Determination device, learning device, determination method, and learning information production method
The determination device processes electroencephalogram data to accurately determine depression by using specific parameters and machine learning, overcoming the accuracy limitations of conventional methods.
Patent Information
- Application Number
- PCT/JP2024/017828
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-19
- Filing Date
- 2024-05-14
- Publication Date
- 2024-11-28
AI Technical Summary
Conventional methods for determining depression using brain wave information lack accuracy.
A determination device that receives and processes electroencephalogram data, acquiring specific parameters such as non-REM representative power values and using machine learning processing with teacher data to determine depression, including methods like center of gravity information and arithmetic expressions for improved accuracy.
Enables high-accuracy determination of depression using brain wave information by employing advanced parameter acquisition and machine learning techniques.
Smart Images

Figure JP2024017828_28112024_PF_FP_ABST
Abstract
Description
Determination device, learning device, determination method, and learning information production method
[0001] The present invention relates to a determination device or the like that determines whether or not a person has a mental illness such as depression using one or more parameters acquired from electroencephalogram information.
[0002] Conventionally, there has been a method for assessing a user's mental state using questionnaire data and measurement data obtained by measuring changes in the high frequency (HF) and low frequency (LF) components of the user's heart rate variability (HRV) index (see Patent Document 1).
[0003] JP 2016-16369 A
[0004] However, in conventional techniques, it has not been possible to accurately determine depression using electroencephalogram information.
[0005] The judgment device of the first invention is a judgment device comprising: a reception unit that receives EEG information; a parameter acquisition unit that acquires one or more parameters from the EEG information; a judgment unit that refers to a storage unit in which learning information based on two or more teacher data sets having one or more positive examples having one or more parameters that indicate the characteristics of the EEG information of patients with depression and one or more negative examples having one or more parameters that indicate the characteristics of the EEG information of patients who are not depressed, and uses the one or more parameters and learning information acquired by the parameter acquisition unit to acquire a judgment result indicating whether the patient whose EEG information is received by the reception unit is depressed or not; and an output unit that outputs the judgment result.
[0006] With this configuration, depression can be diagnosed with high accuracy using electroencephalogram information.
[0007] In addition, in the judgment device of the second invention, compared to the first invention, the parameter acquisition unit acquires one or more parameters including a non-REM representative power value from the EEG information, and the judgment unit uses the one or more parameters and learning information to obtain a judgment result indicating whether the patient of the EEG information received by the reception unit is depressed or not, and the parameter acquisition unit acquires the non-REM representative power value, which is a value obtained using the overall power value of the frequency in epochs determined to be non-REM sleep in the EEG information and the number of epochs, and the non-REM representative power value becomes larger as the power value of the frequency becomes larger and becomes smaller as the number of epochs becomes larger.
[0008] With this configuration, depression can be diagnosed with higher accuracy using useful parameters that can be obtained from electroencephalogram information.
[0009] Furthermore, in the determination device of the third invention, compared to the second invention, the parameter acquisition unit adds up power values of frequencies between approximately 8 Hz and approximately 14 Hz calculated by fast Fourier transform in epochs determined to be non-REM sleep to obtain an overall power value, and divides the overall power value by the number of non-REM epochs to obtain a non-REM representative power value, which is the average power value per epoch.
[0010] With this configuration, depression can be diagnosed with higher accuracy using useful parameters that can be obtained from electroencephalogram information.
[0011] Furthermore, the judgment device of the fourth invention is a judgment device in which, compared to any one of the first to third inventions, the learning information is a learning model obtained by performing a machine learning learning process on two or more teacher data, and the judgment unit performs a machine learning prediction process using one or more parameters and the learning model to obtain a judgment result.
[0012] With this configuration, depression can be diagnosed with high accuracy using electroencephalogram information.
[0013] Furthermore, the determination device of the fifth invention is a determination device according to any one of the first to third inventions, wherein the positive examples are positive example vectors whose elements are two or more parameters that indicate characteristics of the EEG information of patients who are depressed, and the negative examples are negative example vectors whose elements are two or more parameters that indicate characteristics of the EEG information of patients who are not depressed, the determination unit acquires a positive example centroid vector that is a vector of the centroids of one or more positive example vectors, acquires a negative example centroid vector that is a vector of the centroids of one or more negative example vectors, acquires a positive example distance that is the distance between the vector whose elements are the two or more parameters acquired by the parameter acquisition unit and the positive example centroid vector, acquires a negative example distance that is the distance between the vector whose elements are the two or more parameters acquired by the parameter acquisition unit and the negative example centroid vector, and acquires a determination result using the positive example distance and the negative example distance.
[0014] With this configuration, depression can be diagnosed with high accuracy using electroencephalogram information.
[0015] Furthermore, the judgment device of the sixth invention is a judgment device according to any one of the first to third inventions, wherein the positive examples are positive example vectors whose elements are one or more parameters that indicate characteristics of EEG information from patients with depression, the negative examples are negative example vectors whose elements are one or more parameters that indicate characteristics of EEG information from patients who are not depressed, the learning information is a correspondence table having correspondence information with one or more positive example vectors and correspondence information with one or more negative example vectors, the judgment unit constructs a vector whose elements are the one or more parameters acquired by the parameter acquisition unit, acquires the similarity between the vector and one or more positive example vectors and the similarity between the vector and one or more negative example vectors, and acquires a judgment result corresponding to the positive example vector or the negative example vector whose similarity satisfies the adoption condition.
[0016] With this configuration, depression can be diagnosed with high accuracy using electroencephalogram information.
[0017] Furthermore, the judgment device of the seventh invention is a judgment device according to any one of the first to third inventions, wherein the positive examples are positive example vectors whose elements are one or more parameters that indicate characteristics of the EEG information of patients with depression, the negative examples are negative example vectors whose elements are one or more parameters that indicate characteristics of the EEG information of patients who are not depressed, the learning information is an arithmetic formula that outputs positive example values that are values corresponding to positive examples for positive example vectors, and outputs negative example values that are values corresponding to negative examples for negative example vectors, and the judgment unit substitutes the one or more parameters acquired by the parameter acquisition unit into the arithmetic formula to acquire a score and acquire a judgment result corresponding to the score.
[0018] With this configuration, depression can be diagnosed with high accuracy using electroencephalogram information.
[0019] In addition, the learning device of the eighth invention is a learning device that includes a parameter acquisition unit that acquires one or more parameters from two or more pieces of electroencephalogram information corresponding to a judgment result of whether or not the patient is depressed, a learning unit that acquires teacher data for each of the two or more pieces of electroencephalogram information, the teacher data having one or more parameters acquired by the parameter acquisition unit and a judgment result corresponding to each piece of electroencephalogram information, and acquires learning information using the two or more pieces of teacher data, and a storage unit that stores the learning information acquired by the learning unit.
[0020] With this configuration, it is possible to create learning information for accurately diagnosing depression using electroencephalogram information.
[0021] According to the determination device of the present invention, depression can be determined with high accuracy using electroencephalogram information.
[0022] FIG. 1 is a block diagram of the determination system A according to the first embodiment; FIG. 2 is a flowchart illustrating an example of the operation of the determination device 1; FIG. 3 is a flowchart illustrating an example of the parameter acquisition process; FIG. 4 is a diagram illustrating the parameter management table; FIG. 5 is a diagram illustrating the parameter management table; FIG. 6 is a diagram illustrating the parameter management table;
[0023] Hereinafter, embodiments of a determination device and the like will be described with reference to the drawings. Note that, in the embodiments, components with the same reference numerals perform similar operations, and therefore, repeated description may be omitted.
[0024] (Embodiment 1) In this embodiment, a description will be given of a determination device that determines whether a person has a mental illness using one or more parameters acquired from electroencephalogram information. Note that the determination of whether a person has a mental illness here is usually a determination of whether or not the person has depression, but it may also be a determination of whether or not the person has manic depression, or a broad determination of whether or not the person has a mental illness. Furthermore, the determination of whether a person has a mental illness may also be a determination of the severity of the mental illness. It is preferable to interpret mental illness broadly to include, for example, not only depression but also bipolar disorder, schizophrenia, adjustment disorder, etc. The determination of whether or not the person has depression may be considered to be synonymous with the determination of whether or not the person has a mental illness.
[0025] In this embodiment, the determination method will be described using, for example, machine learning, center of gravity information, a correspondence table, or an arithmetic expression.
