaluation index of the depression prediction model under the time window l;
(3) Among the max-min+1 prediction models, the prediction model with the smallest average error is used as an optimal depression prediction model, and a sampling time window corresponding to the optimal depression prediction model is an optimal sampling time window.
According to the invention, through the prediction model obtained by training, the user depression state can be automatically predicted, the prediction efficiency is improved, and the early warning of psychological depression can be realized.
Drawings
FIG. 1 is a flow chart of a method for generating a predictive model of depression in accordance with one embodiment of the invention;
fig. 2 is a schematic diagram of the structure of a depression prediction system according to an embodiment of the present invention.
Specific dimensions, structures and devices are labeled in the drawings in order to clearly realize the structure of the embodiment of the present invention, but this is only for illustrative purposes and is not intended to limit the present invention to the specific dimensions, structures, devices and environments, and those skilled in the art may make adjustments or modifications to these devices and environments according to specific needs, and the adjustments or modifications made remain included in the scope of the appended claims.
Detailed Description
The method for generating a depression prediction model and the depression prediction system provided by the invention are described in detail below with reference to the accompanying drawings and specific embodiments.
In the following description, various aspects of the present invention will be described, however, it will be apparent to those skilled in the art that the present invention may be practiced with only some or all of the structures or processes of the present invention. For purposes of explanation, specific numbers, configurations and orders are set forth, it is apparent that the invention may be practiced without these specific details. In other instances, well-known features will not be described in detail so as not to obscure the invention.
The idea of the invention is as follows: a PPG (plethysmographic) based method, which uses a photoelectric detection method to irradiate an LED light source into skin tissue of a human body, and then converts a received optical signal into an electrical signal through a photoelectric receiving tube at a receiving end. Since the blood flow in the skin tissue periodically changes with the pulse, and the proportion of the oxyhemoglobin cells in the blood also changes with the pulse, the absorption degree of the oxyhemoglobin cells on the incident light also shows periodic changes with the pulse, and the electric signal received at the receiving end also changes with the pulse. Through an algorithm, the signal can be demodulated to calculate the heart rate, and the heart rate variation time sequence features (RR interval sequences) are extracted according to the heart rate fluctuation, so that a depression prediction model based on the time sequence is trained.
In the present invention, the subject means a subject to be tested; the user is the person who collects the data, and the user is a plurality of people.
The invention provides a method for generating a depression prediction model, which is shown in figure 1 and comprises the following steps:
step 1: collecting heart rate wave band time sequence data and depression measurement values of a user;
step 2: extracting time sequence characteristic data of a user from the heart rate band time sequence data, and constructing training data and test data;
step 3: generating a depression prediction model based on the time sequence characteristic data of the user and the depression measurement value of the user;
step 4: and generating an optimal depression prediction model.
In step 1, the heart rate band time series data is time series data, and can be obtained from a bracelet worn by a user. The depression measurement value may be obtained by a depression self-evaluation scale or the like.
In step 2, time series characteristic data is extracted from the heart rate band time series data, the time series characteristic data is a statistical characteristic calculated from a section of heart rate band time series data recorded in time series, and the method comprises the following steps: maximum value, minimum value, mean value, standard deviation, dynamic range, kurtosis, skewness, slope, intercept, mean square error and the like.
Specifically, firstly, obtaining the minimum value min and the maximum value max of the number of heart rate wave band time sequence data of all users; then setting a sampling time window l, intercepting heart rate wave band time sequence data with the length of l for each user, for example intercepting heart rate wave band time sequence data with the last length of l, then generating time sequence feature data, and dividing the obtained time sequence feature data into user training data and user test data under the time window l according to a preset proportion or randomly; the time series characteristic data, training data and test data all contain depression measurement values of the user.
In step 3, using time sequence characteristic data of user training data under a time window l as input, using a user depression measurement value in the user training data as output, and training to obtain a depression prediction model under the time window l; training neural network models is an existing technology, and there are many open source frameworks that can be used.
In one embodiment, for heart rate band time series data of all users, multiple groups of time series characteristic data can be obtained according to different sampling time window l, training data and test data are constructed by using the data, a depression prediction model is trained, the input of the depression prediction model is the time series characteristic data with the time window length of l of all users, and the output is a depression measurement value of the user.
In the step 4, the value of the time window l is from min to max, and the steps S2-S3 are repeated, so that max-min+1 depression prediction models can be generated; then inputting time sequence characteristic data of user test data under each time window l into a depression prediction model under the time window l, and calculating an error between the obtained prediction result and a user depression measurement value in the user test data; then calculating the average error of the depression prediction model under the time window l, and taking the average error as a performance evaluation index of the depression prediction model; among the max-min+1 depression prediction models, the depression prediction model with the smallest average error is used as an optimal depression prediction model, and the sampling time window corresponding to the optimal depression prediction model is the optimal sampling time window.
Example 1:
the specific generation process of the depression prediction model is as follows:
(1) The heart rate band data received for a user is …,65.00,68.52,77.59,66.70,71.78,72.73,66.86,62.58,66.52,64.48,64.36,71.23.
(2) And (3) carrying out statistical feature extraction on the data, setting l=10, namely intercepting the last 10 heart rate values of the time sequence data, and generating the time sequence feature data of [77.58,62.58,68.48, … ], namely the maximum value, the minimum value, the average value and the like. If l=9, then its corresponding timing characteristic value is [72.73,62.58,67.47, … ], and if l=8, then only the last 8 values thereof are truncated, the extracted timing characteristic value is [72.73,62.58,67.56, … ], and so on. All of these user data are then separated into training data and test data.
(3) The predictive model corresponding to l=10, 9,8 can be trained by taking the time sequence characteristic data of the training data as input and the depression state value of the user as output.
(4) For different values of l 10,9 and 8, selecting test data of N users to evaluate the performance of the prediction model, calculating that the error is [13.19,3.42,6.79] and 3.42 is minimum, and selecting the prediction model corresponding to 3.42 as the optimal prediction model, wherein the time window is l=9.
In this example, the value of l is manually set and may be faster than the calculation speed of traversing from min to max.
According to another aspect of the present invention there is provided a depression prediction system comprising: the system comprises a data acquisition module, a characteristic extraction module, a training sample construction module, a neural network training module, an optimal prediction model acquisition module and a prediction analysis module, wherein,
the data acquisition module is used for receiving heart rate wave band time sequence data to be tested, such as heart rate wave band time sequence data of a bracelet;
the characteristic extraction module is used for generating time sequence characteristic data of heart rate wave band time sequence data under the time window l according to the sampling time window l; the time sequence characteristic data of the user is transmitted to the training sample construction module, and the tested time sequence characteristic data is transmitted to the optimal depression prediction model; features have been described above;
the training sample construction module is used for constructing training data and test data under a time window l for the time sequence characteristic data transmitted by the characteristic extraction module; transmitting the training data to the neural network training module, and transmitting the test data to the optimal prediction model acquisition module;
the neural network training module is used for training according to the training data under the time window l to obtain a prediction model under the time window l;
the optimal prediction model acquisition module is used for acquiring an optimal prediction model, and the optimal prediction model outputs a depression state score to be tested; and
the prediction analysis module is used for receiving the time sequence data of the heart rate wave band to be tested, transmitting the time sequence data of the heart rate wave band to the feature extraction module, transmitting the returned result to the optimal depression prediction model, and judging the depression state to be tested according to the returned depression state score to be tested.
Assessment can be made using depression measurement scoring criteria as described in the background. The initial depression measurement and the later evaluation criteria are integral.
In the feature extraction module, heart rate fluctuation time sequence data with the length of l can be taken down in a time window l to generate time sequence feature data; the time series characteristic data may also include depression measurements of the user. The time sequence characteristic data is a statistical characteristic calculated from a section of heart rate wave band time sequence data recorded in time sequence, and comprises the following steps: maximum value, minimum value, mean value, standard deviation, dynamic range, kurtosis, skewness, slope, intercept, mean square error and the like. In one embodiment, the heart rate fluctuation time series data with the final length of l can be selected to generate time series characteristic data.
In a training sample construction module, randomly selecting time sequence characteristic data with a set proportion as training data, and taking the rest samples as test data; the sampling time window l has the value of [ min, max ], and the max-min+1 group of training data and test data can be obtained.
In the neural network training module, time sequence characteristic data of users in training data under a time window l are used as input of a depression prediction model, and depression measurement values corresponding to each user are used as output of the depression prediction model, so that the depression prediction model under the time window l is obtained; when the value of the sampling time window l is [ min, max ], max-min+1 depression prediction models can be obtained, and the output of each depression prediction model is a depression state score.
In the optimal prediction model acquisition module, an optimal prediction model is generated using the following steps:
(1) Calculating an error between a prediction result obtained after heart rate band time sequence data of user test data under a time window l are input into a prediction model under the time window l and a user depression measurement value in the user test data;
(2) Calculating the average error of a depression prediction model under a time window l, and taking the average error as a performance evaluation index of the depression prediction model;
(3) Among the max-min+1 depression prediction models, the depression prediction model with the smallest average error is used as an optimal depression prediction model, and a sampling time window corresponding to the optimal depression prediction model is an optimal sampling time window.