[0026] In this specification, information X being associated with information Y means that information Y can be obtained from information X, or information X can be obtained from information Y, and the method of association is not important. Information X and information Y may be linked, may exist in the same buffer, information X may be included in information Y, or information Y may be included in information X, etc.
[0027] 1 is a conceptual diagram of a determination system A according to this embodiment. The determination system A includes, for example, a determination device 1, a learning device 2, and one or more terminal devices 3.
[0028] The determination device 1 is a device that determines whether or not a subject is depressed using electroencephalogram information. The determination device 1 is usually a so-called server. The determination device 1 is, for example, a cloud server or an ASP server, but the type is not important. The determination device 1 may operate as a standalone device. When the determination device 1 operates as a standalone device, the terminal device 3 is not required.
[0029] The learning device 2 is a device that acquires learning information used for determination by the determination device 1. Details of the learning device 2 will be described in the second embodiment.
[0030] The terminal device 3 is a terminal used by a user. The terminal device 3 is, for example, a device that provides electroencephalogram information to the determination device 1 and receives and outputs a determination result from the determination device 1. The terminal device 3 is, for example, a personal computer, a smartphone, a tablet terminal, or an electroencephalograph, but the type is not limited thereto.
[0031] It is preferable that the determination device 1 and one or more terminal devices 3, and the learning device 2 and one or more terminal devices 3, are capable of communicating via a network such as the Internet.
[0032] 2 is a block diagram of a determination system A according to the present embodiment. The determination device 1 includes a storage unit 11, a reception unit 12, a processing unit 13, and an output unit 14. The processing unit 13 includes a parameter acquisition unit 131 and a determination unit 132.
[0033] Various types of information are stored in the storage unit 11. The various types of information include, for example, learning information and brain wave information, which will be described later.
[0034] The learning information is information for determining whether a patient is depressed using EEG information. The learning information is, for example, a learning model, but may also be center of gravity information, a correspondence table, or an arithmetic formula. The learning information is preferably information acquired by the learning device 2 described in the second embodiment.
[0035] A learning model is a model obtained by performing machine learning learning processing on two or more pieces of training data. The training data is, for example, a positive example obtained from the EEG information of patients with depression, or a negative example obtained from the EEG information of patients without depression. The training data is, for example, a positive example including one or more parameters obtained from the EEG information of patients with depression, or a negative example including one or more parameters obtained from the EEG information of patients without depression.
[0036] The center of gravity information includes a positive example representative vector and a negative example representative vector. The positive example representative vector is a vector whose elements are representative values of one or more parameters obtained from the EEG information of one or more patients with depression. The negative example representative vector is a vector whose elements are representative values of one or more parameters obtained from the EEG information of one or more patients who do not have depression.
[0037] The positive example representative vector is, for example, a positive example centroid vector. The positive example centroid vector is a vector of the centroids of one or more positive example vectors. The positive example representative vector is, for example, a vector whose elements are the representative values of each element of two or more positive example vectors. The negative example representative vector is, for example, a negative example centroid vector. The negative example centroid vector is a vector of the centroids of one or more negative example vectors. The negative example representative vector is, for example, a vector whose elements are the representative values of each element of two or more negative example vectors. The representative value is, for example, the average value or the median.
[0038] A correspondence table is a table that has two or more pieces of correspondence information. The correspondence information is information that associates a positive example vector having one or more parameters obtained from the EEG information of a patient with depression, with depression. The correspondence information is information that associates a negative example vector having one or more parameters obtained from the EEG information of a patient who is not depressed, with depression. The correspondence information has, for example, a positive example vector or a negative example vector, and a flag that indicates whether or not the patient is depressed. The flag is, for example, "1" that indicates depression, or "0" that indicates not depression.
[0039] The arithmetic expression is an arithmetic expression that outputs a positive example value (e.g., "1") that corresponds to a positive example for a positive example vector, and outputs a negative example value (e.g., "-1") that corresponds to a negative example for a negative example vector.
[0040] Electroencephalogram information is information on brain waves. Electroencephalogram information is information that records the potential difference between two points on the scalp. Electroencephalogram information is a collection of amplitude numbers in a time series. Electroencephalogram information is usually information acquired by an electroencephalograph.
[0041] The receiving unit 12 receives one or more pieces of electroencephalogram information. The electroencephalogram information is received, for example, in combination with a user identifier. The user identifier is information for identifying a user who is to be determined as having depression. The user identifier is, for example, a user ID, name, telephone number, or email address.
[0042] Here, "reception" usually refers to the reception of information transmitted via a wired or wireless communication line, but may also be a concept that includes the reception of information input from an input device such as a keyboard, mouse, or touch panel, or the reception of information read from a recording medium such as an optical disk, magnetic disk, or semiconductor memory. The reception may be, for example, reception from an electroencephalograph.
[0043] When the determination device 1 is a standalone device, the means for inputting electroencephalogram information may be any means, such as an electroencephalograph, a mouse, a menu screen, etc. The reception unit 12 may be realized by a device driver for an input means such as an electroencephalograph or a mouse, or control software for a menu screen, etc.
[0044] The processing unit 13 performs various types of processing, such as processing performed by a parameter acquisition unit 131 or a determination unit 132.
[0045] The parameter acquisition section 131 acquires one or more parameters from one or more pieces of electroencephalogram information received by the reception section 12 .
[0046] The parameter acquisition unit 131 acquires, for example, one or more epoch attribute values for each of one or more epochs in the electroencephalogram information, and acquires one or more parameters using the one or more epoch attribute values. The epoch attribute value is information indicating the characteristics of the electroencephalogram information of one epoch. Note that an epoch is the length of a unit interval, such as 30 seconds, but may also be 20 seconds, 40 seconds, etc. Preferably, the one or more epoch attribute values are, for example, one or more pieces of information selected from the group consisting of frequency, average amplitude, duration of a specific frequency range, and power value of a specific frequency range.
[0047] The parameter acquisition unit 131 acquires parameters using, for example, one or more pieces of raw information. Note that the raw information may be, for example, the above-mentioned epoch attribute value, the number of non-REM epochs, and the number of REM epochs.
[0048] It is preferable that the parameter acquisition section 131 acquires one or more parameters including a power value ratio parameter (also referred to as a seventh parameter) described later from one or more pieces of electroencephalogram information received by the reception section 12.
[0049] The parameter acquiring unit 131 preferably acquires, for example, a first parameter to a 36th parameter (described later) from one or more pieces of electroencephalogram information accepted by the accepting unit 12. The parameter acquiring unit 131 typically acquires the parameters after detecting one or more epochs determined to be non-REM sleep (hereinafter referred to as "non-REM epochs") in the electroencephalogram information. Furthermore, the parameter acquiring unit 131 acquires the parameters after detecting, for example, one or more epochs determined to be REM sleep (hereinafter referred to as "REM epochs") in the electroencephalogram information.
[0050] For example, the parameter acquiring unit 131 determines whether each section of the electroencephalogram information to be determined is a non-REM sleep section or a REM sleep section. This process is a well-known technique. Note that the section is, for example, an epoch. Next, the parameter acquiring unit 131 acquires, for example, the following first to 36th parameters using, for example, the determination result of whether the section is a non-REM sleep section or a REM sleep section.
[0051] It is highly preferable that the parameter acquisition unit 131 acquires the seventh parameter (7) from among the following 36 parameters.
[0052] Furthermore, it is preferable that the parameter acquisition unit 131 acquires one or more parameters from among the following 36 parameters: (2), (8), (9), and (15).
[0053] In addition, it is preferable that the parameter acquisition unit 131 acquires one or more parameters from the following 36 parameters: (1) (3) (4) (5) (6) (10) (11) (12) (13) (14) (16) (17) (18) (19) (20) (21).
[0054] Furthermore, the parameter acquisition unit 131 may acquire one or more parameters from (22) to (36) of the following 36 parameters: (1) First parameter
[0055] The parameter acquiring unit 131 acquires the number of first specific spindles in the non-REM epochs. The parameter acquiring unit 131 also acquires the number of non-REM epochs.
[0056] The first specific spindle is a spindle having a first specific frequency, a first specific duration, and a first specific average amplitude in a non-REM epoch. Here, the first specific frequency is, for example, a frequency of approximately 8 Hz to approximately 14 Hz, although it goes without saying that some degree of error is acceptable. The first specific duration is, for example, approximately 0.25 seconds to approximately 2.5 seconds, although it goes without saying that some degree of error is acceptable. The first specific average amplitude is, for example, approximately 15 μV or more, although it goes without saying that some degree of error is acceptable.
[0057] Next, the parameter acquisition unit 131 acquires a first parameter that increases as the number of first specific spindles increases and increases as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the first parameter using, for example, a first arithmetic expression that is an increasing function with the number of first specific spindles as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the first arithmetic expression is, for example, "number of first specific spindles / number of non-REM epochs." In this case, the first parameter is the average number of the first specific spindles per non-REM epoch. (2) Second Parameter
[0058] The parameter acquiring unit 131 acquires the duration of one or more first specific spindles, and acquires a total duration by adding up the durations of the one or more first specific spindles. The parameter acquiring unit 131 acquires the number of non-REM epochs.
[0059] Next, the parameter acquisition unit 131 acquires a second parameter whose value increases as the total time length increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the second parameter using, for example, a second arithmetic expression that is an increasing function with the total time length as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the second arithmetic expression is, for example, "total time length / number of non-REM epochs." In this case, the second parameter is the average time length of the first specific spindle per non-REM epoch. (3) Third Parameter
[0060] The parameter acquiring unit 131 detects one or more second specific spindles, acquires the number of second specific spindles, and acquires the number of non-REM epochs.
[0061] The second specific spindle is a spindle having a first specific frequency, a first specific duration, and a second specific average amplitude during the NREM epoch. The second specific average amplitude is, for example, approximately 20 μV or more, although some margin of error may be allowed.
[0062] Next, the parameter acquisition unit 131 acquires a third parameter, which increases as the number of second specific spindles increases and increases as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the third parameter using, for example, a third arithmetic expression that is an increasing function with the number of second specific spindles as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the third arithmetic expression is, for example, "number of second specific spindles / number of non-REM epochs." In this case, the third parameter is the average number of the second specific spindles per non-REM epoch. (4) Fourth Parameter
[0063] The parameter acquisition unit 131 detects one or more second specific spindles. Then, the parameter acquisition unit 131 acquires the duration of each of the one or more second specific spindles, and acquires a total duration by adding up the durations of the one or more second specific spindles. The parameter acquisition unit 131 acquires the number of non-REM epochs.
[0064] Next, the parameter acquisition unit 131 acquires a fourth parameter whose value increases as the total time length increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the fourth parameter using, for example, a fourth arithmetic expression that is an increasing function with the total time length as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the fourth arithmetic expression is, for example, "total time length / number of non-REM epochs." In this case, the fourth parameter is the average time length of the second specific spindle per non-REM epoch. (5) Fifth Parameter
[0065] The parameter acquiring unit 131 detects one or more third specific spindles, acquires the number of the third specific spindles, and acquires the number of non-REM epochs.
[0066] The third specific spindle is a spindle having a first specific frequency, a first specific duration, and a third specific mean amplitude in the NREM epoch. The third specific mean amplitude is, for example, approximately 30 μV or more, although some margin of error may be allowed.
[0067] Next, the parameter acquisition unit 131 acquires a fifth parameter, which increases as the number of third specific spindles increases and increases as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the fifth parameter using, for example, a fifth arithmetic expression that is an increasing function with the number of third specific spindles as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the fifth arithmetic expression is, for example, "number of third specific spindles / number of non-REM epochs." In this case, the fifth parameter is the average number of the third specific spindles per non-REM epoch. (6) Sixth Parameter
[0068] The parameter acquisition unit 131 detects one or more third specific spindles. Then, the parameter acquisition unit 131 acquires the duration of each of the one or more third specific spindles, and acquires a total duration by adding up the durations of the one or more third specific spindles. The parameter acquisition unit 131 acquires the number of non-REM epochs.
[0069] Next, the parameter acquisition unit 131 acquires a sixth parameter, which increases as the total time length increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the sixth parameter using, for example, a sixth arithmetic expression that is an increasing function with the total time length as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the sixth arithmetic expression is, for example, "total time length / number of non-REM epochs." In this case, the sixth parameter is the average time length of the third specific spindle per non-REM epoch. (7) Seventh Parameter
[0070] The parameter acquisition unit 131 acquires the power value of the first specific frequency in each of the one or more non-REM epochs, and acquires a total non-REM power value by adding up the one or more power values. The parameter acquisition unit 131 also acquires the number of the non-REM epochs.
[0071] The parameter acquisition unit 131 calculates the power value for each of one or more non-REM epochs by fast Fourier transform (FFT).
[0072] Next, the parameter acquisition unit 131 acquires a seventh parameter, which increases as the non-REM total power value increases and as the number of non-REM epochs decreases. The seventh parameter can be said to be a non-REM representative power value. The parameter acquisition unit 131 acquires the seventh parameter using, for example, a seventh arithmetic expression, which is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the seventh arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the seventh parameter is the average power value per non-REM epoch. (8) Eighth Parameter
[0073] The parameter acquisition unit 131 acquires the total non-REM power value and the number of non-REM epochs.
[0074] The parameter acquiring unit 131 acquires the power values of the first specific frequency in one or more REM epochs, and acquires a total REM power value by adding up the one or more power values. The parameter acquiring unit 131 also acquires the number of the REM epochs.
[0075] Next, the parameter acquisition unit 131 acquires an eighth parameter which increases as the non-REM total power value increases, increases as the number of non-REM epochs decreases, decreases as the REM total power value increases, and increases as the number of REM epochs increases.
[0076] The parameter acquisition unit 131 acquires the eighth parameter using, for example, an eighth arithmetic expression, which is an increasing function with the non-REM total power value and the number of REM epochs as parameters, and a decreasing function with the REM total power value and the number of non-REM epochs as parameters. Note that the eighth arithmetic expression is, for example, the ratio of (non-REM total power value / number of non-REM epochs) to (REM total power value / number of REM epochs). (9) Ninth Parameter
[0077] The parameter acquiring unit 131 detects one or more ninth waveforms in the non-REM epochs and acquires the number of the ninth waveforms.
[0078] The ninth waveform is a waveform of a ninth specific frequency and a ninth specific amplitude. Here, the ninth specific frequency is, for example, a frequency of approximately 8 Hz to approximately 12 Hz, although it goes without saying that some degree of error may be allowed. The ninth specific amplitude is, for example, a frequency of approximately 20 μV to approximately 200 μV, although it goes without saying that some degree of error may be allowed.
[0079] Next, the parameter acquisition unit 131 acquires a ninth parameter, which increases as the number of ninth waveforms in a non-REM epoch increases and increases as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the ninth parameter using, for example, a ninth arithmetic expression that is an increasing function with the number of ninth waveforms as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the ninth arithmetic expression is, for example, "number of ninth waveforms in a non-REM epoch / number of non-REM epochs." In this case, the ninth parameter is the average number of ninth waveforms in a non-REM epoch. (10) Tenth Parameter
[0080] The parameter acquiring unit 131 acquires one or more power values of the ninth waveform in one or more non-REM epochs, and acquires a total non-REM power value by adding up the one or more power values. The parameter acquiring unit 131 also acquires the number of the non-REM epochs.
[0081] The parameter acquisition unit 131 calculates the power value of the ninth waveform by FFT for each of one or more non-REM epochs.
[0082] Next, the parameter acquisition unit 131 acquires a tenth parameter that increases as the non-REM total power value increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the tenth parameter using, for example, a tenth arithmetic expression that is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the tenth arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the tenth parameter is the average power value per one non-REM epoch. (11) Eleventh Parameter
[0083] The parameter acquisition unit 131 applies a filter to a specific band for each of one or more non-REM epochs to remove noise. Then, the parameter acquisition unit 131 performs a process similar to the process for acquiring the tenth parameter to acquire an eleventh parameter.
[0084] That is, the parameter acquisition unit 131 acquires an eleventh parameter that increases as the non-REM total power value increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the eleventh parameter using, for example, an eleventh arithmetic expression that is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the eleventh arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the eleventh parameter is the average power value per one non-REM epoch. (12) Twelfth Parameter
[0085] The parameter acquisition unit 131 acquires a tenth parameter. Furthermore, the parameter acquisition unit 131 acquires a REM tenth parameter, which is the same parameter as the tenth parameter, for one or more REM epochs. Next, the parameter acquisition unit 131 acquires a twelfth parameter, which is the ratio of the tenth parameter to the REM tenth parameter.
[0086] The parameter acquisition unit 131 acquires a 12th parameter that becomes larger as the non-REM total power value increases, becomes larger as the number of non-REM epochs decreases, becomes smaller as the REM total power value increases, and becomes larger as the number of REM epochs increases.
[0087] The parameter acquisition unit 131 acquires the twelfth parameter using a twelfth arithmetic expression, which is an increasing function with the non-REM total power value and the number of REM epochs as parameters, and a decreasing function with the REM total power value and the number of non-REM epochs as parameters, for example. As described above, the twelfth arithmetic expression is, for example, the ratio of (non-REM total power value / number of non-REM epochs) to (REM total power value / number of REM epochs). (13) Thirteenth Parameter
[0088] The parameter acquisition unit 131 applies a filter to a specific band to remove noise. Next, the parameter acquisition unit 131 performs a process similar to the process for acquiring the twelfth parameter to acquire the thirteenth parameter. (14) Fourteenth Parameter
[0089] The parameter acquiring unit 131 detects one or more fourteenth waveforms and acquires the power values of the fourteenth waveforms. The parameter acquiring unit 131 acquires a total non-REM power value by adding up one or more power values. The parameter acquiring unit 131 also acquires the number of non-REM epochs.
[0090] The fourteenth waveform is a waveform of the fourteenth specific frequency in the non-REM epoch. Here, the ninth specific frequency is, for example, a frequency of approximately 0.5 Hz to approximately 2 Hz, although it goes without saying that some degree of error may be allowed.
[0091] Next, the parameter acquisition unit 131 acquires a fourteenth parameter, which increases as the non-REM total power value increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the fourteenth parameter using, for example, a fourteenth arithmetic expression that is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the fourteenth arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the fourteenth parameter is the average power value per one non-REM epoch. (15) Fifteenth Parameter
[0092] The parameter acquisition unit 131 acquires a fourteenth parameter. Furthermore, the parameter acquisition unit 131 acquires a fourteenth REM parameter, which is the same as the fourteenth parameter, for one or more REM epochs. Next, the parameter acquisition unit 131 acquires a fifteenth parameter, which is, for example, the ratio of the fourteenth parameter to the fourteenth REM parameter.
[0093] The parameter acquisition unit 131 acquires a 15th parameter that becomes larger as the non-REM total power value increases, becomes larger as the number of non-REM epochs decreases, becomes smaller as the REM total power value increases, and becomes larger as the number of REM epochs increases.
[0094] The parameter acquisition unit 131 acquires the 15th parameter using, for example, a 15th arithmetic expression, which is an increasing function with the NREM total power value and the number of REM epochs as parameters, and a decreasing function with the REM total power value and the number of NREM epochs as parameters. As described above, the 15th arithmetic expression is, for example, the ratio of (NREM total power value / number of NREM epochs) to (REM total power value / number of REM epochs). (16) 16th Parameter
[0095] The parameter acquiring unit 131 detects one or more sixteenth waveforms and acquires the power values of the sixteenth waveforms. The parameter acquiring unit 131 acquires a total non-REM power value by adding up one or more power values. The parameter acquiring unit 131 also acquires the number of non-REM epochs.
[0096] The sixteenth waveform is a waveform of a sixteenth specific frequency in the non-REM epoch. Here, the sixteenth specific frequency is, for example, a frequency of approximately 1 Hz to approximately 2 Hz, although it goes without saying that some degree of error may be allowed.
[0097] Next, the parameter acquisition unit 131 acquires a 16th parameter, which increases as the non-REM total power value increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the 16th parameter using, for example, a 16th arithmetic expression that is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the 16th arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the 16th parameter is the average power value per non-REM epoch. (17) 17th Parameter
[0098] The parameter acquisition unit 131 acquires the 16th parameter. Furthermore, the parameter acquisition unit 131 acquires a REM 16th parameter, which is the same parameter as the 16th parameter, for one or more REM epochs. Next, the parameter acquisition unit 131 acquires a 17th parameter, which is, for example, the ratio of the 16th parameter to the REM 16th parameter.
[0099] The parameter acquisition unit 131 acquires a 17th parameter whose value increases as the non-REM total power value increases, decreases as the number of non-REM epochs decreases, decreases as the REM total power value increases, and increases as the number of REM epochs increases.
[0100] The parameter acquisition unit 131 acquires the 17th parameter using, for example, a 17th arithmetic expression, which is an increasing function with the non-REM total power value and the number of REM epochs as parameters, and a decreasing function with the REM total power value and the number of non-REM epochs as parameters. As described above, the 17th arithmetic expression is, for example, the ratio of (non-REM total power value / number of non-REM epochs) to (REM total power value / number of REM epochs). (18) 18th Parameter
[0101] The parameter acquiring unit 131 detects one or more eighteenth waveforms and acquires the power values of the eighteenth waveforms. The parameter acquiring unit 131 acquires a total non-REM power value by adding up one or more power values. The parameter acquiring unit 131 also acquires the number of non-REM epochs.
[0102] The eighteenth waveform is a waveform of an eighteenth specific frequency in the non-REM epoch. Here, the eighteenth specific frequency is, for example, a frequency of approximately 18 Hz to approximately 35 Hz, although it goes without saying that some degree of error may be allowed.
[0103] Next, the parameter acquisition unit 131 acquires an 18th parameter, which increases as the non-REM total power value increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the 18th parameter using, for example, an 18th arithmetic expression, which is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the 18th arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the 18th parameter is the average power value per non-REM epoch. (19) 19th Parameter
[0104] The parameter acquisition unit 131 applies a filter to a specific band to remove noise. Then, the parameter acquisition unit 131 performs a process similar to the process for acquiring the 18th parameter to acquire the 19th parameter. (20) 20th parameter
[0105] The parameter acquiring unit 131 acquires the 18th parameter. Furthermore, the parameter acquiring unit 131 acquires a REM 18th parameter, which is the same parameter as the 18th parameter, for one or more REM epochs. Next, the parameter acquiring unit 131 acquires a 20th parameter, which is, for example, the ratio of the 18th parameter to the REM 18th parameter.
[0106] The parameter acquisition unit 131 acquires a 20th parameter that becomes larger as the non-REM total power value increases, becomes larger as the number of non-REM epochs decreases, becomes smaller as the REM total power value increases, and becomes larger as the number of REM epochs increases.
[0107] The parameter acquisition unit 131 acquires the 20th parameter using, for example, a 20th arithmetic expression, which is an increasing function with the non-REM total power value and the number of REM epochs as parameters, and a decreasing function with the REM total power value and the number of non-REM epochs as parameters. Note that, as described above, the 20th arithmetic expression is, for example, the "ratio of (non-REM total power value / number of non-REM epochs) to (REM total power value / number of REM epochs)." (21) 21st Parameter
[0108] The parameter acquisition unit 131 applies a filter to a specific band to remove noise. Then, the parameter acquisition unit 131 performs a process similar to the process for acquiring the 20th parameter to acquire the 21st parameter. (22) 22nd parameter
[0109] The parameter acquiring unit 131 detects one or more twenty-second waveforms in the non-REM epochs and acquires the power values of the twenty-second waveforms. The parameter acquiring unit 131 acquires a total non-REM power value by adding up the one or more power values. The parameter acquiring unit 131 also acquires the number of non-REM epochs.
[0110] The twenty-second waveform is a waveform of a twenty-second specific frequency. Here, the twenty-second specific frequency is, for example, a frequency of approximately 2 Hz to approximately 4 Hz, although it goes without saying that some degree of error may be allowed.
[0111] Next, parameter acquisition unit 131 acquires a 22(1) parameter that increases as the non-REM total power value increases and as the number of non-REM epochs decreases. Parameter acquisition unit 131 acquires the 22(1) parameter using, for example, a 22(1) arithmetic expression that is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the 22(1) arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the 22(1) parameter is the average power value per non-REM epoch.
[0112] Furthermore, the parameter acquiring unit 131 acquires a 22(2) parameter, which is the same as the 22(1) parameter, for one or more REM epochs. Next, the parameter acquiring unit 131 acquires a 22(2) parameter, which is, for example, the ratio of the 22(1) parameter to the 22(2) parameter.
[0113] The parameter acquisition unit 131 acquires a 22nd parameter that becomes larger as the non-REM total power value increases, becomes larger as the number of non-REM epochs decreases, becomes smaller as the REM total power value increases, and becomes larger as the number of REM epochs increases.
[0114] The parameter acquisition unit 131 acquires the 22nd parameter using, for example, a 22nd arithmetic expression, which is an increasing function with the NREM total power value and the number of REM epochs as parameters, and a decreasing function with the REM total power value and the number of NREM epochs as parameters. As described above, the 22nd arithmetic expression is, for example, the ratio of (NREM total power value / number of NREM epochs) to (REM total power value / number of REM epochs). (23) 23rd Parameter
[0115] The parameter acquisition unit 131 acquires the power value of the twenty-third waveform instead of the twenty-second waveform in the process described in (22).The parameter acquisition unit 131 then performs the same process as described in (22) to acquire the twenty-third parameter.
[0116] The 23rd waveform is a waveform of a 23rd specific frequency. Here, the 23rd specific frequency is, for example, a frequency of approximately 4 Hz to approximately 8 Hz, but it goes without saying that some degree of error may be allowed. (24) 24th Parameter
[0117] The parameter acquisition unit 131 applies a filter to a specific band to remove noise. Then, the parameter acquisition unit 131 performs a process similar to the process for acquiring the 23rd parameter to acquire the 24th parameter. (25) 25th Parameter
[0118] The parameter acquisition unit 131 acquires the power value of the 25th waveform instead of the 23rd waveform in the process described in (23).The parameter acquisition unit 131 then performs the same process as described in (23) to acquire the 25th parameter.
[0119] The 25th waveform is a waveform of the 25th specific frequency. Here, the 25th specific frequency is, for example, a frequency of approximately 4 Hz to approximately 6 Hz, but it goes without saying that some degree of error may be allowed. (26) 26th Parameter
[0120] The parameter acquisition unit 131 acquires the power value of the 26th waveform instead of the 25th waveform in the process described in (25).The parameter acquisition unit 131 then performs the same process as described in (25) to acquire the 26th parameter.
[0121] The 26th waveform is a waveform of the 26th specific frequency. Here, the 26th specific frequency is, for example, a frequency of approximately 6 Hz to approximately 8 Hz, but it goes without saying that some degree of error may be allowed. (27) 27th Parameter
[0122] The parameter acquisition unit 131 applies a filter to a specific band to remove noise. Then, the parameter acquisition unit 131 performs a process similar to the process for acquiring the 26th parameter to acquire the 27th parameter. (28) 28th Parameter
[0123] The parameter acquisition unit 131 acquires the power value of the 28th waveform instead of the 26th waveform in the process described in (26).The parameter acquisition unit 131 then performs the same process as described in (25) to acquire the 28th parameter.
[0124] The 28th waveform is a waveform of the 28th specific frequency. Here, the 28th specific frequency is, for example, a frequency of approximately 8 Hz to approximately 10 Hz, but it goes without saying that some degree of error may be allowed. (29) 29th Parameter
[0125] The parameter acquisition unit 131 applies a filter to a specific band to remove noise. Then, the parameter acquisition unit 131 performs a process similar to the process for acquiring the 28th parameter to acquire the 29th parameter. (30) 30th Parameter
[0126] The parameter acquisition unit 131 acquires the power value of the 30th specific frequency in each of the one or more non-REM epochs, and acquires the total non-REM power value by adding up the one or more power values. The parameter acquisition unit 131 also acquires the number of the non-REM epochs.
[0127] The 30th specific frequency is, for example, a frequency of approximately 10 Hz to approximately 12 Hz, but it goes without saying that some degree of error may be allowed.
[0128] The parameter acquisition unit 131 calculates the power value by FFT for each of one or more non-REM epochs.
[0129] Next, the parameter acquisition unit 131 acquires a seventh parameter, which increases as the non-REM total power value increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the thirtieth parameter using, for example, a thirtieth arithmetic expression, which is an increasing function with the non-REM total power value as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the thirtieth arithmetic expression is, for example, "non-REM total power value / number of non-REM epochs." In this case, the thirtieth parameter is the average power value per one non-REM epoch. (31) Thirty-first Parameter
[0130] The parameter acquisition unit 131 acquires the power value of the 30th waveform instead of the 28th waveform in the process described in (28). Then, the parameter acquisition unit 131 performs the same process as described in (28) to acquire the 31st parameter. The 30th waveform is a waveform of the 30th specific frequency. (32) 32nd Parameter
[0131] The parameter acquisition unit 131 acquires the power value of the thirty-second waveform instead of the thirty-tenth waveform in the process described in (31). Then, the parameter acquisition unit 131 performs the same process as described in (31) to acquire the thirty-second parameter. The thirty-second waveform is a waveform of a thirty-second specific frequency. The thirty-second specific frequency is, for example, a frequency of approximately 0.5 Hz to approximately 45 Hz, but it goes without saying that some error may be allowed. (33) Thirty-third Parameter
[0132] The parameter acquisition unit 131 applies a filter to a specific band to remove noise. Then, the parameter acquisition unit 131 performs a process similar to the process for acquiring the 32nd parameter to acquire the 33rd parameter. (34) 34th parameter
[0133] The parameter acquiring unit 131 applies a filter to a specific band to remove noise. Then, the parameter acquiring unit 131 detects one or more ninth waveforms. Then, the parameter acquiring unit 131 acquires the duration of each of the one or more ninth waveforms and acquires a total duration by adding up the durations of the one or more ninth waveforms. The parameter acquiring unit 131 acquires the number of non-REM epochs.
[0134] Next, the parameter acquisition unit 131 acquires a 34th parameter, which increases as the total time length increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the 34th parameter using, for example, a 34th arithmetic expression, which is an increasing function with the total time length as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the 34th arithmetic expression is, for example, "total time length / number of non-REM epochs." In this case, the 34th parameter is the average time length of the 34th waveform per one non-REM epoch. (35) 35th Parameter
[0135] The parameter acquiring unit 131 detects one or more thirty-fifth waveforms. Then, the parameter acquiring unit 131 acquires the duration of each of the one or more thirty-fifth waveforms and acquires a total duration by adding up the durations of the one or more thirty-fifth waveforms. The parameter acquiring unit 131 acquires the number of non-REM epochs.
[0136] The thirty-fifth waveform is a waveform having a thirty-tenth specific frequency and a ninth specific amplitude in the non-REM epoch.
[0137] Next, the parameter acquisition unit 131 acquires a 35th parameter, which increases as the total time length increases and as the number of non-REM epochs decreases. The parameter acquisition unit 131 acquires the 35th parameter using, for example, a 35th arithmetic expression, which is an increasing function with the total time length as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the 35th arithmetic expression is, for example, "total time length / number of non-REM epochs." In this case, the 35th parameter is the average time length of the 35th waveform per one non-REM epoch. (36) 36th Parameter
[0138] The parameter acquiring unit 131 detects one or more thirty-fifth waveforms and acquires the number of the thirty-fifth waveforms. The parameter acquiring unit 131 acquires the number of non-REM epochs.
[0139] Next, the parameter acquiring unit 131 acquires a 36th parameter, which increases as the number of the 35th waveforms increases and increases as the number of non-REM epochs decreases. The parameter acquiring unit 131 acquires the 36th parameter using, for example, a second arithmetic expression that is an increasing function with the number of the 35th waveforms as a parameter and a decreasing function with the number of non-REM epochs as a parameter. Note that the 36th arithmetic expression is, for example, "number of 35th waveforms / number of non-REM epochs." In this case, the 36th parameter is the average number of 35th waveforms in the non-REM epochs.
[0140] The determination unit 132 obtains a determination result using one or more parameters obtained by the parameter acquisition unit 131 and the learning information in the storage unit 11. The determination result is information indicating whether or not the person whose EEG information has been received by the reception unit 12 has a mental illness, or information indicating the degree of the mental illness. The determination result is, for example, information indicating whether or not the patient whose EEG information has been received by the reception unit 12 is depressed, or information indicating the degree of depression. The determination result is, for example, "1 for depression" or "0 for not depression." The determination result may have a score. The score can be said to be the certainty, likelihood, or probability of the determination result.
[0141] The learning information is information based on two or more pieces of training data, each of which has one or more positive examples having one or more parameters that indicate characteristics of the EEG information of patients with depression and one or more negative examples having one or more parameters that indicate characteristics of the EEG information of patients without depression. The learning information is stored in the storage unit 11.
[0142] It is highly preferable that the determination unit 132 obtains a seventh parameter (parameter (7)) from the electroencephalogram information received by the reception unit 12 and obtains a determination result using the seventh parameter.
[0143] In addition, it is preferable that the determination unit 132 obtains one or more parameters from among the parameters (2), (8), (9), and (15), and obtains a determination result using the one or more parameters.
[0144] In addition, it is preferable that the judgment unit 132 obtains the judgment result using one or more parameters from (1), (3), (4), (5), (6), (10), (11), (12), (13), (14), (16), (17), (18), (19), (20), and (21).
[0145] The determination unit 132 may also obtain the determination result using one or more parameters from (22) to (36).
[0146] Furthermore, the determination unit 132 may obtain the determination result using all of the parameters (1) to (36).
[0147] The determination unit 132 obtains the determination result by, for example, any one of the following methods (A), (B), (C), and (D): (A) Machine Learning Method
[0148] The determination unit 132 performs machine learning prediction processing using one or more parameters and the learning model acquired by the parameter acquisition unit 131, and acquires a determination result.
[0149] More specifically, the determination unit 132 provides one or more parameters and a learning model to a machine learning prediction module, executes the prediction module, and obtains a determination result.
[0150] The machine learning algorithm may be any of deep learning, random forest, decision tree, SVM, etc. For machine learning, various machine learning functions such as the TensorFlow (registered trademark) library, the random forest module of the R language, and TinySVM, as well as various existing libraries, can be used. (2) Method Using Center of Gravity Information
[0151] The determination unit 132 obtains a determination result using a vector having elements each consisting of two or more parameters obtained by the parameter obtaining unit 131, and the centroid information of the positive examples and the centroid information of the negative examples.
[0152] More specifically, the determination unit 132 obtains a vector whose elements are the two or more parameters obtained by the parameter obtaining unit 131. Next, the determination unit 132 obtains a positive example distance, which is the distance between the vector and centroid information of a positive example. Furthermore, the determination unit 132 obtains a negative example distance, which is the distance between the vector and centroid information of a negative example. Next, the determination unit 132 obtains a determination result using the positive example distance and the negative example distance. Typically, the determination unit 132 obtains a determination result of "not depressed" if "positive example distance < negative example distance," and obtains a determination result of "depressed" if "positive example distance > negative example distance." Furthermore, the determination unit 132 may obtain a degree of depression (such as a score) using the difference between the positive example distance and the negative example distance.
[0153] The centroid information of a positive example is a vector whose elements are two or more parameters contained in the training data of a positive example. The centroid information of a negative example is a vector whose elements are two or more parameters contained in the training data of a negative example. The centroid information of a positive example can be called a positive example centroid vector, and the centroid information of a negative example can be called a negative centroid vector. (3) Method using a correspondence table
[0154] The determination unit 132 constructs a vector whose elements are one or more parameters acquired by the parameter acquisition unit 131. Next, the determination unit 132 acquires the similarity with the vectors contained in two or more pieces of correspondence information in the correspondence table. The vectors contained in the correspondence information are positive example vectors or negative example vectors. Next, the determination unit 132 acquires a determination result corresponding to one or more vectors whose similarity satisfies the adoption conditions. The adoption conditions include the similarity being maximum, the similarity being equal to or greater than a threshold, and the similarity ranking being equal to or greater than a threshold.
[0155] When the determination unit 132 acquires a determination result corresponding to one vector that satisfies the adoption condition, the determination unit 132 acquires the determination result as the final determination result. When the determination unit 132 acquires determination results corresponding to two or more vectors that satisfy the adoption condition, the determination unit 132 acquires the final determination result using the two or more determination results. The determination unit 132 acquires the final determination result using the two or more determination results, for example, by majority vote. (4) Method Using an Arithmetic Expression
[0156] The determination unit 132 substitutes one or more parameters acquired by the parameter acquisition unit 131 into an arithmetic expression to acquire a score, and acquires a determination result corresponding to the score. For example, the determination unit 132 acquires a determination result of "1, indicating depression" when the score is equal to or greater than a threshold, and acquires a determination result of "0, indicating not depression" when the score is less than or equal to the threshold.
[0157] The output unit 14 outputs the determination result acquired by the determination unit 132. The output unit 14 transmits the determination result to, for example, the terminal device 3. The output unit 14 displays the determination result on a display, for example.
[0158] Here, output means, for example, transmission to terminal device 3, but it may also be a concept that includes display on a display, projection using a projector, printing on a printer, sound output, storage on a recording medium, and handing over processing results to other processing devices or other programs.
[0159] The storage unit 11 is preferably a non-volatile recording medium, but may also be a volatile recording medium.
[0160] There is no restriction on the process by which information is stored in the storage unit 11. For example, information may be stored in the storage unit 11 via a recording medium, information transmitted via a communication line or the like may be stored in the storage unit 11, or information input via an input device may be stored in the storage unit 11.
[0161] The reception unit 12 is preferably realized by a wireless or wired communication means, but may also be realized by a means for receiving broadcasts, a device driver for an input means such as a touch panel or keyboard, or control software for a menu screen.
[0162] The processing unit 13, parameter acquisition unit 131, and determination unit 132 can typically be realized by a processor, memory, etc. The processing procedures of the processing unit 13, etc. are typically realized by software, and the software is recorded on a recording medium such as a ROM. However, the processing unit 13, etc. may also be realized by hardware (dedicated circuitry). The processor may be a CPU, MPU, GPU, etc., and the type does not matter.
[0163] The output unit 14 is preferably realized by a wireless or wired communication means, but may also be realized by an output device such as a display or speaker.
[0164] Next, an example of the operation of the determination device 1 will be described with reference to the flowchart of FIG.
[0165] (Step S301) The receiving unit 12 determines whether or not electroencephalogram information has been received. If electroencephalogram information has been received, the process proceeds to step S302, and if electroencephalogram information has not been received, the process returns to step S301.
[0166] (Step S302) The parameter acquiring unit 131 acquires one or more parameters from the electroencephalogram information received in step S301. An example of such parameter acquiring processing will be described with reference to the flowchart of FIG.
[0167] (Step S303) The determining unit 132 acquires the learning information from the storage unit 11.
[0168] (Step S304) The determination unit 132 performs a determination process. That is, the determination unit 132 obtains a determination result using the one or more parameters obtained in step S302 and the learning information obtained in step S303.
[0169] (Step S305) The output unit 14 outputs the determination result acquired in step S304. The process returns to step S301.
[0170] In the flowchart of FIG. 3, the process ends when the power is turned off or an interrupt occurs to end the process.
[0171] Next, an example of the parameter acquisition process in step S302 will be described with reference to the flowchart of FIG.
[0172] (Step S401) The parameter acquiring unit 131 acquires the received electroencephalogram information.
[0173] (Step S402) The parameter acquiring unit 131 assigns 1 to a counter i.
[0174] (Step S403) The parameter acquisition unit 131 determines whether the i-th epoch is present in the electroencephalogram information acquired in step S401. If the i-th epoch is present, the process proceeds to step S404. If the i-th epoch is not present, the process proceeds to step S407. An epoch is, for example, 30 seconds of electroencephalogram information.
[0175] (Step S404) The parameter acquiring unit 131 acquires state information for the i-th epoch using the electroencephalogram information corresponding to the i-th epoch, and associates the state information with the i-th epoch. Note that the state information is information indicating the state of sleep. The state information is, for example, "non-REM sleep" or "REM sleep."
[0176] (Step S405) The parameter acquiring unit 131 acquires one or more epoch attribute values of the i-th epoch using the electroencephalogram information corresponding to the i-th epoch.
[0177] (Step S406) The parameter acquiring unit 131 increments the counter i by 1. The process returns to step S403.
[0178] (Step S407) The parameter acquiring unit 131 assigns 1 to a counter j.
[0179] (Step S408) The parameter acquisition unit 131 determines whether or not the jth parameter to be acquired exists. If the jth parameter exists, the process proceeds to step S409. If the jth parameter does not exist, the process returns to the upper level process. It is preferable that the parameters to be acquired are determined in advance.
[0180] (Step S409) The parameter acquiring unit 131 assigns 1 to a counter k.
[0181] (Step S410) The parameter acquisition unit 131 determines whether the kth raw information to be used when acquiring the jth parameter exists. If the kth raw information exists, the process proceeds to step S411. If the kth raw information does not exist, the process proceeds to step S413.
[0182] (Step S411) The parameter acquiring unit 131 acquires the k-th piece of raw information. The parameter acquiring unit 131 acquires the k-th piece of raw information, for example, by using one or more epoch attribute values acquired in step S405.
[0183] (Step S412) The parameter acquiring unit 131 increments the counter k by 1. The process returns to step S410.
[0184] (Step S413) The parameter acquiring unit 131 acquires the j-th parameter using the one or more pieces of raw information acquired in step S411, and temporarily stores the parameter in a buffer (not shown).
[0185] (Step S414) The parameter acquiring unit 131 increments the counter j by 1. The process returns to step S408.
[0186] A specific example of the determination device 1 in this embodiment will be described below. The parameter acquisition unit 131 of the determination device 1 acquires, for example, 36 parameters shown in the parameter management tables of Figs. 5 to 8. The parameter management table has 36 records each having an "ID" and a "parameter description." The "ID" is the ID of the parameter. The "parameter description" is information indicating the content of the parameter.
[0187] The determination unit 132 then obtains a determination result using one or more parameters and the learning information from among the 36 parameters obtained by the parameter obtaining unit 131. Next, the output unit 14 outputs the determination result.
[0188] As described above, according to this embodiment, depression can be diagnosed with high accuracy using electroencephalogram information.
[0189] Furthermore, according to this embodiment, depression can be diagnosed with higher accuracy using useful parameters that can be acquired from electroencephalogram information.
[0190] (Embodiment 2) In this embodiment, a description will be given of a learning device for acquiring learning information used for determination by the determination device 1. Note that in this embodiment, the description will be given of a case where the learning information is, for example, a learning model, center of gravity information, a correspondence table, or an arithmetic expression acquired by a machine learning learning process.
[0191] 9 is a block diagram of the learning device 2. The learning device 2 includes a learning storage unit 21, a parameter acquisition unit 131, a learning unit 22, and an accumulation unit 23.
[0192] The learning storage unit 21 constituting the learning device 2 stores various types of information. The various types of information include, for example, two or more pieces of EEG information, two or more pieces of teacher data, and learning information. The EEG information in the learning storage unit 21 is associated with a determination result indicating whether or not the subject (patient, etc.) of the EEG information is depressed. The teacher data in the learning storage unit 21 is associated with a determination result indicating whether or not the subject of the teacher data is depressed. Note that this determination result is the result of a person's (usually a doctor's) judgment. Furthermore, the teacher data is, for example, a vector whose elements are two or more parameters obtained from the EEG information. The teacher data is, for example, a vector whose elements are 36 parameters obtained from the EEG information.
[0193] When EEG information exists in the learning storage unit 21 but teacher data does not exist, the parameter acquisition unit 131 acquires one or more parameters for each of two or more pieces of EEG information by the operation described in Fig. 4, composes teacher data having the one or more parameters and the determination result, and stores the teacher data in the learning storage unit 21. It is preferable that the two or more teacher data are one or more positive examples and one or more negative examples.
[0194] The learning unit 22 acquires learning information using two or more pieces of teacher data stored in the learning storage unit 21. The learning unit 22 acquires learning information by performing, for example, any one of the following processes (A), (B), (C), and (D): (A) When the learning information is a learning model
[0195] The learning unit 22 performs machine learning learning processing using two or more pieces of teacher data in the learning storage unit 21 to acquire a learning model. The machine learning algorithm may be deep learning, random forest, decision tree, SVM, or the like. For machine learning, various machine learning functions such as the TensorFlow (registered trademark) library, the random forest module of the R language, and TinySVM, as well as various existing libraries, may be used. (B) When the learning information is centroid information
[0196] The learning unit 22 acquires one or more positive examples from the two or more pieces of training data stored in the learning storage unit 21. Here, a positive example is a vector whose elements are two or more parameters acquired from the electroencephalogram information of a patient suffering from depression. The learning unit 22 acquires centroid vectors of the one or more positive examples and associates them with information indicating that they are positive example vectors. Such vectors are positive example centroid vectors.
[0197] The learning unit 22 acquires one or more negative examples from the two or more pieces of teacher data stored in the learning storage unit 21. Here, the negative examples are vectors whose elements are two or more parameters acquired from the electroencephalogram information of patients with depression. The learning unit 22 acquires centroid vectors of the one or more negative examples and associates them with information indicating that the negative example vectors are negative example vectors. This vector is a negative example centroid vector. (C) When the learning information is a correspondence table: (C-1) The learning unit 22 acquires, for each of two or more pieces of teacher data, correspondence information having a vector whose elements are one or more parameters contained in the teacher data and a determination result. The learning unit 22 constructs a correspondence table having two or more pieces of correspondence information. (C-2) The learning unit 22 acquires an average vector of vectors whose elements are one or more parameters that pair with the determination result of "being depressed," and acquires correspondence information having the average vector and the determination result of "being depressed." The learning unit 22 acquires an average vector of vectors whose elements are one or more parameters that are paired with the "not depressed" judgment result, and acquires correspondence information having the average vector and the "not depressed" judgment result. Next, the learning unit 22 constructs a correspondence table having two pieces of correspondence information. (D) When the learning information is an arithmetic formula
[0198] The learning unit 22 performs multiple regression analysis on each of two or more pieces of teacher data, for example, to obtain an arithmetic formula. More specifically, the learning unit 22 obtains an arithmetic formula that takes one or more parameters contained in each of the two or more pieces of teacher data as input and outputs a judgment result (usually a score in this case).
[0199] The accumulation unit 23 accumulates the learning information acquired by the learning unit 22. For example, the accumulation unit 23 accumulates the learning information acquired by the learning unit 22 in the learning storage unit 21 or the storage unit 11 of the determination device 1. It does not matter where the accumulation unit 23 accumulates the learning information.
[0200] The learning storage unit 21 is preferably a non-volatile recording medium, but can also be realized as a volatile recording medium.
[0201] There is no restriction on the process by which information is stored in the learning storage unit 21. For example, information may be stored in the learning storage unit 21 via a recording medium, information transmitted via a communication line or the like may be stored in the learning storage unit 21, or information input via an input device may be stored in the learning storage unit 21.
[0202] The learning unit 22 and the storage unit 23 can usually be realized by a processor, memory, etc. The processing procedures of the learning unit 22, etc. are usually realized by software, and the software is recorded on a recording medium such as a ROM. However, the learning unit 22, etc. may also be realized by hardware (dedicated circuit). The processor may be a CPU, MPU, GPU, etc., and the type does not matter.
[0203] Next, an example of the operation of the learning device 2 will be described with reference to the flowchart of FIG.
[0204] (Step S1001) The parameter acquiring unit 131 assigns 1 to a counter i.
[0205] (Step S1002) The parameter acquisition unit 131 determines whether or not the i-th electroencephalogram information exists in the learning storage unit 21. If the i-th electroencephalogram information exists, the process proceeds to step S1003; if the i-th electroencephalogram information does not exist, the process proceeds to step S1007.
[0206] (Step S1003) The parameter acquiring section 131 acquires one or more parameters from the i-th electroencephalogram information. An example of such parameter acquiring processing has been described with reference to the flowchart of FIG.
[0207] (Step S1004) The learning unit 22 acquires the determination result paired with the i-th electroencephalogram information.
[0208] (Step S1005) The learning unit 22 constructs teacher data using one or more parameters acquired in step S1003 as explanatory variables and the judgment result acquired in step S1004 as the objective variable, and stores the data in the learning storage unit 21.
[0209] (Step S1006) The parameter acquiring unit 131 increments the counter i by 1. The process returns to step S1002.
[0210] (Step S1007) The learning unit 22 acquires learning information using the two or more pieces of teacher data accumulated in step S1005.
[0211] (Step S1008) The storage unit 23 stores the learning information acquired in step S1007, and the process ends.
[0212] As described above, according to this embodiment, learning information for accurately diagnosing depression using electroencephalogram information can be constructed.
[0213] The processing in this embodiment may be implemented by software. This software may be distributed by software download or the like. This software may also be recorded on a recording medium such as a CD-ROM and distributed. The software implementing the determination device 1 in this embodiment is the following program. That is, this program causes a computer to function as: a reception unit that receives electroencephalogram information; a parameter acquisition unit that acquires one or more parameters from the electroencephalogram information; a determination unit that references a storage unit that stores learning information based on two or more sets of teacher data, each set having one or more positive examples with one or more parameters that indicate characteristics of the electroencephalogram information of a patient with depression and one or more negative examples with one or more parameters that indicate characteristics of the electroencephalogram information of a patient without depression, and that uses the one or more parameters acquired by the parameter acquisition unit and the learning information to obtain a determination result indicating whether or not the patient associated with the electroencephalogram information received by the reception unit is depressed; and an output unit that outputs the determination result.
[0214] Furthermore, the software that realizes the learning device 2 in this embodiment is the following program: That is, this program causes a computer to function as a parameter acquisition unit that acquires one or more parameters from two or more pieces of electroencephalogram information associated with the determination result of whether or not the patient is depressed, a learning unit that acquires, for each of the two or more pieces of electroencephalogram information, teacher data having the one or more parameters acquired by the parameter acquisition unit and the determination result associated with each piece of electroencephalogram information, and acquires learning information using the two or more pieces of teacher data, and a storage unit that stores the learning information acquired by the learning unit.
[0215] 11 shows the appearance of a computer that executes the programs described in this specification to realize the determination device 1, learning device 2, and the like of the various embodiments described above. The above-described embodiments can be realized by computer hardware and a computer program executed thereon. FIG. 11 is an overview of this computer system 300, and FIG. 12 is a block diagram of the system 300.
[0216] In FIG. 11, a computer system 300 includes a computer 301 including a CD-ROM drive, a keyboard 302, a mouse 303, and a monitor 304.
[0217] 12, in addition to a CD-ROM drive 3012, the computer 301 includes an MPU 3013, a bus 3014 connected to the CD-ROM drive 3012 etc., a ROM 3015 for storing programs such as a boot-up program, a RAM 3016 connected to the MPU 3013 for temporarily storing instructions for application programs and providing temporary storage space, and a hard disk 3017 for storing application programs, system programs, and data. Although not shown here, the computer 301 may further include a network card for providing connection to a LAN.
[0218] A program that causes computer system 300 to execute the functions of the determination device and the like of the above-described embodiment may be stored on CD-ROM 3101, inserted into CD-ROM drive 3012, and then transferred to hard disk 3017. Alternatively, the program may be transmitted to computer 301 via a network (not shown) and stored on hard disk 3017. The program is loaded into RAM 3016 when executed. The program may also be loaded directly from CD-ROM 3101 or the network.
[0219] The program does not necessarily include an operating system (OS) or a third-party program that causes the computer 301 to execute the functions of the determination device and the like of the above-described embodiment. The program only needs to include instructions that call appropriate functions (modules) in a controlled manner to achieve the desired results. How the computer system 300 operates is well known, and a detailed description thereof will be omitted.
[0220] The computer that executes the program may be a single computer or a plurality of computers, that is, it may perform centralized processing or distributed processing.
[0221] Furthermore, in each of the above embodiments, it goes without saying that two or more communication means present in one device may be physically realized by one medium.
[0222] Furthermore, in each of the above embodiments, each process may be realized by centralized processing in a single device, or may be realized by distributed processing in a plurality of devices.
[0223] The present invention is not limited to the above-described embodiment, and various modifications are possible, and it goes without saying that these modifications are also included within the scope of the present invention.
[0224] As described above, the determination device 1 according to the present invention has the effect of being able to accurately determine depression using electroencephalogram information, and is useful as a determination device for determining depression, etc.
Claims
1. A judgment device comprising: a reception unit that receives electroencephalogram information; a parameter acquisition unit that acquires one or more parameters from the electroencephalogram information; a judgment unit that refers to a storage unit in which learning information based on two or more teacher data having one or more positive examples having one or more parameters that indicate characteristics of the electroencephalogram information of persons with a mental illness and one or more negative examples having one or more parameters that indicate characteristics of the electroencephalogram information of persons without a mental illness is stored, and uses the one or more parameters acquired by the parameter acquisition unit and the learning information to acquire a judgment result indicating whether or not the patient of the electroencephalogram information received by the reception unit has a mental illness; and an output unit that outputs the judgment result.
2. The parameter acquisition unit acquires one or more parameters including a non-REM representative power value from the EEG information, and the judgment unit uses the one or more parameters and the learning information to acquire a judgment result indicating whether or not the patient of the EEG information accepted by the reception unit has a mental illness, and the parameter acquisition unit acquires the non-REM representative power value, which is a value acquired using the overall power value of frequencies in epochs judged to be non-REM sleep in the EEG information and the number of epochs, and which becomes larger the greater the power value of the frequency and becomes smaller the greater the number of epochs.
3. The determination device according to claim 2, wherein the parameter acquisition unit adds up power values of frequencies between approximately 8 Hz and approximately 14 Hz calculated by fast Fourier transform in epochs determined to be non-REM sleep to obtain an overall power value, and divides the overall power value by the number of non-REM epochs to obtain the non-REM representative power value which is the average power value per epoch.
4. The learning information is a learning model obtained by performing a machine learning learning process on the two or more teacher data, and the judgment unit performs a machine learning prediction process using the one or more parameters and the learning model to obtain the judgment result. A judgment device as described in any one of claims 1 to 3.
5. The positive example is a positive example vector having as its elements two or more parameters which indicate characteristics of the electroencephalogram information of a person with a mental illness, and the negative example is a negative example vector having as its elements two or more parameters which indicate characteristics of the electroencephalogram information of a person without a mental illness, and the determination unit obtains a positive example centroid vector which is a vector of the centroids of one or more positive example vectors, obtains a negative example centroid vector which is a vector of the centroids of one or more negative example vectors, obtains a positive example distance which is the distance between the vector having as its elements two or more parameters obtained by the parameter acquisition unit and the positive example centroid vector, obtains a negative example distance which is the distance between the vector having as its elements two or more parameters obtained by the parameter acquisition unit and the negative example centroid vector, and obtains the determination result using the positive example distance and the negative example distance.
6. The positive example is a positive example vector whose elements are one or more parameters that indicate characteristics of EEG information of persons with a mental illness, the negative example is a negative example vector whose elements are one or more parameters that indicate characteristics of EEG information of persons without a mental illness, the learning information is a correspondence table having correspondence information having one or more positive example vectors and correspondence information having one or more negative example vectors, and the judgment unit constructs a vector whose elements are the one or more parameters acquired by the parameter acquisition unit, acquires a similarity between the vector and each of the one or more positive example vectors and a similarity between the vector and each of the one or more negative example vectors, and acquires a judgment result corresponding to the positive example vector or the negative example vector whose similarity satisfies an adoption condition.
7. A judgment device as claimed in any one of claims 1 to 3, wherein the positive example is a positive example vector whose elements are one or more parameters which indicate characteristics of EEG information of persons with a mental illness, the negative example is a negative example vector whose elements are one or more parameters which indicate characteristics of EEG information of persons without a mental illness, the learning information is an arithmetic formula which outputs a positive example value which is a value corresponding to a positive example for the positive example vector, and outputs a negative example value which is a value corresponding to a negative example for the negative example vector, and the judgment unit substitutes the one or more parameters acquired by the parameter acquisition unit into the arithmetic formula to acquire a score, and acquires a judgment result corresponding to the score.
8. A learning device comprising: a parameter acquisition unit that acquires one or more parameters from two or more pieces of electroencephalogram information corresponding to a judgment result of whether or not a subject has a mental illness; a learning unit that acquires, for each of the two or more pieces of electroencephalogram information, teacher data having one or more parameters acquired by the parameter acquisition unit and a judgment result corresponding to each piece of electroencephalogram information, and acquires learning information using the two or more pieces of teacher data; and a storage unit that stores the learning information acquired by the learning unit.
9. A judgment method realized by a reception unit, a parameter acquisition unit, a judgment unit, and an output unit, comprising: a reception step in which the reception unit receives electroencephalogram information; a parameter acquisition step in which the parameter acquisition unit acquires one or more parameters from the electroencephalogram information; a judgment step in which the judgment unit refers to a storage unit in which learning information based on two or more teacher data having one or more positive examples having one or more parameters that indicate characteristics of the electroencephalogram information of a person with a mental illness and one or more negative examples having one or more parameters that indicate characteristics of the electroencephalogram information of a person without a mental illness is stored, and the one or more parameters acquired by the parameter acquisition unit and the learning information are used to acquire a judgment result indicating whether or not the patient of the electroencephalogram information accepted by the reception unit has a mental illness; and an output step in which the output unit outputs the judgment result.
10. A method for producing learning information realized by a parameter acquisition unit, a learning unit, and an accumulation unit, comprising: a parameter acquisition step in which the parameter acquisition unit acquires one or more parameters from two or more pieces of electroencephalogram information corresponding to a judgment result of whether or not a subject has a mental illness; a learning step in which the learning unit acquires teacher data for each of the two or more pieces of electroencephalogram information, the teacher data having one or more parameters acquired in the parameter acquisition step and a judgment result corresponding to each piece of electroencephalogram information, and acquires learning information using the two or more teacher data; and a storage step in which the storage unit stores the learning information acquired in the learning step.
Citation Information
Patent Citations
Inference model generation method and discrimination device
JP2024033833A
Organism light measuring device
WO2005025421A1
Mental illness analysis device, mental illness analysis method, and program
WO2012176790A1