Finally, it should be noted that the above embodiments are only intended to describe the technical solution of the present invention and not to limit the technical method, the present invention extends to other modifications, variations, applications and embodiments in application, and therefore all such modifications, variations, applications, embodiments are considered to be within the spirit and scope of the teachings of the present invention.
Claims (7)
1. A method of generating a predictive model of depression, comprising:
step 1: collecting heart rate wave band time sequence data and depression measurement values of at least one user;
step 2: extracting time sequence characteristic data of a user based on the heart rate band time sequence data, and constructing training data and test data;
step 3: generating a depression prediction model based on the time sequence characteristic data of the user and the depression measurement value of the user;
step 4: generating an optimal depression prediction model;
in the step 2, the method comprises the following steps:
s21: acquiring the minimum value min and the maximum value max of the number of the heart rate band time sequence data of the user;
s22: setting a sampling time window l, extracting time sequence characteristic data from the heart rate wave band time sequence data with the length of l for each user, and constructing user training data and user test data under the time window l;
in the step 4, it includes:
s41: l is traversed [ min, max ], and the steps S2-S3 are repeated to generate max-min+1 depression prediction models;
s42: calculating an error between a prediction result obtained after time sequence characteristic data of user test data under a time window l are input into a prediction model under the time window l and a user depression measurement value in the user test data;
s43: calculating the average error of a depression prediction model under a time window l, and taking the average error as a performance evaluation index of the depression prediction model under the time window l;
s44: among the max-min+1 prediction models, a depression prediction model with the smallest average error is used as an optimal depression prediction model, and a sampling time window corresponding to the optimal depression prediction model is an optimal sampling time window.
2. The method according to claim 1, wherein the time sequence characteristic data is extracted from the intercepted heart rate band time sequence data with the final length of l; the time sequence characteristic data is a statistical characteristic calculated from a section of heart rate wave band time sequence data recorded in time sequence, and comprises the following steps: maximum, minimum, mean, standard deviation, dynamic range, kurtosis, skewness, slope, intercept, mean square error.
3. The generating method according to claim 1, characterized in that in the step 3, it comprises the steps of:
s31: using time sequence characteristic data of user training data under a time window l as input, using user depression measurement values in the user training data as output, and training to obtain a depression prediction model under the time window l.
4. A depression prediction system, the system comprising: the system comprises a data acquisition module, a characteristic extraction module, a training sample construction module, a neural network training module, an optimal prediction model acquisition module and a prediction analysis module, wherein,
the data acquisition module is used for receiving the heart rate wave band time sequence data to be tested;
the characteristic extraction module is used for generating time sequence characteristic data of heart rate wave band time sequence data under the time window l according to the sampling time window l; transmitting the time sequence characteristic data of the user to the training sample construction module, and transmitting the time sequence characteristic data to be tested to an optimal depression prediction model;
the training sample construction module is used for constructing training data and test data under a time window according to the time sequence characteristic data transmitted by the characteristic extraction module; transmitting the training data to the neural network training module, and transmitting the test data to the optimal prediction model acquisition module; the sampling time window l has the value of [ min, max ], and can obtain the training data and the test data of the max-min+1 group;
the neural network training module is used for training through training data under the time window l to obtain a depression prediction model under the time window l;
the optimal depression prediction model acquisition module is used for acquiring an optimal depression prediction model, and the optimal depression prediction model outputs a depression state score to be tested; and
the prediction analysis module is used for receiving the time sequence data of the heart rate wave band to be tested, transmitting the time sequence data of the heart rate wave band to the feature extraction module, transmitting a returned result to the optimal depression prediction model, and judging the depression state to be tested according to the returned depression state score to be tested;
in the optimal prediction model acquisition module, an optimal prediction model is generated by using the following steps:
(1) Calculating an error between a prediction result obtained after time sequence characteristic data of user test data under a time window l are input into a prediction model under the time window l and a user depression measurement value in the user test data;
(2) Calculating the average error of a depression prediction model under a time window l, and taking the average error as a performance evaluation index of the depression prediction model under the time window l;
(3) Among the max-min+1 depression prediction models, the depression prediction model with the smallest average error is used as an optimal depression prediction model, and a sampling time window corresponding to the optimal depression prediction model is an optimal sampling time window.
5. The depression prediction system of claim 4, wherein in the feature extraction module, time series feature data of heart rate band time series data of final length l is generated under a time window l; the time sequence characteristic data is a statistical characteristic calculated from a section of heart rate wave band time sequence data recorded in time sequence, and comprises the following steps: maximum, minimum, mean, standard deviation, dynamic range, kurtosis, skewness, slope, intercept, mean square error; the time series characteristic data also includes depression measurements of the user.
6. The depression prediction system of claim 4, wherein in the training sample construction module, a set proportion of the time series characteristic data is randomly selected as training data, and the remaining time series characteristic data is used as test data.
7. The depression prediction system according to claim 4, wherein in the neural network training module, time sequence characteristic data of a user in training data under a time window l is used as an input of a prediction model, a depression measurement value corresponding to the user is used as an output of the prediction model, and the depression prediction model under the time window l is obtained through training; and when the value of the sampling time window l is [ min, max ], obtaining max-min+1 depression prediction models, wherein the output of the depression prediction models is a depression state score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010220741.6A CN111419249B (en) | 2020-03-26 | 2020-03-26 | Depression prediction model generation method and prediction system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010220741.6A CN111419249B (en) | 2020-03-26 | 2020-03-26 | Depression prediction model generation method and prediction system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111419249A CN111419249A (en) | 2020-07-17 |
| CN111419249B true CN111419249B (en) | 2023-04-25 |
Family
ID=71548875
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010220741.6A Active CN111419249B (en) | 2020-03-26 | 2020-03-26 | Depression prediction model generation method and prediction system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111419249B (en) |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104127193B (en) * | 2014-07-14 | 2017-08-18 | 华南理工大学 | Assessment system and its appraisal procedure that a kind of depression degree quantifies |
| CN104127194B (en) * | 2014-07-14 | 2016-05-04 | 华南理工大学 | A kind of evaluating system of the depression based on heart rate variance analyzing method |
| CN104462808B (en) * | 2014-12-04 | 2017-06-16 | 河海大学 | Level of security displacement and the slip variable window dynamic data approximating method of water level |
| CA2988416A1 (en) * | 2015-06-15 | 2016-12-22 | Medibio Limited | Method and system for assessing mental state |
| US20200152332A1 (en) * | 2017-06-12 | 2020-05-14 | Koninklijke Philips N.V. | Systems and methods for dynamic monitoring of patient conditions and prediction of adverse events |
| CN108389631A (en) * | 2018-02-07 | 2018-08-10 | 平安科技（深圳）有限公司 | Varicella morbidity method for early warning, server and computer readable storage medium |
| CN110232483B (en) * | 2019-06-18 | 2021-05-04 | 国网河北省电力有限公司经济技术研究院 | Deep learning load prediction method and device and terminal equipment |
-
2020
- 2020-03-26 CN CN202010220741.6A patent/CN111419249B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111419249A (en) | 2020-07-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2020013302A1 (en) | Mental/nervous system disorder estimation system, estimation program, and estimation method | |
| CN109770921B (en) | Method and device for screening early stage language and cognitive ability of autistic children | |
| Zhu et al. | Detecting emotional reactions to videos of depression | |
| KR102302071B1 (en) | Method for predicting of depression and device for predicting of depression risk using the same | |
| Doan et al. | Predicting dementia with prefrontal electroencephalography and event-related potential | |
| CN113397546B (en) | Method and system for constructing emotion recognition model based on machine learning and physiological signals | |
| Bartman et al. | Relationship between health status and utility measures in older claudicants | |
| Mahesh et al. | Requirements for a reference dataset for multimodal human stress detection | |
| CN115500843A (en) | Sleep stage staging method based on zero sample learning and contrast learning | |
| CN115844393A (en) | Anxiety disorder evaluation device and anxiety disorder evaluation system based on near infrared data | |
| CN117133464B (en) | Intelligent monitoring system and monitoring method for health of old people | |
| Jiao et al. | Feasibility study for detection of mental stress and depression using pulse rate variability metrics via various durations | |
| CN111419249B (en) | Depression prediction model generation method and prediction system | |
| CN112328072A (en) | Multi-mode character input system and method based on electroencephalogram and electrooculogram | |
| Chapman et al. | A psychophysiological causal model of pain report validity | |
| Sintonen | Comparing properties of the 15D and the EQ-5D in measuring health-related quality of life | |
| KR101130761B1 (en) | Thematic Apperception Test device Based on BCI. | |
| CN111415680A (en) | Method for generating anxiety prediction model based on voice and anxiety prediction system | |
| CN116052877A (en) | Diabetes patient depression risk assessment method and assessment system construction method | |
| Li et al. | Predicting PTSD symptoms in firefighters using a fear-potentiated startle paradigm and machine learning | |
| Buettner et al. | Machine Learning Based Diagnostics of Developmental Coordination Disorder using Electroencephalographic Data | |
| Ross et al. | The Memory Assessment Scales in the detection of incomplete effort in mild head injury | |
| RU2663642C1 (en) | Method of differential diagnostics of behavioral reactions | |
| EP4088666A1 (en) | Device for estimating mental/nervous system diseases using speech | |
| Kumar et al. | A survey on Neuropsychiatric Tools and Machine Learning Approaches used in the Diagnosis of Depression |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |