Pain classification and instantaneous pain discrimination using sparse modeling

Abstract

The present invention provides a method for discriminating or classifying pain experienced by an estimated subject based on the subject's brain waves. The method comprises: a) stimulating the estimated subject with multiple levels of stimulation intensity; b) obtaining brain wave data or analysis data corresponding to the stimulation intensity; c) extracting brain wave features from the brain wave data or analysis data; and d) introducing the features into a sparse model analysis to approximate at least one of the quantitative and qualitative levels of pain, thereby estimating or discriminating the pain level. The present invention also provides a method for discriminating or evaluating pain, comprising the steps of comparing all or a portion of the brain wave data or analysis data for a period of 2000 milliseconds from the earliest time point among the evoked brain wave component, the initial event-related potential component, and 250 milliseconds after the subject is stimulated, with brain wave data or analysis data after the same time period has elapsed since the application of a reference stimulus.

Description

Technical Field

[0001] The present invention relates to a technique for analyzing biological signals such as brain waves obtained from a subject using sparse modeling, thereby classifying the quality and quantity of pain using a small amount of information. More specifically, the present invention relates to objectively classifying or discriminating pain levels (e.g., mild pain, severe pain, etc.) that vary among individuals.

[0002] The present invention also relates to a technique for determining instantaneous pain using brain waves. More specifically, it relates to a technique for determining whether instantaneous pain has occurred by determining a signal after a specific time has passed since stimulation. Background Art

[0003] Pain is essentially subjective, but it is desirable to evaluate it objectively for treatment. It is common for patients to suffer disadvantages due to underestimation of pain. Therefore, a method has been proposed to objectively estimate pain using brain waves (for example, see Patent Document 1). However, the intensity of pain is subjective and difficult to evaluate objectively. Whether it is extremely unbearable pain or pain that can be tolerated to a certain extent, it is impossible to express it with the subjective expression "pain" alone, and personal expressions are also diverse, so it is difficult to evaluate objectively. However, when observing the effect of treatment, it is desirable to classify pain, but no such technology has yet been provided.

[0004] Furthermore, various sensations are often expressed using vectors in one direction. For example, when observing pain, it is often distinguished as pain or no pain. Discerning instantaneous pain is difficult.

[0005] Prior art literature

[0006] Patent Literature

[0007] Patent Document 1: Japanese Patent Application No. 2010-523226 Summary of the Invention

[0008] Means for solving problems

[0009] The present invention provides a pain estimation method and apparatus that can objectively and accurately estimate the pain experienced by an estimation subject using sparse modeling, and further can easily classify the pain into qualitative and quantitative values. The present invention also provides a technique for generating pain classification values ​​for such pain classification.

[0010] Furthermore, the inventors of the present invention have conducted intensive research and, as a result, discovered a discrimination technique capable of discriminating instantaneous pain.

[0011] The present invention provides, for example, the following solutions.

[0012] (1) A method for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, comprising the following steps:

[0013] a) obtaining model electroencephalogram data corresponding to the stimulation intensity for the model or its analysis data;

[0014] b) extracting model-use electroencephalogram feature quantities from the electroencephalogram data or its analysis data;

[0015] c) setting a target pain level, subjecting the model's electroencephalogram features and the pain level to sparse model analysis, determining an appropriate λ value, and determining parameters (partial regression coefficients) of the model's electroencephalogram features and an algorithm constant (intercept) corresponding to the appropriate λ to generate a regression model;

[0016] d) obtaining the measurement electroencephalogram data or analysis data of the estimated subject;

[0017] e) extracting a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0018] f) fitting the measurement brain wave feature value to a regression model to calculate the corresponding pain level; and

[0019] g) Steps to display the pain level as needed.

[0020] (2) A method comprising the following steps:

[0021] c) providing a sparse model analytical regression model for estimating the subject's pain level;

[0022] d) obtaining the measurement electroencephalogram data or analysis data of the estimated subject;

[0023] e) extracting a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0024] f) fitting the measurement brain wave feature value to a regression model to calculate the corresponding pain level; and

[0025] g) Steps to display the pain level as needed.

[0026] (3) A method for generating a regression model for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the method comprising the following steps:

[0027] a) obtaining model electroencephalogram data corresponding to the stimulation intensity for the model or its analysis data;

[0028] b) extracting model-use electroencephalogram feature quantities from the electroencephalogram data or its analysis data;

[0029] c) Set the target pain level, import the model's brain wave characteristics and the pain level into the sparse model analysis, calculate the appropriate λ, determine the parameters (partial regression coefficients) of the model's brain wave characteristics and the algorithm's constant (intercept) corresponding to the appropriate λ, and generate a regression model.

[0030] (4) The method according to any one of items 1 to 3, wherein

[0031] The regression model was validated by split cross validation.

[0032] (5) The method according to item 4, wherein the split cross validation is a ten-fold cross validation.

[0033] (6) The method according to any one of items 1 to 5, wherein the model electroencephalogram data is electroencephalogram data from an estimation subject.

[0034] (7) The method according to any one of items 1 to 5, wherein the model electroencephalogram data is electroencephalogram data from a subject different from the estimation subject.

[0035] (8) The method according to item 7, wherein the regression model is further calibrated with respect to the estimation object after being generated.

[0036] (9) The method according to any one of items 1 to 8, wherein the pain level includes at least two types and / or at least two patterns of pain.

[0037] (10) The method according to any one of items 1 to 9, wherein the regression model is generated for a plurality of subjects.

[0038] (11) According to the method of item 1, there are at least two types of the electroencephalogram feature quantities.

[0039] (12) A device for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, comprising:

[0040] A) a model data acquisition unit that acquires model electroencephalogram data corresponding to the stimulation intensity for the model or analysis data thereof;

[0041] B) a model feature extraction unit that extracts model-use electroencephalogram feature values ​​from the electroencephalogram data or its analysis data;

[0042] C) a regression model generation unit that sets a target pain level, introduces the model electroencephalogram feature quantity and the pain level into sparse model analysis, finds an appropriate λ, determines parameters (partial regression coefficients) of the model electroencephalogram feature quantity and an algorithm constant (intercept) corresponding to the appropriate λ, and generates a regression model;

[0043] D) a measurement data acquisition unit that acquires (measurement-use) electroencephalogram data of the estimated subject or analysis data thereof;

[0044] E) a measurement feature value extraction unit that extracts a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0045] F) a pain level calculation unit that fits the measurement electroencephalogram feature value to a regression model to calculate a corresponding pain level; and

[0046] G) A pain level display section that displays the pain level as needed.

[0047] (13) A device for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, comprising:

[0048] C) a regression model providing unit that provides a regression model for estimating the pain level of the subject using sparse model analysis;

[0049] D) a measurement data acquisition unit that acquires (measurement-use) electroencephalogram data of the estimated subject or analysis data thereof;

[0050] E) a measurement feature value extraction unit that extracts a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0051] F) a pain level calculation unit that fits the measurement electroencephalogram feature value to a regression model to calculate a corresponding pain level; and

[0052] G) A pain level display section that displays the pain level as needed.

[0053] (14) A device for generating a regression model for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the device comprising:

[0054] A) a model data acquisition unit that acquires model electroencephalogram data corresponding to the stimulation intensity for the model or analysis data thereof;

[0055] B) a feature extraction unit that extracts a model-use electroencephalogram feature from the electroencephalogram data or its analysis data; and

[0056] C) A regression model generation unit, which sets the pain level as a target, imports the model's brain wave characteristics and the pain level into a sparse model analysis, calculates an appropriate λ, determines the parameters (partial regression coefficients) of the model's brain wave characteristics corresponding to the appropriate λ and the algorithm's constant (intercept), and generates a regression model.

[0057] (15) The device according to any one of items 12 to 14, comprising any one or more of the features described in items 4 to 11.

[0058] (16) A program for causing a computer to execute a method for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the method comprising the following steps:

[0059] a) obtaining model electroencephalogram data corresponding to the stimulation intensity for the model or its analysis data;

[0060] b) extracting model-use electroencephalogram feature quantities from the electroencephalogram data or its analysis data;

[0061] c) setting a target pain level, subjecting the model electroencephalogram features and the pain level to sparse model analysis, determining an appropriate λ, and determining parameters (partial regression coefficients) of the model electroencephalogram features and an algorithm constant (intercept) corresponding to the appropriate λ to generate a regression model;

[0062] d) a step of obtaining (measurement) electroencephalogram data of the estimation subject or its analysis data;

[0063] e) extracting a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0064] f) fitting the measurement brain wave feature value to a regression model to calculate the corresponding pain level; and

[0065] g) Steps to display the pain level as needed.

[0066] (17) A program causing a computer to execute a method for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the method comprising the following steps:

[0067] c) providing a sparse model analytical regression model for estimating the subject's pain level;

[0068] d) a step of obtaining (measurement) electroencephalogram data of the estimation subject or its analysis data;

[0069] e) extracting a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0070] f) fitting the measurement brain wave feature value to a regression model to calculate the corresponding pain level; and

[0071] g) Steps to display the pain level as needed.

[0072] (18) A program causing a computer to execute a method for generating a regression model for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the method comprising the following steps:

[0073] a) obtaining model electroencephalogram data corresponding to the stimulation intensity for the model or its analysis data;

[0074] b) a step of extracting a model-use electroencephalogram feature quantity from the electroencephalogram data or its analysis data; and

[0075] c) Set the target pain level, import the model brain wave characteristics and the pain level into the sparse model analysis, calculate the appropriate λ, determine the parameters (partial regression coefficients) of the model brain wave characteristics and the algorithm constant (intercept) corresponding to the appropriate λ, and generate a regression model.

[0076] (19) The program according to any one of items 16 to 18, comprising any one or more of the features described in items 4 to 11.

[0077] (20) A recording medium storing a program for causing a computer to execute a method for determining or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the method comprising the following steps:

[0078] a) obtaining model electroencephalogram data corresponding to the stimulation intensity for the model or its analysis data;

[0079] b) extracting model-use electroencephalogram feature quantities from the electroencephalogram data or its analysis data;

[0080] c) setting a target pain level, subjecting the model electroencephalogram features and the pain level to sparse model analysis, determining an appropriate λ, and determining parameters (partial regression coefficients) of the model electroencephalogram features and an algorithm constant (intercept) corresponding to the appropriate λ to generate a regression model;

[0081] d) a step of obtaining (measurement) electroencephalogram data of the estimation subject or its analysis data;

[0082] e) extracting a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0083] f) fitting the measurement brain wave feature value to a regression model to calculate the corresponding pain level; and

[0084] g) Steps to display the pain level as needed.

[0085] (21) A recording medium storing a program for causing a computer to execute a method for determining or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the method comprising the following steps:

[0086] c) providing a sparse model analytical regression model for estimating the subject's pain level;

[0087] d) a step of obtaining (measurement) electroencephalogram data of the estimation subject or its analysis data;

[0088] e) extracting a measurement electroencephalogram feature value from the measurement electroencephalogram data or its analysis data;

[0089] f) fitting the measurement brain wave feature value to a regression model to calculate the corresponding pain level; and

[0090] g) Steps to display the pain level as needed.

[0091] (22) A recording medium storing a program for causing a computer to execute a method for generating a regression model for discriminating or classifying pain experienced by an estimated subject based on brain waves of the estimated subject, the method comprising the following steps:

[0092] a) obtaining model electroencephalogram data corresponding to the stimulation intensity for the model or its analysis data;

[0093] b) a step of extracting a model-use electroencephalogram feature quantity from the electroencephalogram data or its analysis data; and

[0094] c) Set the target pain level, import the model brain wave characteristics and the pain level into the sparse model analysis, calculate the appropriate λ, determine the parameters (partial regression coefficients) of the model brain wave characteristics and the algorithm constant (intercept) corresponding to the appropriate λ, and generate a regression model.

[0095] (23) The recording medium according to any one of items 20 to 22, comprising any one or more of the features described in items 4 to 11.

[0096] Furthermore, the present invention also provides, for example, the following solutions.

[0097] (A1) A method for identifying or evaluating pain, comprising the following steps:

[0098] All or part of the brain wave data or its analysis data from the induced brain wave component, the initial event-related potential component and the 2000 milliseconds earliest time point after the object stimulation is applied are compared with the brain wave data or its analysis data after the same time period after the reference stimulation is applied.

[0099] (A2) The method according to item A1 is characterized in that the judgment criterion includes whether there is a persistence characteristic in the electroencephalogram data during the period of 2000 milliseconds from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component and the 250 milliseconds.

[0100] (A3) The method according to item A1 or A2, wherein the brain wave data or the analysis data thereof includes all or part of the brain wave data or the analysis data thereof within a range of 2000 milliseconds from the mid-term time period.

[0101] (A4) The method of item A3, wherein the intermediate time period includes values ​​ranging from 250 milliseconds to 600 milliseconds.

[0102] (A5) The method according to any one of items A1 to A4, wherein the whole or part includes a range of at least 100 milliseconds in length.

[0103] (A6) The method according to item A5, wherein persistence is determined to be observed when a statistically significant difference is observed between the range of at least 100 milliseconds and the range of at least 100 milliseconds.

[0104] (A7) The method according to any one of items A1 to A6, wherein

[0105] The comparison includes:

[0106] determining whether there is a duration for which the value of the electroencephalogram data or the analysis data thereof obtained by the target stimulation differs from the value of the electroencephalogram data or the analysis data thereof obtained by the reference stimulation, and determining whether, if there is a duration for which the difference exists, the values ​​return from being different to being the same; and

[0107] If the duration lasts for this period and does not return to the same value, it is determined that unpleasant pain exists.

[0108] (A8) The method according to any one of items A1 to A7, wherein the electroencephalogram data or analysis data thereof is potential, duration, or a combination thereof.

[0109] (A9) The method according to any one of items A1 to A8, wherein the determination of pain is determination of discomfort of pain.

[0110] (A10) The method according to any one of items A1 to A9, wherein the electroencephalogram data or its analysis data are compared over the entire range from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds to 2000 milliseconds.

[0111] (A11) The method according to any one of items A1 to A10, further comprising the step of analyzing the compared data using Sigmoid function fitting.

[0112] (A12) The method according to any one of items A1 to A11, wherein the determination is made based on a positive component of electroencephalogram data or analysis data thereof.

[0113] (A13) The method according to item A12, wherein when the positive component persists even after the mid-term period, it is determined to be pain.

[0114] (A14) A program for causing a computer to execute a method for identifying or evaluating pain, the method comprising the following steps:

[0115] All or part of the brain wave data or its analysis data during the period of 2000 milliseconds from the earliest time point among the induced brain wave component, the initial event-related potential component and 250 milliseconds after the object stimulation is applied are compared with the brain wave data or its analysis data after the same time period after the reference stimulation is applied.

[0116] (A14A) The program according to item A14, further comprising one or more features of items A1 to A13.

[0117] (A15) A recording medium storing a program for causing a computer to execute a method for identifying or evaluating pain, the method comprising the following steps:

[0118] All or part of the brain wave data or its analysis data during the period of 2000 milliseconds from the earliest time point among the induced brain wave component, the initial event-related potential component and 250 milliseconds after the object stimulation is applied are compared with the brain wave data or its analysis data after the same time period after the reference stimulation is applied.

[0119] (A15A) The recording medium according to item A15, further comprising one or more features of items A1 to A13.

[0120] (A16) A system or device for identifying or evaluating pain, the system or device comprising:

[0121] an electroencephalogram data input section that inputs electroencephalogram data or analysis data thereof; and

[0122] An analysis unit compares all or part of the brain wave data or its analysis data during the period of 2000 milliseconds from the earliest time point among the induced brain wave component, the initial event-related potential component and 250 milliseconds after the application of the object stimulus with the brain wave data or its analysis data after the same time period after the application of the reference stimulus.

[0123] (A16A) The system or apparatus according to item A16, further comprising one or more features of items A1 to A13.

[0124] In the present invention, the above one or more features are intended to be provided in combination in addition to the combinations explicitly described. More embodiments and advantages of the present invention will be recognized by those skilled in the art as long as they read and understand the following detailed description as needed.

[0125] Effects of the Invention

[0126] The present invention can easily classify pain using fewer parameters. In a preferred embodiment, the obtained parameters can be used to distinguish the nature of various pains, thereby enabling classification of pain, or performing various treatments with minimal effort, or classifying treatment effects.

[0127] The present invention can also distinguish between instantaneous pain and delayed pain, enabling more detailed treatment and / or surgery that is consistent with subjective needs, and is therefore useful in the medical-related industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0128] Figure 1A Graph showing the relationship between electrical stimulation and pain rating (VAS).

[0129] Figure 1B This is a graph showing the relationship between electrical stimulation and pain level (paired comparison).

[0130] Figure 1C This is a graph showing the relationship between electrical stimulation and brain wave amplitude.

[0131] Figure 1D This is a graph showing an example of an electroencephalogram waveform.

[0132] Figure 1E This is a graph showing the relationship between the pain level (VAS) caused by electrical stimulation and the brain wave amplitude.

[0133] Figure 1F This is a graph showing the relationship between the pain level (paired comparison) caused by electrical stimulation and the brain wave amplitude.

[0134] Figure 1GThis is a graph showing the relationship between the pain level (VAS) caused by thermal stimulation and the brain wave amplitude.

[0135] Figure 2 This is a graph showing the relationship between the pain level (paired comparison) caused by thermal stimulation and the brain wave amplitude.

[0136] Figure 3 The heat stimulation experimental paradigm of Example 1 is shown. The heat stimulation was performed with a baseline temperature of 35°C and a temperature increase of 2°C per level from 40°C in Level 1 to 50°C in Level 6. Each level included three stimuli, each lasting 15 seconds.

[0137] Figure 4 This is an example of a flowchart showing the process of the present invention, and shows the process of sparse model analysis.

[0138] Figure 5 The sparse model of the characteristic quantity coefficients of Example 1 is shown. Discrimination verification was performed 1000 times (X-axis). The characteristic quantities include 24 average amplitude absolute values ​​(4) and frequency powers (20). In the 1000 discrimination accuracy verifications, the coefficient averages of the highly contributing characteristic quantities were constant and high. For example, the coefficient of the variation range of the delta band in the frontal lobe showed a value of "-2.17."

[0139] Figure 6 The figure shows the discrimination accuracy distribution of the test data of Example 1. The average discrimination accuracy of 1000 validations is approximately 80%, which is approximately 30% higher than the average discrimination accuracy when the discrimination accuracy is estimated by randomizing the pain level labels (strong or weak).

[0140] Figure 7 The following shows a sparse model of feature quantity coefficients used in the discriminant model construction of Example 2. Data from one subject (4 samples) was randomly removed from the data of 40 subjects, and the feature quantity coefficients to be used in the discriminant model were calculated using 10-fold cross-validation using the 156 sample data.

[0141] Figure 8 The figure shows the transition of the intercept value in the ten-fold cross validation when creating the discriminant model in Example 2.

[0142] Figure 9 The following shows the transition of the discrimination accuracy in the ten-fold cross validation when creating the discrimination model in Example 2.

[0143] Figure 10 The following shows a multiple regression model for pain discrimination created by ten-fold cross validation in Example 2. The pain index estimated by the discrimination model is classified into pain levels according to a threshold value (eg, 50%).

[0144] Figure 11The figure shows the estimated pain value of a certain subject calculated by the pain discrimination model of Example 2 and the pain level discrimination accuracy based on a threshold value (50%).

[0145] Figure 12 This is an example of a flowchart showing the flow of the present invention.

[0146] Figure 13 This is an example of a block diagram showing the functional configuration of the present invention.

[0147] Figure 14 This is an example of a block diagram showing the functional configuration of the present invention.

[0148] Figure 15 is shown with Figure 14 Linked graph of sparse model parsing in .

[0149] Figure 16 This is another example of a block diagram showing the functional configuration of the present invention.

[0150] Figure 17 A schematic diagram related to evoked electroencephalogram components and event-related potential components is shown, showing examples of P1, P2, N1, N2, and P3.

[0151] Figure 18 An example of a persistent signal (characteristic) of the left frontal lobe (F3) found in the present invention is shown. Usually, due to the difficulty of the subject, the peak time of the mid-term positive and late positive components such as P300 sometimes changes. In the case of non-persistence, the peak is clear and returns to the baseline (zero potential line). The persistent component of the present invention does start from around 400 milliseconds, but the effect lasts for a very long time, and the peak does not appear around 400 milliseconds, but is offset to around 1500 milliseconds. In addition, by template-ing this persistent component and regressing it online to the brain wave data, the approximation index (R 2 The time point at which the occasional instantaneous pain occurs can be determined by using the time point at which the value and / or correlation coefficient are high.

[0152] Figure 19 The present invention demonstrates a method for determining a persistent ERP that is effective for discriminating instantaneous pain, as well as a statistical verification method. The time point and duration of a persistent deviation of a pain stimulus from a reference stimulus are determined, and this interval is divided into 100-millisecond intervals, for example, and verification is performed using a representative value comparison method such as a t-test or ANOVA.

[0153] Figure 20The instantaneous pain experimental method used in the embodiment of the present invention is shown. In this embodiment, the "pain oddball paradigm" is used. This paradigm randomly presents a reference stimulus (39°C) with a high frequency (about 70%) and a deviation stimulus (52°C) with a low frequency (about 30%) of strong pain. At the same time, the brain waves are recorded, and the weighted average waveform of the reference stimulus and the deviation stimulus is calculated. Figure 19 Compare the methods shown.

[0154] Figure 21 A persistent positive potential effect was shown, which is a persistent signal (characteristic) observed under a deviation stimulus (52°C). A "persistent positive potential effect" was observed from 400 (F3, F4) or 600 (C3, C4) to 2000 milliseconds after the application of the 52°C stimulus. On the other hand, the mismatch potential and / or P300 observed in similar deviation subjects of vision and / or hearing were not significantly visible within 200 to 600 milliseconds. The transient pain paradigm used the pain Oddball paradigm, randomly presenting 70% of standard stimuli at 39°C and 30% of deviation stimuli at 52°C. In order to continue paying attention to the stimulus, the subject was instructed to count the number of painful stimuli.

[0155] Figure 22 The figure shows subjective evaluations of pain discomfort for standard stimuli (39°C) and deviant stimuli (52°C). The discomfort of the 52°C stimulus was significantly greater than that of the 39°C stimulus. The experimental paradigm randomly presented 39°C and 52°C stimuli multiple times, and subjects continuously rated the discomfort of the pain as the stimuli were presented.

[0156] Figure 23 The instantaneous pain discrimination value (inflection point in the Sigmoid function) is shown. Using the instantaneous pain discrimination value to classify pain intensity, a discrimination accuracy of 71% was achieved. The instantaneous pain discrimination value is based on the persistence characteristic.

[0157] Figure 24 The following is an overview of the process of discriminant estimation for two levels of instantaneous pain. In the process of the present invention, sample data is collected (for example, 160 samples (2 levels × 80 people), 4 feature quantities (F3, F4, C3, C4)), the training and test data are split (8:2), and a discriminant model is created using the training data (for example, using ten-fold cross-validation and LASSO (regularization) algorithms to calculate the optimal λ value and determine the coefficients of the feature quantities (4) and the intercept of the model). The discriminant model is used to perform discriminant estimation of the test data (discriminant estimation of the test data labels (no pain, severe pain) is performed, and the discrimination accuracy is calculated).

[0158] Figure 25 Shown by Figure 24The discrimination estimation process used 4 continuous ERP features (amplitude of 1000-1600 milliseconds) to determine the accuracy of instantaneous pain discrimination. Figure 24 In the process of , the discrimination rate was calculated 1000 times in order to perform 1000 model validations. The actual discrimination accuracy of the instantaneous pain level was about 70%, which is about 20% higher than the discrimination accuracy of the chance level of about 51%. The estimated value was obtained by using Figure 24 The obtained β coefficient and intercept were calculated using a multiple regression equation. The calculated values ​​were binary classified into those with or without transient pain using the median value, i.e., the discriminant value, and compared with the actual pain level.

[0159] Figure 26 Shown by Figure 24 The discrimination estimation process used the discrimination accuracy of instantaneous pain using four non-sustained ERP features (amplitude of 200 to 600 milliseconds). Figure 25 Using the same discrimination test process as the example shown in [1], a binary classification of transient pain versus non-transient pain was performed using non-persistent features from the time interval of 200 to 600 milliseconds, which did not exhibit persistent characteristics. Discrimination accuracy decreased by approximately 10% compared to using features from the time interval exhibiting persistent characteristics. Random data showed similar accuracy (approximately 50%) for this time interval and the persistent period.

[0160] Figure 27 is a representative flow chart in the practice of the present invention.

[0161] Figure 28 This is an example of a block diagram showing the functional configuration of the present invention.

[0162] Figure 29 This is an example of a block diagram showing the functional configuration of the present invention.

[0163] Figure 30 This is another example of a block diagram showing the functional configuration of the present invention. DETAILED DESCRIPTION

[0164] The present invention is described below. Throughout this specification, expressions in the singular should be understood to include the concept of their plural forms unless otherwise specified. Therefore, articles in the singular (e.g., "a," "an," "the," etc., in the case of English) should be understood to include the concept of their plural forms unless otherwise specified. In addition, the terms used in the specification should be understood to be used in the sense commonly used in the field unless otherwise specified. Therefore, all technical terms and scientific terms used in this specification have the same meaning as those generally understood by those skilled in the art. In the event of a conflict, this specification (including definitions) takes precedence.

[0165] (definition)

[0166] First, the terms and general techniques used in the present invention are explained.

[0167] In this specification, the term "subject" is used synonymously with "patient" or "subject," and refers to any organism or animal used as the subject of the disclosed techniques, such as pain measurement and EEG measurement. The subject is preferably a human, but is not limited to such. In this specification, when pain estimation is performed, the term "estimation subject" may be used, but this term has the same meaning as "subject."

[0168] In this specification, "brain wave" has the same meaning as that commonly used in this technical field, and refers to the electric current generated by placing a pair of electrodes on the scalp due to the potential difference associated with the neural activity of the brain. Brain waves include an electroencephalogram (EEG) obtained by deriving and recording the time changes of the electric current. When at rest, the main component is a wave with an amplitude of about 50μV and a frequency of about 10Hz. It is called an α wave. During mental activity, the α wave is suppressed and a fast wave with a small amplitude of 17 to 30Hz appears, which is called a β wave. During light sleep, the α wave gradually decreases and the θ wave of 4 to 8Hz appears. During deep sleep, the δ wave of 1 to 4Hz appears. These brain waves can be expressed with specific amplitudes and frequencies (power). In the present invention, the analysis of the amplitude may be important.

[0169] In this specification, "brain wave data" refers to any data related to brain waves (also called "brain activity amount", "brain characteristic amount", etc.), including amplitude data (EEG amplitude, frequency characteristics, etc.). The "analysis data" obtained by analyzing these brain wave data can be used in the same way as the brain wave data, so in this specification, they are sometimes collectively referred to as "brain wave data or its analysis data". As analysis data, for example, average amplitude (for example, Fz, Cz, C3, C4), frequency power (for example, Fz(δ), Fz(θ), Fz(α), Fz(β), Fz(γ), Cz(δ), Cz(θ), Cz(α), Cz(β), Cz(γ), C3(δ), C3(θ), C3(α), C3(β), C3(γ), C4(δ), C4(θ), C4(α), C4(β), C4(γ), etc.) can be cited. Of course, other data that are usually used as brain wave data or its analysis data are not excluded.

[0170] In this specification, "amplitude data" is a type of "brain wave data," referring to data on the amplitude of brain waves. It is sometimes simply referred to as "amplitude" or "EEG amplitude." This amplitude data is an indicator of brain activity and is therefore sometimes referred to as "brain activity data," "brain activity level," and so on. Amplitude data can be obtained by measuring the electrical signals of brain waves and can be expressed as a potential (which can be expressed in μV, etc.). Average amplitude can be used as amplitude data, but is not limited to this.

[0171] In this manual, the so-called "frequency power" is to express the frequency component of a waveform as energy, also known as power spectrum. Regarding frequency power, the frequency component of a signal buried in a noise-containing signal in the time domain can be extracted and calculated using a high-speed Fourier transform (FFT) (an algorithm for high-speed calculation of discrete Fourier transform (DFT) on a computer). Regarding the FFT of a signal, the output of a power spectrum density PSD or a power spectrum that is a source of power can be standardized using, for example, a periodogram of a function in MATLAB. PSD represents how the power of a time signal is distributed with respect to frequency, and the unit is watt / hertz (w / Hz). The power spectrum is calculated by integrating each point of the PSD over the frequency range that defines the point (i.e., over the resolution bandwidth of the PSD). The unit of the power spectrum is watt. Regarding the value of power, it is possible to read it directly from the power spectrum without integrating over the frequency range. Both PSD and power spectrum are real numbers and therefore do not contain any phase information. In this way, the calculation of frequency power can be calculated using the standard functions of MATLAB.

[0172] In this specification, "pain" and "ache" have the same meaning, both referring to the feeling produced as a stimulus when there is a generally strong invasion such as injury or inflammation on a part of the body. For humans, the feeling accompanied by a strong sense of discomfort is also included in the general feeling. Moreover, skin pain and the like also have the characteristics of external acceptance to a certain extent, and cooperate with other skin sensations and taste to contribute to the qualitative judgment of the hardness, sharpness, heat (hot pain), coldness (cold pain), spiciness, etc. of external objects. In addition to the skin and mucous membranes, human pain may also be generated in almost all parts of the body (for example, pleura, peritoneum, viscera (visceral pain, excluding the brain), teeth, eyes and ears, etc.), all of which can be perceived in the brain as brain waves or their changes. In addition, the internal pain represented by visceral pain is also included in the pain. The above-mentioned pain is called somatic pain relative to visceral pain. In addition to somatic pain and visceral pain, there are also reports of "referred pain", a phenomenon of surface pain in a part different from the actual injured part. The present invention can classify these pain types and can also classify these various pain types based on whether they are comfortable or not.

[0173] Regarding pain, there are individual differences in sensitivity (pain threshold), and there are qualitative differences depending on the way the pain stimulus occurs and / or the difference in the receptor site. There are classifications such as dull pain and sharp pain, but in the present disclosure, any type of pain can be measured, estimated and classified. In addition, it is also possible to deal with acute pain (A pain), chronic pain (B pain), (acute) local pain and (chronic) diffuse pain. The present invention can deal with allodynia such as allodynia. Among the peripheral nerves that transmit pain, two nerve fibers, "Aδ fibers" and "C fibers", are known. For example, when clapping your hands, the initial pain is transmitted through the conduction of Aδ fibers, transmitting a clearly located sharp pain (primary pain; sharp pain). Afterwards, through the conduction of C fibers, an unclearly located tingling pain (secondary pain; dull pain) is felt. Pain is classified into "acute pain" that lasts for 4-6 weeks and "chronic pain" that lasts for more than 4-6 weeks.

[0174] Although pain is an important vital sign alongside heart rate and / or body temperature, blood pressure, and respiration, it is difficult to characterize as objective data. Representative pain rating methods, such as the visual analogue scale (VAS) and the faces pain rating scale, are subjective assessment methods and cannot compare pain between patients. On the other hand, the inventors of the present invention have deduced that, as an indicator for objectively evaluating pain, by focusing on brain waves, which are less susceptible to the influence of the peripheral circulatory system, observing the changes in their amplitude / latency in response to painful stimuli and applying sparse analysis to them, it is possible to distinguish and classify the type of pain. Furthermore, the inventors of the present invention have deduced that by observing the changes in amplitude / latency time relative to the painful stimulus, it is also possible to classify the type of pain (comfortable or not). This classification can be performed for both instantaneous and continuous stimuli.

[0175] In this specification, "transient pain" refers to pain that occurs in synchronization with a single stimulus, with an evoked potential signal appearing within 50 to 200 milliseconds, and refers to pain similar to that felt by a needle prick. Meanwhile, "breakthrough pain," a term used in the field of pain relief, differs from transient pain in that it causes temporary, intense pain, and therefore encompasses the concept of persistent pain used in this specification.

[0176] Furthermore, breakthrough pain was defined as “transient pain or increased pain that occurs regardless of the presence and / or severity of ongoing pain and regardless of the use of analgesics.”

[0177] Persistent pain generally refers to pain that is not single-shot, or is not perceived as single-shot because single-shot pain occurs continuously, and it is difficult to discern the synchronization of evoked potentials with single-shot stimulation, and the signal does not appear in the form of event-related potentials. Almost all clinical pain is included in this category. The same is true at the animal level, but while transient pain can be treated as an increase in amplitude, persistent pain can also be described as pain that sometimes manifests as a decrease in amplitude.

[0178] In the present invention, being able to distinguish instantaneous pain from persistent pain is one of the key points, which helps to determine the appropriate treatment. Therefore, it is also important to be able to clearly categorize "pain" around the concept of "treatment". Figure 18 As shown in FIG, the persistent brain wave characteristic quantity synchronized with the instantaneous pain shows a slow movement in the positive direction within the range of 2000 milliseconds. Figure 18 By templating as shown, online continuous regression is performed to determine the time point with high approximation, thereby determining the time point when instantaneous pain occurs.

[0179] In the present invention, one of the key points regarding pain determination is the ability to efficiently determine and classify pain even with a small number of parameters.

[0180] In the present invention, one of the key points is to be able to distinguish whether the pain is "needing treatment" compared to the intensity itself. Therefore, it is also important to be able to clearly categorize "pain" around the concept of "treatment". For example, it can be said to be a "qualitative" classification of pain such as "comfortable or not" and "unbearable". For example, the positioning, inflection point and / or width of the classification value and / or its relationship of the "pain classification value" can also be defined. In addition to the case of n=2, it can be assumed that there may be cases where n=3 or more. In addition, in the case of more than 3, it can also be divided into "no pain", "painful but comfortable" and "painful". For example, it is possible to make judgments such as "unbearable, must be treated" pain, "intermediate", and "painful but not a big deal". The judgment using the regression model calculated by the sparse modeling of the present invention can identify "unbearable" and "painful but tolerable, no treatment required".

[0181] In this specification, "subjective pain level" refers to the level of pain felt by a subject and can be expressed using conventional techniques such as the Computerized Visual Analog Scale (COVAS) or other well-known techniques, such as the Support Team Assessment Schedule (STAS-J), the Numerical Rating Scale (NRS), the Faces Pain Scale (FPS), the Abbey Pain Scale (Abbey), the Check List of Nonverbal Pain Indicators (CNPI), the Non-communicative Patient's Pain Assessment Instrument (NOPPAIN), and the Doloplus2 Pain Rating Scale for Alzheimer's Disease. Such subjective pain levels can also be applied to the assessment of instantaneous pain.

[0182] In this specification, "stimulus" refers to a certain reaction caused by an object. When the object is a living thing, it refers to the main cause of temporary changes in the physiological activity of the organism and / or a part thereof. The specific examples of "stimulus" exemplified by events related to pain include any stimulus that can produce pain. For example, electrical stimulation, cold stimulation, heat stimulation, physical stimulation, chemical stimulation, etc. In the present invention, any stimulus can be used to generate a pain classification value, and temperature stimulation (cold stimulation or warm stimulation) or electrical stimulation is usually used. Regarding the stimulation level, usually 3 or more types are used, preferably 4 or more types are used, more preferably 5 or more types are used, and even more preferably 6 or more types are used, or more types of stimulation than these can be used. In the case of temperature stimulation, for example, in the case of low temperature stimulation, it can be reduced with a suitable feeling in the range of 10°C to -15°C, and the temperature can be reduced by 5°C when 6 points are taken, thereby generating 6 temperature levels of stimulation. Regarding the evaluation of stimulation, for example, conventional techniques such as the computerized visual analogue scale (COVAS) or other well-known techniques, such as the Support Team Assessment Schedule (STAS-J), the Numerical Rating Scale (NRS), the Faces Pain Scale (FPS), the Abbey pain scale (Abbey), the Check list of Nonverbal Pain Indicators (CNPI), the Non-communicative Patient's Pain Assessment Instrument (NOPPAIN), the Doloplus2 Pain Rating Scale for Alzheimer's Disease, etc., can be used to make them correspond to the subjective pain sensation level. Examples of values ​​that can be used as stimulus intensity include the nociceptive threshold (the threshold at which nociceptive fibers generate nerve impulses), the pain detection threshold (the intensity of an injurious stimulus that a person can perceive as pain), and the pain tolerance threshold (the strongest stimulus intensity among injurious stimuli that a person can tolerate experimentally).

[0183] In the case of psychological conditions, stimuli include, for example, any element that can be perceived through the five senses (sight, hearing, taste, touch, smell) and complete information processing in the brain, and / or any element that can be perceived mentally, such as social pressure.

[0184] In this specification, "evoked (brainwave) potential" refers to the brainwave component induced by exogenous stimulation, and there are positive and negative brainwaves. For example, P1 (P100) (wave) can be cited as a positive one, and N1 (N100) (wave) or N125 can be cited as a negative one (refer to Donchin, E., Ritter, W., & McCallum, C. (1978). Cognitive psychophysiology: The endogenous components of the ERP. In E. Callaway, P. Tueting, & S. Koslow (Eds.), Brain event-related potentials in man (pp. 349-441). New York: Academic Press). Although it reflects the processing process for physical stimulation, event-related potentials are generated by higher-level processing processes, especially changes in memory, prediction, attention, and psychological state.

[0185] In this specification, "event-related potential component" (event-related potential, ERP) refers to an endogenous brain wave component, and there are positive and negative brain waves. For example, P2 (P200) (wave) can be cited as a positive one, and N2 (N200) (wave) can be cited as a negative one. The event-related potential component is generated after the induced brain wave component in a time series (Donchin et al., supra). As a result of thinking and / or cognition, it can be said to be a reaction of the brain measured in some form. In more detail, it can be said to be an electrophysiological reaction to a type of internal and external stimulation. ERP is measured by brain waves. The same concept based on magnetoencephalography (MEG) can be called an event-related magnetic field (event-related field, (ERF)), which can be analyzed in the same way as ERP.

[0186] In this specification, “P100” or “P1” refers to a positive evoked electroencephalogram component, and is also referred to as a first positive evoked electroencephalogram component.

[0187] In this specification, "N100" or "N1" refers to a negative evoked electroencephalogram component, also called the first negative evoked electroencephalogram component, which occurs after P100.

[0188] In this specification, "P200" or "P2" is also referred to as a positive early event-related potential component, which usually occurs after P1 and N1.

[0189] In this specification, “N200” or “N2” refers to an initial negative event-related potential component that occurs after P200.

[0190] In this specification, "P300" or "P3" refers to a positive mid-term event-related potential component. It is usually generated after P200 and N200. The previous P300 had a peak value and was observed as a transient potential component. In the present invention, a potential component with a persistent characteristic was found at a position overlapping with or after the P300, and it was found that this was associated with transient pain. Such a situation was unpredictable based on previous cognition and provides an indicator useful for treatment. Therefore, the persistent component found in the present invention can also be called a "persistent positive component" containing a "persistent P300".

[0191] In most cases, both the "evoked brain wave component" and the "initial event-related potential component" are generated before 250 milliseconds. Usually, the "evoked brain wave component" is generated first and the "event-related potential component" is generated later. However, due to the relationship with the purpose of the present invention, in the case of being generated later than 250 milliseconds, the measurement range can be determined with 250 milliseconds as the starting point (refer to Cul et al. (2007). Brain Dynamics Underlying the Nonlinear Threshold for Access to Consciousness. PLoS Biol 5 (10): e260). There are also cases where evoked and event-related potential components such as N100 and P200 are not clearly observed due to the state of stimulus perception, low alertness, etc., but they will always exist as long as the vegetative state in which the brain's neocortex loses function is not present. Therefore, in the measurement, 250 milliseconds can also be set as the starting point for convenience. The P200 component is a component that reflects selective attention and is a component that is like the entrance of the brain wave component of high-level cognition that begins to appear in the middle period. Figure 17 The time relationship among P1, N1, P2, N2, P3, etc. is illustrated in FIG. Figure 17This refers to the temporal changes in the period from the onset of the evoked potential to the event-related potential (P2 onwards). Prior to the evoked potential before P1 (<100 milliseconds), there are more detailed responses deep in the brain (P50, P25, etc.). Therefore, the measurement target of the present invention can also be defined as the 2000 milliseconds from the earliest of the three time points: "negative and positive exogenous evoked EEG components," "initial negative and positive endogenous event-related potential components," and 250 milliseconds.

[0192] P300 (P3) is positioned as the third positive ERP component following P100 and P200. The peak latency often occurs around 300 milliseconds, but is sometimes delayed by hundreds of milliseconds in association with cognitive load and / or processing content. In addition, it is said that there are two types of P300. One is called P3a, a reaction to novel stimuli, in other words, a startle reaction to novel stimuli, which is mainly manifested in the frontal lobe. The other is called P3b, which is a conscious reaction to the target of attention and will not be observed if attention is not paid. It is mainly manifested in the central part of the parietal lobe and has a peak time slightly later than P3a. Although it is not desired to be bound by theory, the persistent positive component used in the present invention has no direct relationship with P300. The positive component of the present invention means the third positive component and can be called "persistent P300".

[0193] Furthermore, it has been reported that the P300, unlike brain activities such as the early evoked potential (N100), does not have the property of continuously changing relative to the stimulus characteristics. For example, if the stimulus intensity increases, the arousal component of the evoked component N100 "continuously" produces a potential change. In contrast, the P300 shows a "sigmoid function pattern" that smoothly approximates the discreteness of conscious and unconscious reactions (Cul et al. (2007). Brain Dynamics Underlying the Nonlinear Threshold for Access to Consciousness. PLoS Biol 5 (10): e260). If this cognition is applied to the pain situation of the present invention, this component can be seen, indicating that "there is a consciousness of pain." Moreover, it is not a simple P300 activity but shows a continuous characteristic. This point can be said to be an important cognition.

[0194] The event-related potential of the mid-term as P300 generally greets peak potential at 300 to 400 milliseconds, but this may change because of the content and / or load of processing information. Generally speaking, the deviation to the positive direction shows the non-persistence of starting from before 300 milliseconds and welcoming peak between 300 to 400 milliseconds, but as shown in the present invention, persistence continues unexpectedly, even after 2000 milliseconds after the stimulation is presented, it can not get back to baseline. No matter how long it will not return, this time actually continued for example 2 seconds, and this is an unexpected event, and it is also unexpected that this can be used for the discrimination of pain. In addition, the persistence positive component is one of key points, and this is also an unexpected feature.

[0195] In this specification, "target stimulus" refers to the stimulus being measured (e.g., a stimulus that is the source of pain). In actual diagnosis, there are cases where pain stimuli occur naturally and irregularly, externally or internally, and there are also cases where external pain stimuli are artificially presented as a reference for examination purposes.

[0196] In this specification, "reference stimulus" refers to a stimulus that serves as a comparison benchmark for the measurement of pain, etc. As a reference stimulus, for example, it is a stimulus that has the same type of physical properties as a painful stimulus but does not cause pain. By setting this biological response as a benchmark, it is possible to determine the biological response signal related to pain based on the minimum difference in physical properties. For example, the 39°C thermal stimulus used in the embodiments of this application document is used as a reference stimulus (for investigating the characteristics of thermal stimuli that cause pain (such as high temperature stimuli of 52°C)). However, the reference stimulus is not limited to thermal stimulation, and electrical stimulation can also be cited, but is not limited to these.

[0197] In this specification, the "mid-term time period" refers to the time period from the time point when the initial event-related potential component is generated to the time period when the later event-related potential (P600, later positive component, etc.) is generated. It includes values ​​in the range of about 250 milliseconds to 600 milliseconds. It is the time period after 250 milliseconds when the early time period in which the evoked potential (N100 and / or P100) is seen, and is the time period in which cognitive processing components such as P300 and N400 appear. It can even include the time period until the later positive component or P600 appears (around 600 milliseconds). The mid-term time period can, for example, include 300 milliseconds, 400 milliseconds, 500 milliseconds, 600 milliseconds, and consecutive time periods in between (for example, including 315 milliseconds, 350 milliseconds, etc.).

[0198] In this specification, "persistent characteristics" or "persistent ERP" refers to a state in which the signal continues smoothly, such as when no peaks or valleys occur, or when the peaks are delayed for a long time. Typically, the signal continues for at least 100 milliseconds without a peak. As will be described repeatedly below, there are multiple technical methods for identifying persistent components. 1) After the stimulus is presented, the activity (brain wave waveform) relative to the reference stimulus deviates. If this deviation occurs immediately after the stimulus is presented, it may not be a component related to the perception and / or cognition of pain but a false signal. 2) When observing the brain wave waveform presented by each stimulus, if the baseline of the brain wave itself is significantly shifted upward or downward, and disappears when the baseline is returned to the baseline through linear correction, it is not considered a persistent component. 3) When low-frequency band components such as 0.02Hz, 0.01Hz, or 0.1Hz are cut off, the persistent component disappears as the duration continues. Therefore, in cases that do not belong to 1 or 2 and such low-frequency components are cut off and disappear, it can be identified as a persistent ERP. 4) When performing offline analysis, if a significant correlation is observed between characteristic quantities including the amplitude, frequency power, etc. of the persistent component and stimulus characteristics such as pain intensity, behavioral characteristics and / or psychological characteristics of the subject during examination, it can be further determined that the persistent component is a signal associated with the stimulus.

[0199] In addition, typically, a case where there is a peak that moves like a mountain (for example, a clearly visible width of about 0.2 seconds (200 milliseconds)) is not included in the persistence. Conversely, a case where "the effect does not appear and disappear suddenly, and there is no peak or the peak is not noticeable" can be judged as having persistence. Regarding "mountain (with a clear peak)" and "persistence", as observed Figure 18As can be seen from the examples of etc., it is possible to determine if there is a clear difference. For the situation where the effect begins to appear around at least 400 milliseconds after the stimulus is presented, and there is no obvious peak like a mountain or the appearance is quite late and quite slowly, it can be said to be persistence from experience. The situation where the brain activity deviates from the conditions that become the benchmark and lasts for at least more than 100 milliseconds, preferably more than 200 milliseconds, preferably more than 300 milliseconds, preferably more than 500 milliseconds, and more preferably more than 1000 milliseconds is called a persistence state. If a positive effect is seen in the time interval of the measurement range that becomes the object, for more than a certain time or in the entire interval, it can be said to be a persistence state. For example, if the interval is divided into every 100 milliseconds and conditional differences are seen in all or many intervals, it can be interpreted as persistence. For example, in this manual, a situation where a statistically significant difference is seen between a range that is at least 100 milliseconds long can be determined to be presenting persistence. In addition, the persistent component is also sensitive to the frequency filtering process. For example, if the predicted persistent negative ERP, namely CNV (contingent negative variation), lasts for more than 3 seconds, the low-frequency band above 0.02 Hz is generally included in the analysis. In other words, if this component needs to be judged objectively, it can be judged by whether the persistent negative ERP disappears when the low-frequency band to be cut off is increased. Or, for example, Figure 19 As shown, the time point and duration of the continuous deviation of the pain stimulus relative to the reference stimulus are determined, the interval is divided into, for example, every 100 milliseconds and verified using representative value comparison methods such as t-verification and ANOVA, so that a method for determining the continuous ERP and statistical verification can be performed.

[0200] In one embodiment, for the persistence characteristic, in the comparison, it is determined whether there is a duration in which the value of the brain wave data or its analysis data obtained by the object stimulation is different from the value of the brain wave data or its analysis data obtained by the reference stimulation, and if there is such a different duration, it is determined whether it becomes the same value again from the difference, and if there is such a duration and it does not become the same value again, it is determined that there is uncomfortable pain. As an example of the duration, for example, it is more than 100 milliseconds, preferably lasting for more than 200 milliseconds, preferably lasting for more than 300 milliseconds, preferably lasting for more than 500 milliseconds, and more preferably lasting for more than 1000 milliseconds, but it is not limited to these. The reference stimulus is a stimulus for which the reaction produced by the object is known, for example, if it is a sound stimulus, it is a stimulus that serves as a benchmark such as 1000 Hz, if it is a visual stimulus, it is a simple stimulus such as a graphic, and if it is an electrical stimulus, it is a stimulus of weak intensity. Examples include stimuli that do not produce discomfort and / or discomfort due to physical or psychological characteristics.

[0201] In the determination of the persistence characteristic, in the case of event-related potentials, the brain activity of the reference stimulus is used as a benchmark to determine the corresponding brain activity effect of pain. Therefore, in the case of pain that has already occurred and is persistent, it may not be possible to distinguish only by the characteristic quantity but also require further analysis of the characteristic quantity. On the other hand, the indicator of the present invention can be used to investigate individual differences in pain sensitivity, etc. For example, before monitoring pain, the instantaneous pain stimulus can be randomly presented in two forms, namely "non-painful reference stimulus" and "pain stimulus", about 10 times each, and the persistent effect of the pain stimulus (after z-value transformation) can be applied with a Sigmoid discriminator to determine which pain category it belongs to. It is assumed that this sensitivity can be used as a basis for the correction of the weighting coefficient of the discrimination algorithm used when performing pain monitoring. In addition, in a preferred embodiment, as the characteristic quantity to be targeted, it is preferred to be a characteristic quantity of potential, duration, or a combination of the two. If an individual's pain evaluation shows a significant correlation with the characteristic quantity, it can be indicated that there are individual differences. In the embodiment, in the curve approximated by the Sigmoid function, approximately 30% of the samples made errors in the discrimination of severe pain. Therefore, it can be said that the difference between the error group and the correct group is the individual difference.

[0202] It is known that a mid-term event-related potential such as the P300 peaks between approximately 300 and 400 milliseconds, but this varies depending on the content and / or load of the information being processed. Therefore, generally speaking, a shift in the positive direction appears to be non-persistent, starting 300 milliseconds ago and peaking between 300 and 400 milliseconds. However, as shown in this embodiment, the persistence is unexpectedly long, and sometimes it does not return to the baseline even after 2000 milliseconds after stimulus presentation (see Figure 18 ), by understanding these, it is possible to understand instantaneous pain and other pains, which is a previously unknown and surprising effect. In a preferred embodiment of the present invention, it is hoped that the "polarity" difference of the positive component rather than the persistent negative component will also be focused on.

[0203] Alternatively, when determining whether the persistence feature exists, the following is checked.

[0204] 1) After the stimulus is presented, a deviation occurs in the activity (brain wave waveform) relative to the reference stimulus. If the deviation occurs immediately after the stimulus is presented, it is not a component related to the perception and / or cognition of pain, but may be a false signal. 2) When observing the brain wave waveform of each stimulus presentation, the case where the baseline of the brain wave itself is significantly shifted upward or downward, and disappears when the baseline is returned to the baseline through linear correction, is not called a persistent component. 3) When the low-frequency band components such as 0.02Hz, 0.01Hz, and 0.1Hz are cut off, the persistent component disappears with the duration. Therefore, in the case that does not belong to 1 or 2 and the above-mentioned low-frequency frequency components are cut off and disappear, it can also be specified as a persistent ERP. 4) At the time of offline analysis, when a significant correlation is found between the characteristic quantities including the amplitude and frequency power of the persistent component and the stimulus characteristics such as pain intensity, the behavioral characteristics and / or psychological characteristics of the subject during the examination, it can be further determined that the persistent component is a signal associated with the stimulus.

[0205] In this specification, "sparse" or "sparse modeling" refers to a method of displaying the whole image based on a small amount of information as a scientific mathematical modeling method. Sparse means that there are many gaps and the amount is small. The basic idea is (1) to assume that the explanatory variables of high-dimensional data are fewer than the number of dimensions (sparse), (2) to minimize the number of explanatory variables by regularization, i.e., setting a penalty term (for example, introducing a λ coefficient in L1 regularization), and at the same time to require a smooth model to adapt to the data, so as to (3) enable automatic selection of explanatory variables without manual intervention.

[0206] In this specification, "fitting" to a function refers to a technique for fitting a certain measured value and / or a curve obtained therefrom in order to make it approximate to a target function, and can be implemented based on any method. For example, least squares fitting and nonlinear regression fitting (MATLAB's nlinfit function, etc.) can be cited. For the approximate curve, the regression coefficient is calculated, so that it can be judged whether the curve can be used in the present invention and whether it is preferred. As a regression coefficient, the regression model is meaningful, and the coefficient of determination (R 2 ) is 0.5 or greater, 0.6 or greater, 0.7 or greater, 0.8 or greater, 0.85 or greater, 0.9 or greater, etc. The closer the value is to "1", the more ideal it is, and the higher the value, the higher the reliability. Alternatively, a specific threshold value can be used to classify the estimated value and the measured value, and the fitting accuracy of the two can be verified by comparison (the so-called discrimination accuracy in the analysis mentioned in this invention refers to this situation).

[0207] In this specification, "cross validation" is also called "cross test" (all Cross-validation), which refers to a method of dividing sample data in statistics, first analyzing a part of it, and testing the analysis with the remaining part as a method of verifying and confirming the validity of the analysis itself. The case of dividing into 10 parts is called ten-fold cross validation, and the case of dividing into 5 parts is called five-fold cross validation. It is a method for verifying and confirming the extent to which the analysis of data (and the derived estimates, statistical predictions) can truly cope with the matrix (population) through good approximation. The data initially analyzed is called "learning data" or "training case set (training set)", etc., and the other data is called "test case set (testing set)" or "test data", etc. The 10 divisions of the ten-fold cross validation illustrated in this specification are an example, and leave-one-out cross validation can also be used. There are also cases where they are collectively referred to as "split cross validation (test)". When carrying out sparse modeling, determine the appropriate λ (preferably optimal λ) value used in penalty term by cross validation, and determine the parameter (partial regression coefficient) of characteristic quantity and the constant (intercept) of algorithm.Here, " λ value " refers to the hyperparameter that plays a role in regularization, for making the fitness of model smooth, improving generalization ability.Regularization has L1 regularization and L2 regularization, and the LASSO (Least absolute shrinkageand selection operator) used in a preferred embodiment of the present invention uses L1 regularization.λ coefficient is a positive coefficient, determines the appropriate (preferably optimal) solution by cross validation.As a concrete example, the LASSO function of MATLAB solves the following minimization problem.

[0208] [Mathematical formula 1]

[0209] Min(Dev(β0,β)+λΣ|β j |)

[0210] Min: Minimize

[0211] Deviation: Deviation (the deviation of the estimated value of the regression model using the intercept β0 and the regression coefficient β from the observed value)

[0212] N: number of samples

[0213] λ: Regularization parameter with positive value

[0214] Here, the partial regression coefficient refers to the coefficient of each explanatory variable in the regression equation obtained in the regression analysis, and the intercept refers to the intersection of a curve or graph on a coordinate plane and the coordinate axis.

[0215] Sparse modeling is described in Ozeki et al. (Special Lecture Notes on Electronic and Physical Engineering, Osaka City University, 2015 (Revised Edition: September 9, 2015)), and the sparse model analysis and LASSO described therein can be applied. The entire text of this document is incorporated herein by reference.

[0216] In norm selection used in sparse model analysis, compressed sensing can be used to determine whether the resulting solution is a sparse solution. Here, noisy compressed sensing can be used when noise is observed. Noisy compressed sensing can use a method called LASSO (Least Absolute Shrinkage and Selection Operators; R. Tibshirani: JR Statist. Soc. B, 581a (1996) 26).

[0217] In sparse model analysis, data input, algorithm determination by the discrimination / estimation unit, and discrimination / estimation output can be performed multiple times (e.g., 1,000 times, or more or less) to form an appropriate value (preferably optimized). For example, this can be performed 2,000 times, 3,000 times, 5,000 times, or 10,000 times.

[0218] When using an appropriate (preferably optimal) λ coefficient to determine the parameters (coefficients) of the feature quantity and the constant (intercept) of the algorithm and performing a discrimination estimation on the test data, this operation is repeated 1000 times, and the average is the discrimination accuracy. It can be said that the rigor is significantly different from the accuracy discrimination of the existing technology. In addition, when using a regression model generated for ordinary people, it is preferable to calibrate it for each person. A technique for calibrating the parameters (coefficients) of the feature quantity and / or the constant (intercept) of the algorithm used in the model for each person can be added.

[0219] When performing sparse model analysis during modeling, the following considerations should be taken into account. For example, in LASSO, all coefficients are uniformly multiplied by the lambda value for regularization, so the features used must be treated in the same unit. Therefore, normalization of the features is necessary.

[0220] In this specification, the term "pain classification" can be applied from various perspectives. Typical examples include categorizing whether the subject is experiencing "pain" or "no pain." Other examples include determining whether pain is felt and distinguishing between strong and weak pain. However, this is not limited to these criteria and also includes qualitative distinctions (e.g., "tolerable" pain versus "unbearable" pain).

[0221] (Preferred embodiment)

[0222] The preferred embodiments of the present invention are described below. The embodiments provided below are provided for a better understanding of the present invention, and it should be understood that the scope of the present invention should not be limited to the following description. Therefore, those skilled in the art will appreciate that they can be appropriately changed within the scope of the present invention with reference to the description in this specification. In addition, it should be understood that the following embodiments of the present invention can be used independently or in combination.

[0223] In addition, the embodiments described below are all inclusive or specific examples. The numerical values, shapes, materials, components, configuration positions and connection methods of components, steps, and the order of steps shown in the following embodiments are merely examples and are not intended to limit the scope of the claims. In addition, components in the following embodiments that are not described in the independent claims representing the superordinate concept are described as arbitrary components.

[0224] Therefore, the inventors of the present invention have evaluated various pains using various methods to clarify the relationship between pain and brain waves. The relationship between pain and brain waves clarified by the inventors of the present invention will be described below with reference to the accompanying drawings.

[0225] First, the relationship between pain caused by electrical stimulation and brain waves will be described. The data shown below represents data from a representative subject among multiple subjects.

[0226] Figure 1A Graph showing the relationship between electrical stimulation and pain rating (VAS). Figure 1B This is a graph showing the relationship between electrical stimulation and pain level (paired comparison). Figure 1C This is a graph showing the relationship between electrical stimulation and brain wave amplitude. Figure 1D This is a graph showing an example of an electroencephalogram waveform.

[0227] Figure 1A 、 Figure 1B and Figure 1C The horizontal axis represents the current value of electrical stimulation. Figure 1A The vertical axis represents the pain level reported by the subjects using VAS. Figure 1B The vertical axis represents the pain level reported by the subjects through pairwise comparison. Figure 1C The vertical axis represents the amplitude of the brain wave. Figure 1D In the graph, the horizontal axis represents time, and the vertical axis represents signal level.

[0228] Paired comparison involves pairing two electrical stimulations of varying magnitudes together. For each of multiple stimulations, the subject reports a numerical value indicating which stimulation caused what level of pain. This method reduces the influence of past experience on pain levels by comparing the two pain levels.

[0229] like Figure 1A and Figure 1B As shown in Figure 2, regardless of the VAS or paired comparison method, the relationship between the current value of the electrical stimulation (i.e., the intensity of the stimulation) and the pain level is generally represented by a Sigmoid curve. In addition, the shape of the Sigmoid curve (such as the upper and lower limits, etc.) varies depending on the subject.

[0230] In addition, if Figure 1C As shown in FIG, the relationship between the current value of the electrical stimulation and the amplitude value of the brain wave can also be roughly represented by the Sigmoid curve. Here, the amplitude value of the brain wave uses the difference between the maximum peak value and the minimum peak value (i.e., the peak-to-peak value). For example, in Figure 1D , the largest difference value (N1-P1) among the three difference values ​​(N1-P1, N2-P2, N1-P2) is used as the amplitude value.

[0231] In this way, the relationship between the intensity of electrical stimulation and the pain level, and the relationship between the intensity of electrical stimulation and the amplitude of brain waves can both be represented by Sigmoid curves. That is, the pain level and the amplitude of brain waves both have upper and lower limits relative to the electrical stimulation, and show the same changes relative to the intensity of electrical stimulation. Therefore, after analyzing the relationship between the amplitude of brain waves and the pain level, it was found that the relationship between the amplitude of brain waves and the pain level is as follows: Figure 1E and Figure 1F As shown.

[0232] Figure 1E This is a graph showing the relationship between the pain level (VAS) caused by electrical stimulation and the brain wave amplitude. Figure 1F This is a graph showing the relationship between the pain level (paired comparison) caused by electrical stimulation and the brain wave amplitude. Figure 1E and Figure 1F In the figure, the horizontal axis represents the amplitude of brain waves, and the vertical axis represents the pain level.

[0233] like Figure 1E and Figure 1F As shown in the figure, the pain level caused by electrical stimulation and the amplitude of the brain wave are linear, regardless of whether it is VAS or paired comparison. In other words, the amplitude of the brain wave is proportional to the pain level.

[0234] In addition, in this disclosure, the so-called linearity includes not only strict linearity but also substantial linearity. That is, linearity includes a relationship that can be approximated to linearity within a specified error range. The specified error range is, for example, the coefficient of determination R in regression analysis. 2 To define the coefficient of determination R 2 It is the value obtained by subtracting the sum of squares of the residuals divided by the sum of squares of the differences between the observed values ​​and the mean from 1. The specified error range is, for example, R 2 The range is 0.5 or more.

[0235] Regarding the relationship between pain caused by thermal stimulation and brain waves, similar to the case of electrical stimulation, there is a linear relationship between the pain level and the brain wave amplitude.

[0236] Figure 1G This is a graph showing the relationship between the pain level (VAS) caused by thermal stimulation and the brain wave amplitude. Figure 2 This is a graph showing the relationship between the pain level (paired comparison) caused by thermal stimulation and the brain wave amplitude. Figure 1G and Figure 2 In the figure, the horizontal axis represents the amplitude of brain waves, and the vertical axis represents the pain level.

[0237] like Figure 1G and Figure 2 As shown, the pain level caused by thermal stimulation and the amplitude of the brain wave have a linear relationship, whether in the VAS or paired comparison. Furthermore, the upper and lower limits of the brain wave amplitude vary among subjects, but the inventors' experiments have shown that the upper limit of the amplitude does not exceed approximately 60 μV.

[0238] As described above, the inventors analyzed the relationship between pain levels and EEG amplitude values, obtained using various methods to assess various types of pain. The results revealed a specific relationship between EEG amplitude and pain. Furthermore, the inventors discovered a sparse modeling approach that can estimate pain severity based on this specific relationship between EEG amplitude and pain.

[0239] Pain Classification

[0240] In one embodiment, the present invention provides a method for discriminating or classifying pain experienced by an estimated subject based on the subject's electroencephalogram (EB). The method comprises the following steps: a) obtaining model EB data or analysis data thereof corresponding to a model stimulus intensity; b) extracting model EB features from the EB data or analysis data; c) setting a target pain level, subjecting the model EB features and the pain level to sparse model analysis, determining an appropriate λ (preferably an optimal λ) (using a least squares method to reduce error, preferably a method to minimize error), determining parameters (partial regression coefficients) of the model EB features and an algorithm constant (intercept) corresponding to the appropriate λ (preferably the optimal λ), and generating a regression model; d) obtaining (measurement) EB data or analysis data of the estimated subject; e) extracting measurement EB features from the measurement EB data or analysis data; f) fitting the measurement EB features to the regression model to calculate the corresponding pain level; and g) displaying the pain level as needed. a) to c) are the stages of generating a regression model, and based on the generated regression model, d) to f) are the stages of fitting the brainwave data from the subject (object) or its analysis data to calculate the pain level. Through such steps, high-precision pain discrimination can be performed. When the technology of the present invention is used as a medical device, the pain level calculated is preferably displayed in a manner that is easy for the user to understand as needed. In addition, the numerical value of the error that can be allowed as an appropriate λ value can be appropriately determined by those skilled in the art based on each case. For example, it can be determined in a manner that is within the range of a specific λ value obtained through cross-validation (for example, 0.0001 to 0.01 or 0.00001, 0.0001, 0.001, 0.01, etc. or any range therebetween). In addition, the optimal λ is a value set or calculated in a manner that minimizes the error by the least squares method, and can be easily determined by those skilled in the art by any known method.

[0241] In one embodiment, the present invention provides an apparatus for discriminating or classifying pain experienced by an estimated subject based on the subject's electroencephalogram (EB) data. The apparatus includes: a) a model data acquisition unit that acquires model EB data or analysis data thereof corresponding to a model stimulus intensity; b) a model feature extraction unit that extracts model EB features from the EB data or analysis data; c) a regression model generation unit that sets a target pain level, subjects the model EB features and the pain level to sparse model analysis, calculates an appropriate λ (preferably an optimal λ), determines parameters (partial regression coefficients) of the model EB features and an algorithm constant (intercept) corresponding to the appropriate λ (preferably the optimal λ), and generates a regression model; d) a measurement data acquisition unit that acquires (measurement) EB data or analysis data of the estimated subject; e) a measurement feature extraction unit that extracts measurement EB features from the measurement EB data or analysis data; f) a pain level calculation unit that fits the measurement EB features to the regression model to calculate the corresponding pain level; and g) a pain level display unit that displays the pain level as needed.

[0242] In another embodiment, the present invention provides a program that causes a computer to execute a method for discriminating or classifying pain experienced by an estimated subject based on the subject's electroencephalogram (EB) data. The method executed by the program includes the following steps: a) obtaining model EB data or analysis data corresponding to the stimulus intensity used in the model; b) extracting model EB features from the EB data or analysis data; c) setting a target pain level, subjecting the model EB features and the pain level to a sparse model analysis, determining an appropriate λ (preferably an optimal λ), determining parameters (partial regression coefficients) and an algorithm constant (intercept) for the model EB features corresponding to the appropriate λ (preferably the optimal λ), and generating a regression model; d) obtaining (measurement) EB data or analysis data of the estimated subject; e) extracting measurement EB features from the measurement EB data or analysis data; f) fitting the measurement EB features to the regression model to calculate the corresponding pain level; and g) displaying the pain level as needed. Alternatively, the present invention provides a recording medium storing the program. Representatively, step a) is implemented by A) the data acquisition unit for model, step b) is implemented by B) the brain wave feature extraction unit for model, step c) is implemented by C) the regression model generation unit, step d) is implemented by the data acquisition unit for measurement (which may also be the same as the brain wave data acquisition unit), step e) is implemented by the brain wave feature extraction unit for measurement (which may also be the same as the brain wave feature extraction unit for model), step f) is implemented by the pain level calculation unit (which may also have the function of the regression model generation unit, also called the pain classification value generation unit), and step g) is implemented by the pain level display unit.

[0243] In another embodiment, the present invention provides a method for estimating the pain experienced by an estimated object based on the brain waves of the estimated object, the method comprising the following steps: a) providing an approximate amount of at least one of the quantitative and qualitative levels of pain obtained by importing brain wave feature quantities into a sparse model analysis; b) obtaining brain wave data or its analysis data from the estimated object; c) extracting brain wave feature quantities from the brain wave data or its analysis data; and d) estimating or discriminating the pain level of the estimated object based on the approximate amount according to the brain wave feature quantities.

[0244] When performing pain discrimination on a subject, it is preferred that, rather than collecting data and creating a regression model from scratch using that data, the subject's real-time data be substituted into a regression model previously created based on model data and / or clinical data for discrimination / estimation. In this case, it is preferred that a plurality of strong and weak painful stimuli (e.g., 10 times x 2 types) are initially presented using reference stimuli, these stimuli are estimated using the discrimination model, and the parameters of the characteristic quantities (partial regression coefficients) and the algorithm constant (intercept) are calibrated while confirming accuracy.

[0245] Furthermore, the data acquisition step is very important when creating a model. It is preferable to acquire data that is as diverse as possible in terms of the number of adults and age groups and that is also universal for clinical populations.

[0246] In the present invention, a regression model can be used to calculate (fit) the pain level, thereby "estimating" or "discriminating" the subject's pain. In this case, a pain classification value can also be generated based on the regression model. It can be appreciated that knowing the pain level and intensity can achieve the following benefits: It allows for surgery to be performed without causing intense stimulation, or allows for objectively determining the therapeutic effects of analgesics, etc.

[0247] In this specification, a "pain classification value" refers to brainwave data (e.g., amplitude) or its analyzed value or range determined to classify pain types. In this disclosure, the component, device, or apparatus that generates a "pain classification value" (and therefore predicts pain) is sometimes referred to as a "pain classifier" or "pain predictor." In this disclosure, a stimulus estimation target is determined based on a regression model curve obtained by plotting the stimulus intensity or the subjective pain level corresponding to the stimulus intensity and fitting it to a regression model, using, for example, points where a certain change is observed. Once a pain classification value is generated, it can be calibrated and improved. Pain classification values ​​are sometimes referred to as pain classifiers, pain predictors, etc., which are all synonymous. Using a "pain classification value" allows for distinction between "is the change within the severe pain level" or "is the change, deviating from the severe pain level, exhibiting a low pain level?" If there is a deviation reaction that exceeds the fluctuation within the severe pain level, the pain classification value of the present invention can be used to identify the fluctuation within the severe pain level. If it is a fluctuation within the severe pain level, it is not an error and can be identified. If it exceeds the level, it can be treated as a deviation reaction.

[0248] The following diagram illustrates how to calculate the pain level: Figure 12 ).

[0249] In the step (S200) of obtaining brain wave data for the model or its analysis data corresponding to the stimulation intensity for the model (presentation of pain stimulation for discriminative model making, S100) as step a), the estimated object is stimulated with stimulation of multiple levels (intensity or size) (such as cold and warm stimulation, electrical stimulation, etc.) to obtain brain waves. The number of types of stimulation intensity can be the number required for the production of the function pattern. For example, there must usually be at least three types: weak, medium, and strong. It can also be one or two types. Sometimes, it is possible to apply to sparse modeling by combining with information obtained in advance, so it is not necessarily necessary to have this number of types. On the other hand, in the case of a new application, it may be advantageous to stimulate with at least three, preferably four, five, six or more levels of stimulation. If there are three types, weak, medium, and strong can be distinguished, so it is preferred. If there are more, the function pattern can be known in more detail, so it can be said to be ideal, but it is not limited to this. Here, because the burden on the estimated object should be minimized, it is preferred that the number of stimulation intensities that are highly invasive to the estimated object (in other words, the intensity that the subject cannot tolerate) is the minimum or zero. On the other hand, highly invasive stimulation to the estimated object may be necessary for a more accurate fit, so the minimum number can be obtained according to the purpose. For example, the number of types of such highly invasive levels can be at least 1, at least 2, or at least 3, and can also be 4 or more if the estimated object can tolerate it. Brain wave data or its analysis data also refers to brain activity data, brain activity amount, etc. For example, it includes amplitude data ("EEG amplitude"), frequency characteristics, etc. Such brain wave data can be obtained using any method known in the art. Brain wave data can be obtained by measuring the electrical signal of the brain wave, and expressed as amplitude data using potential (which can be expressed in μv, etc.). Frequency characteristics can be expressed using power spectrum density, etc.

[0250] In a preferred embodiment, in order to implement the present invention, it is preferred that the brain wave data be recorded using a simple method such as 1) using as few electrodes as possible (about 2), 2) avoiding the scalp with hair as much as possible, and 3) being able to record even when the patient is asleep. However, the number of electrodes can be increased as needed (for example, 3, 4, 5, etc.).

[0251] Step b) is a step of extracting model-use electroencephalogram feature quantities from the electroencephalogram data or its analysis data. The brain wave characteristic quantities can use the average amplitude (Fz, Cz, C3, C4), Fz: δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), γ (30-100Hz), Cz: 5 frequency bands <δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), γ (30-100Hz) >, C3: 5 frequency bands <δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), γ (30-100Hz) >, C4: 5 frequency bands <δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), γ (30-100Hz) >, etc. The above-mentioned frequency power, etc. can also be used.

[0252] Step c) is a step of setting the pain level as the target, importing the model brain wave feature and the pain level into the sparse model analysis, finding the appropriate λ (preferably the optimal λ), determining the parameters (partial regression coefficients) of the model brain wave feature corresponding to the appropriate λ (preferably the optimal λ) and the constant (intercept) of the algorithm, and generating a regression model. This is a step of performing the so-called sparse model analysis (S300). Here, the pain level is set, and a regression model (pain classifier / predictor (model regression formula)) is prepared using the brain wave feature obtained in step b) (S400). The regression model can be made using any method known in the art. As such a specific analytical method, there is LASSO, which solves the following optimization problem suitable for such a model.

[0253] When carrying out sparse modeling, determine the appropriate λ (preferably optimal λ) value used in penalty term by cross validation, and determine the parameter (partial regression coefficient) of characteristic quantity and the constant (intercept) of algorithm.Here, " λ value " refers to the hyperparameter that plays a role in regularization, for making the fitness of model smooth, improving generalization ability.Regularization has L1 regularization and L2 regularization, and in LASSO, it uses L1 regularization.λ coefficient is positive coefficient, determines appropriate (preferably optimal) solution by cross validation.As a concrete example, the LASSO function of MATLAB solves following mathematical formula by cross validation.

[0254] [Mathematical formula 2]

[0255] Min(Dev(β0,β)+λΣ|βj|)

[0256] Min: Minimize

[0257] Deviation: Deviation (the deviation of the estimated value of the regression model using the intercept β0 and the regression coefficient β from the observed value)

[0258] N: number of samples

[0259] λ: Regularization parameter with positive value

[0260] Step d) is a step of obtaining the (measurement) electroencephalogram data of the estimation subject or its analysis data. The measurement data are obtained in the same manner as in step c) (S500).

[0261] Step e) is a step of extracting measurement electroencephalogram feature quantities from the measurement electroencephalogram data or the analysis data. Step e) extracts measurement electroencephalogram feature quantities using the same method as step b). Extracting feature quantities presupposes that feature quantity extraction is performed through electroencephalogram signal processing (such as filtering) and secondary processing (signal processing).

[0262] Step f) is a step of fitting the measured electroencephalogram feature quantity to a regression model to calculate the corresponding pain level. Here, the measured electroencephalogram feature quantity can be substituted into the regression model to calculate a value, and the pain level corresponding to the value is classified or calculated (S600).

[0263] Step g) is a step of displaying the pain level as needed. The calculated pain level can be displayed numerically and / or graphically, or provided as an audio prompt. For the same subject, a step of inheriting or updating the classification value using previous classification value data may also be included.

[0264] In the apparatus of the present invention, A) the model data acquisition unit (which obtains model electroencephalogram data or analysis data corresponding to the stimulation intensity used in the model) is configured to implement step a). Specifically, the unit has means and / or functions capable of providing multiple stimulation intensities, is configured to apply such stimulation to a subject, and is further configured to obtain electroencephalogram data of an estimated subject. In addition to implementing step a), the electroencephalogram data acquisition unit may also have other functions (e.g., step d).

[0265] B) The model feature extraction unit (which extracts model electroencephalogram feature values ​​from the electroencephalogram data or its analysis data) is configured to obtain model feature values. In addition to performing step b), the model feature extraction unit may also have other functions (e.g., step e)).

[0266] C) Regression model generation unit (sets the target pain level, imports the model brain wave feature and the pain level into the sparse model analysis, finds the appropriate λ (preferably the optimal λ), determines the parameters (partial regression coefficients) of the model brain wave feature corresponding to the appropriate λ (preferably the optimal λ) and the algorithm constant (intercept) and generates a regression model), which can have the function of generating a regression model. Usually, step c) is implemented by C) regression model generation unit, and step f is implemented as the case may be. These two functions can be implemented by different devices, components, CPUs or terminals, etc., or they can be implemented as a part. Usually, the program for implementing these calculations is installed or can be installed in one CPU or computing device.

[0267] Figure 4 for Figure 12 An example of more detailed steps of sparse modeling illustrated in the embodiment.

[0268] In S1000 , data is input, and feature value data and pain level data are input.

[0269] In S2000 , data segmentation is performed, where data is segmented into learning data and test data. The learning data is used for model determination, and the test data is used for testing model accuracy.

[0270] In S3000 , an appropriate (preferably optimal) λ value is determined by cross-validation using the learning data (a ten-fold cross-validation is shown as an example in the figure) (eg, LASSO analysis).

[0271] In S4000 , the parameters of the feature value (partial regression coefficient) and the constant (intercept) of the regression model are determined.

[0272] In S5000, the pain level estimate for the test data is compared with the actual pain level. For example, an existing regression model estimates the pain level as follows: strong pain ≥ 0.3 (0.3 or above), weak pain < 0.3 (less than 0.3). Therefore, a value of ≥ 0.3 is "2," and a value of < 0.3 is "1." Here, the actual pain level is also expressed as "strong = 2" and "weak = 1." Therefore, the two are compared. If they match, the correct answer is obtained, and the discrimination accuracy is calculated.

[0273] In S6000, the discrimination accuracy (%) is calculated. Return to S2000 from S6000 and repeat several times ( Figure 4 1000 times) calculation accuracy.

[0274] Figure 13A schematic diagram of the device of the present invention is recorded in . The embodiment of the measuring device is described therein. The reference stimulation unit 1000 is provided together with the electroencephalogram data acquisition unit 2000 or separately. Regarding the actual pain discrimination / estimation process (solid line), the electroencephalogram data acquisition unit implements step a) of obtaining electroencephalogram data from the electroencephalogram 2500). The electroencephalogram feature extraction unit 2600 extracts electroencephalogram feature quantities from the original electroencephalogram data. In the pain level discrimination / estimation unit 3200, the pain level is fitted or estimated using the discrimination model obtained in advance by the sparse model analysis of the discrimination model generation unit 3000. When the confirmation and / or correction process (dashed arrow) of the discrimination algorithm carried out, the electroencephalogram data synchronized with the stimulation issued from the reference stimulation unit 1000 to the object (1500) is obtained by the electroencephalogram data acquisition unit 2000. In addition, the reference stimulation unit 1000 sends the reference stimulation intensity level data to the (pain level) discrimination / estimation unit 3200 for the correction of the discrimination algorithm. The estimated value of the pain level determination / estimation unit 3200 is visualized by the pain visualization unit 4000. For example, the estimated value is classified based on a threshold value (0 = weak, 1 = strong).

[0275] Figure 14 This is a block diagram illustrating the functional configuration of a pain identification and classification system 5100 according to one embodiment (note that several components in this configuration diagram are arbitrary and may be omitted). System 5100 includes an electroencephalogram measurement unit 5200, which internally or externally includes an electroencephalogram recording sensor 5250 and, as needed, internally or externally includes an electroencephalogram amplifier 5270. A pain identification / estimation unit 5300 performs signal processing and identification / estimation of pain using a regression model. Within the pain identification / estimation unit 5300, an electroencephalogram signal processing unit 5400 processes electroencephalogram signals (and, if necessary, extracts electroencephalogram features through an electroencephalogram feature extraction unit 5500). Pain is identified / estimated by a pain level identification / estimation unit 5600, and (if necessary) visualized by a pain level visualization unit 5800. The pain level discrimination / estimation unit 5600 is equipped with an algorithm determined by the discrimination model production unit using an existing database and sparse model analysis to discriminate / estimate the real-time pain level. In addition, a stimulation device unit 5900 is provided inside or outside, and the stimulation device unit 5900 has a reference stimulus presentation device unit (terminal) 5920, which helps to discriminate the patient's pain level. The stimulation device unit also includes a reference stimulus generating unit 5940. A reference stimulus level visualization unit 5960 may also be provided as needed. The sparse model analysis of the present invention or the discrimination regression model obtained by the sparse model analysis is stored inside the discrimination unit. Figure 4 The specific process is shown.

[0276] Figure 15 The situations where sparse modeling is involved are shown in more detail.

[0277] As shown above, the pain discrimination and classification system 5100 includes an electroencephalogram measurement unit 5200 and a pain discrimination / estimation device unit 5300, and, if necessary, a stimulation device unit (reference stimulation unit) 5900. The pain level discrimination / estimation unit 5600 is implemented, for example, by a computer having a processor and memory. In this case, when the processor executes the program (discrimination algorithm) produced by the discrimination model generation unit 3000, the pain level discrimination / estimation unit 5600 causes the processor to function as an electroencephalogram amplification unit 5270, an electroencephalogram signal processing unit 5400, (if necessary) a pain level discrimination / estimation unit 5600, and (if necessary) a pain level visualization unit 5800, as needed. The processor can also be configured to produce and visualize reference stimulation as needed. In addition, the system 5100 or device 5300 of the present invention can also be implemented, for example, by a dedicated electronic circuit. The dedicated electronic circuit can be a single integrated circuit or multiple electronic circuits. The electroencephalogram data acquisition unit and the pain classification value generation unit can also adopt the same structure as the pain estimation device.

[0278] The electroencephalogram measuring unit 5200 obtains a plurality of electroencephalogram data from an estimation subject by performing multiple electroencephalogram measurements via an electroencephalogram meter (electroencephalogram recording sensor 5250). The estimation subject is a living body whose electroencephalogram changes due to pain, and is not limited to humans.

[0279] The pain level discrimination / estimation unit 5600 generates or stores a regression model for discrimination / estimation, created by the discrimination model generation unit 3000, either internally or externally. The discrimination algorithm (regression model) generated through sparse model analysis is used to estimate or classify pain intensity based on the amplitude of multiple electroencephalogram (EEG) data. Specifically, the pain level discrimination / estimation unit 5600 can generate or store a regression model for estimating or classifying a subject's pain based on EEG data.

[0280] The electroencephalogram sensor 5250 uses electrodes on the scalp to measure electrical activity within the subject's brain. The electroencephalogram sensor 5250 then outputs electroencephalogram data as the measurement result. The electroencephalogram data can be amplified as needed.

[0281] Next, the processing or method of the device configured as above will be described. Figure 12 This is a flowchart showing a series of processes. S100 to S400 are involved in the scheme of generating a model regression formula. In S400, a pain level discrimination / classification device is generated.

[0282] By referring to the stimulation part 1000 (refer to Figure 13), applying stimulation of multiple levels (sizes) of stimulation intensity to the subject (S100).

[0283] Next, the electroencephalogram data (e.g., electroencephalogram amplitude reference data, such as amplitude data) is obtained (S200). Figure 13 In general, it is performed by the electroencephalogram data acquisition unit 2000. Figure 14 For example, the electroencephalogram measuring unit 5200 performs electroencephalogram measurements multiple times via the electroencephalogram (electroencephalogram recording sensor) 5250 to obtain multiple electroencephalogram data from the estimated subject, and uses the data as electroencephalogram data (e.g., amplitude data). The electroencephalogram measuring unit 5200 may also perform electroencephalogram measurements at multiple times. Figure 13 ) performs sparse model analysis (S300). Fitting is performed based on the sparse model analysis, and if the regression coefficient is judged to be an appropriate value, the pain level discrimination / estimation unit 3200 (refer to Figure 13 ) can use the regression model (also called discriminant model) to classify pain levels (S400). After the regression model is generated, the pain level can be classified by referring to the stimulation part 1000 (refer to Figure 13 ) for correction (calibration).

[0284] The sparse modeling of the present invention can also be expressed as follows. That is, in one embodiment, the present invention provides an algorithm for classifying pain experienced by an estimated subject based on the subject's brain waves, comprising the following steps: a) stimulating the estimated subject with multiple levels of stimulation intensity; b) obtaining brain wave data of the estimated subject corresponding to the stimulation intensity; c) extracting brain wave feature quantities from the brain wave data; d) fitting the feature quantities to the sparse model analysis to approximate the quantitative and qualitative levels of pain, and estimating and discriminating the pain level; and finally, algorithmic determination (including feature quantity coefficients and intercepts).

[0285] Alternatively, the present invention provides a device for discriminating or classifying the pain experienced by an estimated object based on the brain waves of the estimated object, the device comprising: a) a stimulation unit for stimulating the estimated object with multiple levels of stimulation intensity; b) a data acquisition unit for obtaining brain wave data or analysis data thereof corresponding to the stimulation intensity of the estimated object; c) a brain wave feature extraction unit for extracting brain wave feature quantities from the brain wave data or analysis data thereof; and d) a pain level discrimination / estimation unit for fitting the feature quantity to the sparse model analysis to approximate at least one of the quantitative level and qualitative level of the pain, and estimating or discriminating the pain level.

[0286] In another embodiment, the present invention provides a program that causes a computer to execute a method for discriminating or estimating pain experienced by an estimated subject based on the subject's brain waves. The method executed by the program includes the following steps: a) stimulating the estimated subject with multiple levels of stimulation intensity; b) obtaining brain wave data or analysis data corresponding to the stimulation intensity; c) extracting brain wave features from the brain wave data or analysis data; and d) fitting the features to the sparse model analysis to approximate at least one of the quantitative and qualitative levels of pain, thereby estimating or discriminating the pain level. Alternatively, the present invention provides a recording medium storing the program.

[0287] By using the present invention, it is possible to have the generality to cope with a variety of pain types and pain patterns. In addition, the present invention provides a known discrimination algorithm produced or installed in the above text, so that it is not only effective for discriminating the pain level of a known subject, but also can predict and classify the pain discrimination of an unknown subject. In this case, the parameters (regression coefficient, intercept, threshold, etc.) of the discrimination / estimation algorithm determined based on the pain database of the known subject and the real-time pain feature of the unknown subject are used to obtain an estimated value, and the pain level of the estimated value is converted using a specific threshold. For example, by using a specific regression model with the help of the LASSO analytical method, the following known discrimination algorithm can be adapted to the unknown measurement data of the subject.

[0288] [Mathematical formula 3]

[0289] Est i =∑(Test i ×β j )+C

[0290] i. Low pain level: Est i <Threshold

[0291] ii. High pain level: Est i Threshold

[0292] Est: estimated value

[0293] Test: Determine the characteristic quantity

[0294] β: partial regression coefficient

[0295] C:Intercept

[0296] i: number of observations

[0297] j: number of characteristic quantities

[0298] Furthermore, the present invention provides a discrimination algorithm, thereby enabling classification or prediction of pain levels of known and unknown subjects using as few electroencephalogram feature quantities as possible in order to discriminate the pain level of a known subject.

[0299] Examples of such feature quantities include the following.

[0300] (Representative feature quantity)

[0301] Average amplitude (Fz, Cz, C3, C4)

[0302] ·Fz: δ(1-3Hz), θ(4-7Hz), α(8-13Hz), β(14-30Hz), γ(31-100Hz)

[0303] ·Cz: 5 frequency bands <δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), γ (31-100Hz) >

[0304] ·C3: 5 frequency bands <δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), γ (31-100Hz) >

[0305] ·C4: 5 frequency bands <δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), γ (31-100Hz)>, etc.

[0306] The above-mentioned frequency power etc. can also be used.

[0307] In another embodiment, the present invention can provide a technique for discriminating / estimating pain levels without having to create a regression model each time. In this case, the method of the present invention includes the following steps: c) providing a regression model for the pain level of the estimated subject using a sparse model analysis; d) obtaining (measurement) electroencephalogram data or analysis data of the estimated subject; e) extracting measurement electroencephalogram features from the measurement electroencephalogram data or analysis data; f) fitting the measurement electroencephalogram features to the regression model to calculate the corresponding pain level; and g) displaying the pain level as needed. The regression model for the pain level of the estimated subject using a sparse model analysis can be prepared by precalculating a sparse model analysis regression model for the pain level of the estimated subject and storing it in a suitable storage device or recording medium. Alternatively, such a regression model can be provided as a model that has been standardized to a certain degree. In such cases, sparse model analysis need not necessarily be performed for the same estimated subject; a regression model calculated for subjects of similar age groups and / or similar or identical attributes can also be used. In this case, a regression model generation method as described in the following <Regression Model Generation> may be used.

[0308] Alternatively, the present invention provides a device for discriminating or classifying pain experienced by an estimated subject based on the subject's electroencephalogram (EEG) waves, pre-provided with a regression model (also called a discriminant model). The device comprises: c) a regression model providing unit that provides a regression model for sparse model analysis of the subject's pain level; d) a measurement data obtaining unit that obtains (measurement) EEG data or its analysis data of the subject; e) a measurement feature extraction unit that extracts measurement EEG feature values ​​from the measurement EEG data or its analysis data; f) a pain level calculation unit that fits the measurement EEG feature values ​​to the regression model to calculate the corresponding pain level; and g) a pain level display unit that displays the pain level as needed.

[0309] Alternatively, the present invention provides a program that enables a computer to execute a method of discriminating or classifying the pain of an estimated object based on the brain waves of the estimated object. The method provided by the program includes the following steps: c) a step of providing a regression model of a sparse model analysis of the pain level of the estimated object; d) a step of obtaining (measurement) brain wave data or its analysis data of the estimated object; e) a step of extracting a measurement brain wave feature from the measurement brain wave data or its analysis data; f) a step of fitting the measurement brain wave feature to a regression model to calculate the corresponding pain level; and g) a step of displaying the pain level as needed. Alternatively, the present invention provides a recording medium storing the above-mentioned program. The regression model can be calculated by any method as described in the section <Regression model generation>, but can also be generated by another method or pre-generated ( Figure 12 , to the case where the steps of S400 are performed separately).

[0310] Step d) is the step of obtaining EEG data (e.g., amplitude data) of the subject to be estimated (S500). This step involves obtaining EEG data from the subject to be measured, regardless of whether the subject has been stimulated or treated. Any method capable of obtaining EEG data may be used. The same method as used in step a) can be used, and the same method is typically used.

[0311] Then, in the present invention, in step e), the brain wave feature is extracted. The same or different method as that used in step b) can be used, but the same method is usually used.

[0312] Step f) is a step of fitting the measured brain wave feature quantity to a regression model to calculate the corresponding pain level (pain classification, S600). The regression model is associated with the pain level of the estimated object and is called a "pain classifier" or "pain predictor". For example, when the brain wave amplitude related to pain shows a decreasing pattern and the pain classification value of pain is classified into "strong pain" and "weak pain", when brain wave data (such as amplitude data) lower than the value is detected, it is classified as "strong pain", and when large brain wave data (such as amplitude data) is detected, it is classified as "weak pain". For example, when the value of the pain classifier shows a standardized brain wave absolute amplitude of "0.7", after the brain wave amplitude data recorded online is absolutized and standardized according to existing data, when it shows "0.8", it is classified as feeling "weak pain", and when it shows "0.2", it is classified as feeling "strong pain".

[0313] In one embodiment, the brainwave data (eg, amplitude data) may be fitted to the regression model using an original average value or a normalized average value, such as an average value between 15 seconds and 120 seconds.

[0314] based on Figure 13 The solution is described below. Figure 13In addition to the pain level discrimination / estimation unit 3200, the electroencephalogram data acquisition unit 2000 is also referenced. In this case, as described in the "Regression Model Generation" section, electroencephalogram data can be obtained from the subject 1500 via the electroencephalogram 2500. That is, the electroencephalogram data acquisition unit 2000 is configured to be connected to the subject (1500), and the electroencephalogram data acquisition unit 2000 is configured to have an electroencephalogram or to be connected to the electroencephalogram (2500), and the electroencephalogram is connected to the subject (1500) or can be connected to the subject (1500) to obtain electroencephalogram data obtained from the subject (1500). The pain level discrimination / estimation unit 3200, also known as the pain level calculation unit, stores the regression model generated by the discrimination model generation unit 3000, or is configured to receive a separately generated regression model and is configured to refer to it as needed. Such a connection configuration can be either wired or wireless. For model calibration, artificial pain stimulation level data is sent from the reference stimulation unit 1000 to the pain level discrimination estimation (classification) unit 3200. The discrimination model generation unit 3000 imports or calculates a discrimination model implemented by the sparse model analysis of the present invention.

[0315] Figure 14 This is a block diagram illustrating the functional structure of a pain level discrimination estimation or regression model generation system 5100 according to one embodiment. The system 5100 includes an electroencephalogram measurement unit 5200, which internally or externally includes an electroencephalogram recording sensor 5250, and, as needed, an electroencephalogram amplifier 5270, which internally or externally includes an electroencephalogram amplifier 5270. A pain level discrimination / estimation device unit 5300 processes pain signals and discriminates / estimates pain. In the pain level discrimination / estimation device unit 5300, an electroencephalogram signal processing unit 5400 processes the electroencephalogram signal. A pain discrimination / estimation device unit 5600 stores a regression model (as needed) and, based on this, discriminates and estimates pain. The pain is visualized (as needed) by a pain level visualization unit 5800. Furthermore, a stimulation device unit 5900 is internally or externally provided, and the stimulation device unit 5900 includes a reference stimulus presentation device unit (terminal) 5920 to assist in calibrating the patient's pain level discrimination / estimation device. The stimulation device section (as needed) has a reference stimulation generating section 5940 .

[0316] As shown above, the pain level discrimination / estimation system 5100 includes an electroencephalogram measurement unit 5200 and a pain discrimination / estimation device unit 5300. The pain discrimination / estimation device unit 5300, which stores the regression model, is implemented, for example, by a computer having a processor and memory. In this case, when the processor executes a program stored in the memory, the pain discrimination / estimation device unit 5300 causes the processor to function as an electroencephalogram amplification unit 5270, an electroencephalogram signal processing unit 5400, a pain discrimination / estimation unit 5600 (as needed), and a pain level visualization unit 5800 (as needed). The processor can also generate and visualize reference stimuli as needed. In addition, the system 5100 or device 5300 of the present invention can be implemented, for example, by a dedicated electronic circuit. The dedicated electronic circuit can be a single integrated circuit or multiple electronic circuits. The electroencephalogram data acquisition unit and the pain classification value generation unit can also adopt the same structure as the pain level estimation device.

[0317] The electroencephalogram measurement unit 5200 obtains multiple sets of electroencephalogram data from an estimated subject by performing multiple electroencephalogram measurements using an electroencephalogram (electroencephalogram recording sensor 5250). The estimated subject is a living organism whose electroencephalogram changes due to pain. This entity is any living organism with pain-sensing nerves (e.g., vertebrates such as mammals and birds (including livestock and pets)), and is not limited to humans.

[0318] The pain level identification / estimation unit 5600 identifies or classifies the pain level based on the amplitude of the plurality of electroencephalogram data using a regression model. In other words, the pain identification / estimation unit 5600 identifies or classifies the subject's pain based on the electroencephalogram data using a regression model.

[0319] The electroencephalogram (EEG) sensor 5250 uses electrodes on the scalp to measure electrical activity within the subject's brain. The EEG sensor 5250 then outputs EEG data as the measurement result. The EEG data can be amplified as needed.

[0320] Figure 15 The accompanying description is shown in Figure 14 The following table shows the relationship between the pain discrimination / estimation device and sparse model analysis, which is explained in the basic block diagram. The discrimination model creation unit 3000 is shown here, and a basis for estimating the pain level based on the discrimination model (also called a regression model) obtained through sparse model analysis is provided. The pain discrimination correction unit 5700 is an optional component, but here, by applying "strong" and "weak" reference stimuli multiple times and correcting the coefficients of the feature value based on the discrimination accuracy. In the pain level visualization unit 5800, a threshold value can be set and the pain level can be expressed in several stages.

[0321] Next, the processing or method of the device configured as described above will be described. Figure 12 This is a flowchart showing a series of processes. S400 to S600 can be included in this solution. These steps occur after the generation of a regression model (also called a discriminant model, also known as a pain classifier / pain predictor) in S400. Alternatively, if a regression model is already available (for example, if it has been previously obtained and saved), this process begins at S400.

[0322] This regression model can also be generated by a different discriminant model generating unit 3000 (see Figure 13 ) is created and stored in advance in the pain level determination / estimation unit 3200 (refer to Figure 13 ), the pain level discrimination / estimation unit 3200 may also be configured to receive value data. Alternatively, if the discrimination model generation unit 3000 is provided, the value data may be stored in the generation unit or in a separate recording medium. Alternatively, the value data may be received via communication.

[0323] Next, electroencephalogram data is obtained from the subject (S500). This electroencephalogram data can be obtained using the same technology and implementation as described in S200, but does not need to use the same apparatus or device as S200 and may be different or the same.

[0324] Next, the EEG data (e.g., amplitude data) obtained in S500 is fitted to a regression model generated by sparse model analysis, and the pain level corresponding to the EEG data is classified (S600). This pain classification can also be configured to display specific text (such as strong pain, weak pain, etc.) or a voice when a predetermined value is output in advance, or the actual value and the regression model can be displayed side by side to enable the user (clinician) to study it. Figure 16 An exemplary embodiment is shown. Figure 16 As shown in , the regression model generated by sparse model analysis is stored in the discriminant model generation unit 3000 and used in the pain discrimination / estimation device. The results obtained by the sparse model analysis of the present invention are imported into the discriminant model generation unit 3000. The reference stimulation unit 5900 determines the individual's pain sensitivity and sends data for calibrating the discrimination algorithm. The discriminant / estimation unit 5600 uses the generated model to discriminate / estimate the pain level. The visualization unit 5800 can arbitrarily determine the threshold value and express pain in multiple stages.

[0325] <Regression model generation>

[0326] In another embodiment, the present invention provides a technique for making a regression model. Here, the present invention provides a method for generating a regression model (for discriminating or estimating the pain of the estimated object based on the brain wave of the estimated object). The method includes the following steps: a) obtaining model brain wave data or its analysis data corresponding to the stimulation intensity used by the model; b) extracting the model brain wave feature from the brain wave data or its analysis data; and c) setting the pain level as a target, importing the model brain wave feature (independent variable) and the pain level (dependent variable) into the sparse model analysis, finding the appropriate λ (preferably the optimal λ), determining the parameters (partial regression coefficients) of the model brain wave feature corresponding to the appropriate λ (preferably the optimal λ) and the algorithm constant (intercept) and generating a regression model.

[0327] Alternatively, the present invention provides an apparatus for generating a regression model for discriminating or classifying pain experienced by an estimated subject based on the subject's electroencephalogram (EBV). The apparatus comprises: a) a model data acquisition unit that acquires model EBV data or analysis data corresponding to the stimulus intensity used in the model; b) a feature extraction unit that extracts model EBV features from the EBV data or analysis data; and c) a regression model generation unit that sets a target pain level, subjects the model EBV features and the pain level to sparse model analysis, determines an appropriate λ (preferably an optimal λ), determines parameters (partial regression coefficients) of the model EBV features and an algorithm constant (intercept) corresponding to the appropriate λ (preferably the optimal λ), and generates a regression model.

[0328] Alternatively, the present invention provides a program that causes a computer to execute a method for generating a regression model (for discriminating or classifying the pain of an estimated object based on the brain waves of the estimated object). The method executed by the program includes the following steps: a) obtaining model brain wave data or its analysis data corresponding to the stimulation intensity used by the model; b) extracting model brain wave features from the brain wave data or its analysis data; and c) setting a target pain level, importing the model brain wave features and the pain level into a sparse model analysis, finding an appropriate λ (preferably an optimal λ), determining the parameters (partial regression coefficients) of the model brain wave features corresponding to the appropriate λ (preferably the optimal λ) and the constant (intercept) of the algorithm, and generating a regression model. Alternatively, the present invention provides a recording medium storing the above-mentioned program.

[0329] based on Figure 13 This scheme is described. Figure 13In addition to the discriminant model generation unit 3000, the electroencephalogram data acquisition unit 2000 is also referenced. In this case, as described in the section "Regression Model Generation", the electroencephalogram data can be obtained from the subject via an electroencephalogram. That is, the electroencephalogram data acquisition unit 2000 is configured to be connectable to the subject 1500. The electroencephalogram data acquisition unit 2000 is configured to include an electroencephalogram or to be connectable to the electroencephalogram (2500), and the electroencephalogram is connected to the subject (1500) or is connectable to the subject (1500) to obtain the electroencephalogram data obtained from the subject (1500).

[0330] Furthermore, in the above embodiment, the pain severity value Pmax corresponding to the upper limit Amax of the EEG amplitude is set to 1, and the pain severity value Pmin corresponding to the upper limit Amin of the EEG amplitude is set to 0. However, this is not limiting. For example, the pain severity can also be expressed on a scale of 0 to 100. In this case, the pain level determination / estimation unit 3200 can simply estimate the pain severity value Px using the following equation.

[0331] [Formula 4]

[0332] Px=Pmax×(Ax-Amin) / (Amax-Amin)

[0333] Furthermore, Amax and Amin are electroencephalogram amplitudes, but may be estimated values ​​calculated using a regression model. In this case, they are as follows.

[0334] [Formula 5]

[0335] Px=Pmax×(Estx-Modmin) / (Modmax-Modmin)

[0336] Estx: Model estimation value calculated based on feature quantity

[0337] Modmin: Minimum value of the discriminant estimate determined by the existing model

[0338] Modmax: The maximum value of the discriminant estimate determined by the existing model

[0339] In addition, in the above, curve fitting is described as an example of generating pain classification values ​​by analyzing multiple brain wave data, but it is not limited to this. For example, a learning model (used to estimate the brain wave amplitude corresponding to a large stimulus based on the brain wave amplitude corresponding to a small stimulus) can be used to determine the value corresponding to the large stimulus. In this case, a large stimulus can be not applied to the estimated object, thereby reducing the physical burden of the estimated object. In addition, a predetermined value can be used as the upper limit of the brain wave amplitude. For example, the predetermined value is, for example, 50μV to 100μV, which can be determined based on experiments or experience. In this way, in conventional analysis, as a method of removing false signals, data from plus or minus 50μV to about 100μV is excluded. In the present invention, such false signal removal can also be implemented as needed.

[0340] In addition, the stimulation device section 5900 (see Figure 14 ) The stimulation applied to the estimated subject is not limited to electrical stimulation and thermal stimulation. Any type of stimulation may be applied as long as the magnitude of the pain felt by the estimated subject varies according to the magnitude of the stimulation.

[0341] Furthermore, some or all of the components of the pain estimation devices in the above-described embodiments may be implemented as a single system LSI (Large Scale Integration). For example, the pain discrimination estimation device unit 5300 may be implemented as a system LSI that includes the optional (electroencephalogram) measurement unit 5200 and the optional stimulation device unit 5900.

[0342] Furthermore, in each of the above-described embodiments, each component can be implemented using dedicated hardware or by executing a software program suitable for each component. Alternatively, each component can be implemented by a program execution unit, such as a CPU or processor, reading and executing a software program stored on a recording medium, such as a hard disk or semiconductor memory. The software that implements the pain estimation device and the like in each of the above-described embodiments is referred to in this specification as the aforementioned program.

[0343] That is, the program causes a computer to execute a method for discriminating or classifying the pain experienced by an estimated object based on the brain waves of the estimated object, the method comprising the following steps: a) a step of obtaining model brain wave data or its analysis data corresponding to the stimulation intensity used for the model; b) a step of extracting model brain wave feature quantities from the brain wave data or its analysis data; c) a step of setting a target pain level, importing the model brain wave feature quantities and the pain level into a sparse model analysis, obtaining an appropriate λ (preferably the optimal λ), determining the parameters (partial regression coefficients) of the model brain wave feature quantities and the algorithm constant (intercept) corresponding to the appropriate λ (preferably the optimal λ) and generating a regression model; d) a step of obtaining (measurement) brain wave data or its analysis data of the estimated object; e) a step of extracting measurement brain wave feature quantities from the measurement brain wave data or its analysis data; f) a step of fitting the measurement brain wave feature quantities to a regression model to calculate the corresponding pain level; and g) a step of displaying the pain level as needed.

[0344] Alternatively, a computer is caused to execute a method for discriminating or classifying the pain experienced by an estimated object based on the brain waves of the estimated object, the method comprising the following steps: a) stimulating the estimated object with multiple levels of stimulation intensity; b) obtaining brain wave data or analysis data thereof corresponding to the stimulation intensity of the estimated object; c) extracting brain wave feature quantities from the brain wave data or analysis data thereof; d) substituting the feature quantities into a sparse model analysis to approximate at least one of the quantitative level and qualitative level of pain and estimating or discriminating the pain level.

[0345] Alternatively, a computer is caused to execute a method for discriminating or classifying the pain experienced by an estimated object based on the brain waves of the estimated object, the method comprising the following steps: c) a step of providing a regression model for sparse model analysis of the pain level of the estimated object; d) a step of obtaining (measurement) brain wave data or its analysis data of the estimated object; e) a step of extracting a measurement brain wave feature quantity from the measurement brain wave data or its analysis data; f) a step of fitting the measurement brain wave feature quantity to a regression model to calculate the corresponding pain level; and g) a step of displaying the pain level as needed.

[0346] Alternatively, a computer is caused to execute a method for generating a regression model, wherein the regression model is used to discriminate or estimate the pain experienced by an estimated object based on the brain waves of the estimated object, the method comprising the following steps: a) a step of obtaining model brain wave data or its analysis data corresponding to the stimulation intensity used by the model; b) a step of extracting model brain wave feature quantities from the brain wave data or its analysis data; and c) a step of setting a target pain level, introducing the model brain wave feature quantities and the pain level into a sparse model analysis, obtaining an appropriate λ (preferably an optimal λ), determining the parameters (partial regression coefficients) of the model brain wave feature quantities and the algorithm constant (intercept) corresponding to the appropriate λ (preferably the optimal λ), and generating a regression model.

[0347] (Determination of instant pain)

[0348] In one embodiment, the present invention provides a method for distinguishing or evaluating pain, comprising the steps of comparing all or part of the brain wave data or its analysis data during the period of 2000 milliseconds from the earliest time point among the induced brain wave component, the initial event-related potential component, and 250 milliseconds after the application of the object stimulus with the brain wave data or its analysis data after the same time period after the application of the reference stimulus. In the past, pain was usually determined by observing the induced brain wave component (P100) to the P300 component. In the present invention, by also observing the area behind the mid-term event-related potential component P300, a detailed analysis of the pain can be performed, and the occurrence and stage of instantaneous pain can be distinguished using brain wave feature quantities that are difficult to capture synchronization in persistent pain. This can be said to be an important aspect.

[0349] In one embodiment, the determination criteria include whether the electroencephalogram data for a period of 2000 milliseconds from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds exhibits a persistence characteristic. It should be noted that if a potential component with a peak exists after the initial event-related potential component, it is determined to be a conventional P300.

[0350] In one embodiment, the brain wave data or the analysis data thereof includes all or part of the brain wave data or the analysis data thereof within a range of 2000 milliseconds from the mid-term time period.

[0351] In an embodiment of the present invention, the mid-term period generally corresponds to the event-related potential component, including the range that produces the P200, N200, and P300. Typically, the mid-term period may include values ​​within the range of 250 milliseconds to 600 milliseconds. Furthermore, the mid-term period may also continuously or comprehensively include values ​​within this range, or may also extend to 250-600 milliseconds or 300-600 milliseconds. Alternatively, if it is after the P200, it may be any range of values.

[0352] In one embodiment, all or part of the electroencephalogram data from the earliest time point of 2000 milliseconds among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds includes a range of at least 100 milliseconds. In one embodiment, the mid-term time period is a time width continuously covering from 300 milliseconds or earlier to around 600 milliseconds.

[0353] In one embodiment, if a statistically significant difference is observed within the range of at least 100 milliseconds, the data is considered to be persistent within the range of at least 100 milliseconds. Even if a difference is observed upon visual inspection but is not statistically significant, a difference may be considered to be present. However, in preferred embodiments, a statistically significant difference is advantageous. In another embodiment, if the data clearly show persistence within 250 to 2000 milliseconds, even if there are only sporadic significant differences, the data may be considered persistent.

[0354] In one embodiment, the comparison includes determining whether there is a duration of time during which the values ​​of the EEG data or its analysis data obtained by the subject stimulation differ from the values ​​of the EEG data or its analysis data obtained by the reference stimulation, and determining whether, if the duration of the difference exists, the values ​​return to the same value. If the duration of the difference exists and the values ​​do not return to the same value, it is determined that unpleasant pain is present. Specifically, for example, to identify the persistent component, verification is first performed using the following signal processing.

[0355] 1) After the stimulus is presented, whether there is a deviation from the reference stimulus activity. If the deviation occurs from the time of stimulus presentation, it is not a component related to the perception and / or cognition of pain, but may be a false signal. 2) Observe the brain wave waveform presented by each stimulus. The situation where the baseline of the brain wave itself is significantly shifted upward or downward, and disappears when the baseline returns to the baseline through linear correction, is not called a persistent component. 3) When low-frequency band components such as 0.02Hz, 0.01Hz, and 0.1Hz are cut off, the persistent component disappears with the duration. Therefore, in the case where it does not belong to 1 or 2 but such low-frequency band components can be cut off, it can also be determined as a persistent ERP. In the above process, after verifying the persistent component, such as Figure 19 As shown, in order to determine the starting point and duration of the persistent ERP, a time period was set, and statistical verification such as t-test and ANOVA was performed to determine the significant difference between the reference stimulus and the painful stimulus.

[0356] In one embodiment, the brain wave data or analysis data thereof is potential, duration or a combination thereof.

[0357] In a further embodiment, the pain discrimination is the discrimination of the discomfort of pain. Although not wishing to be bound by theory, the present invention has the potential to be able to identify whether it is "identifiable pain," that is, pain emanating from a specific body part or pain caused by external stimuli.

[0358] In one embodiment, the electroencephalogram data or the analyzed data thereof can be compared over the entire range of 2000 milliseconds from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds.

[0359] In a preferred embodiment, the method of the present invention further comprises analyzing the compared data using a sigmoid function fit. Specifically, the sample data (normalized within individuals) for all subjects are arranged in the order of no transient pain and with transient pain. A function (e.g., a sigmoid function and / or a step function) is created that approximates the sample data. This function approximation can be performed using the MATLAB code nlinfit, for example.

[0360] In one embodiment, it is characterized in that the discrimination is performed on the positive component of the brain wave data or its analysis data. In a specific embodiment, it is characterized in that when the positive component continues even after the mid-term time period, it is determined that there is pain. In particular, the delayed persistence effect is believed to reflect the uncomfortable pain that gradually appears later and is estimated to be a C-fiber nerve reflex.

[0361] Typically, mid-term event-related potentials, such as the P300, typically peak between 300 and 400 milliseconds. However, this may vary depending on the content and / or load of the information being processed. Therefore, generally speaking, a shift in the positive direction indicates a non-persistent nature, starting before 300 milliseconds and peaking between 300 and 400 milliseconds. However, as demonstrated in the present invention, this persistence unexpectedly persists, with some cases not returning to baseline even after 2000 milliseconds after stimulus presentation. This can be used as an indicator for detailed pain discrimination. The present invention has even observed instances where the response persisted for 2 seconds, despite not returning to baseline for an extended period of time. This is an unexpected observation; the discovery that persistent positive components are important indicators of pain is an unexpected finding. In particular, ERP components that persist beyond the mid-term indicate that even a single painful stimulus can cause discomfort, and that the discomfort persists consciously even after the stimulus disappears. Therefore, it is possible to discern acute pain, such as nociceptive pain, and the associated pain pressure.

[0362] Each step is described below.

[0363] The method of the present invention is described below using a schematic diagram ( Figure 27 ).

[0364] Step S10100: Data collection under reference stimulation

[0365] S10100 is an arbitrary step, which is a step of providing data that serves as a baseline (reference). In order to extract the brain wave feature quantity related to instantaneous pain, it is desired to set the type of pain stimulus and / or stimulus characteristics to be as similar as possible, and to minimize the difference in pain characteristics such as the presence or absence of pain or the intensity of pain. For example, in the case of high temperature stimulation, a reference stimulus used as a benchmark when evaluating pain, such as a stimulus that is painless or slightly painful, is presented multiple times, and brain wave data or its analysis data is collected. The reference stimulus is a standard (background) stimulus that is mixed with the object stimulus described later. In order for the subject to form a short-term memory of the standard stimulus, the frequency must be very high, accounting for about 70% of the total. Providing pre-acquired brain wave data or its analysis data or reading the stored data is also interpreted as being equivalent to the step of collecting the data.

[0366] Step S10200: Applying stimulus to the object and obtaining data

[0367] S10200 is a step of applying the object stimulus multiple times in a manner of randomly mixing it with the reference stimulus, or treating the naturally applied stimulus as the object stimulus, and collecting brain wave data or its analysis data relative to the stimulus. In this step, a certain stimulus (such as a stimulus that causes pain) is applied to the object, or the naturally occurring stimulus is treated as the object stimulus, and the brain wave data or its analysis data of the model or actual measurement object is measured or obtained. Regarding the application of the stimulus, in the case of a model system, various stimuli (such as cold and warm stimuli, electrical stimulation, etc.) are used to stimulate the estimated object, and the brain wave data corresponding to the stimulus intensity (also called brain activity data, brain activity amount, etc., including amplitude data (EEG amplitude), frequency characteristics, etc.) are obtained. Such brain wave data can be obtained using any method known in the art. The brain wave data can be obtained by measuring the electrical signal of the brain wave, and is represented by a potential as amplitude data, etc. (can be represented by μV, etc.). The frequency characteristics are represented by power spectrum density, etc. The brain wave data or its analysis data can be used to associate the difference points with conditional parameters (such as pain stimulation discomfort, etc.) based on appropriate methods. The conditional parameters include parameters related to stimulation and environment, such as stimulation type and stimulation presentation environment.

[0368] Step S10300: Compare the EEG data of the reference stimulus or its analysis data with the EEG data of the target stimulus or its analysis data in the whole or part of the period from the earliest time point of 2000 milliseconds among the evoked EEG component, the initial event-related potential component, and 250 milliseconds (reference Figure 19 )

[0369] The method determines the time point at which the reference stimulus deviates from the target stimulus within a range of 2000 milliseconds from the earliest of the evoked EEG component, the initial event-related potential component, and 250 milliseconds, and compares the EEG data or analyzed data from the two. In other words, the method compares the EEG data or analyzed data from the range of 2000 milliseconds from the earliest of the evoked EEG component, the initial event-related potential component, and 250 milliseconds with the EEG data or analyzed data from the same time period after the application of the reference stimulus. Preferably, the method determines whether the EEG data from the range of 2000 milliseconds from the earliest of the evoked EEG component, the initial event-related potential component, and 250 milliseconds exhibits a persistence characteristic. If the persistence characteristic is statistically confirmed, it can be determined that a specific type of pain exists, where the discomfort persists even after the painful stimulus physically disappears. This allows for discrimination between unpleasant and transient pain. In a preferred embodiment, the method measures the EEG data or analyzed data from the range of 2000 milliseconds from the mid-term period. Such measurements can be adjusted by any control mechanism included in the system and / or program implementing the method of the present invention. The mid-term time period can be set to 250 to 600 milliseconds, but due to individual differences and even differences within the same person, it can be appropriately calibrated or changed after a single measurement. The mid-term time period represents the range of the event-related potential component, which can be calculated using any method known in the art.

[0370] Difference judgment can be achieved through statistical methods (refer to Figure 19 ), such a method can be arbitrarily added to the system or program for implementing the present invention, or can be provided from the outside. For example, the duration of time during which the value of the brain wave data or its analysis data obtained by the object stimulation is different from the value of the brain wave data or its analysis data obtained by the reference stimulation is measured, and in the case of the existence of the different duration, whether it becomes the same value again from the different time, and in the case of the existence of the duration and not becoming the same value again, it is determined that there is uncomfortable pain, which can also be arbitrarily added to the system or program for implementing the present invention, or can be provided from the outside. The brain wave data or its analysis data is preferably potential, duration or a combination thereof, but other parameters can also be used, and the present invention can be constructed so that it can be arbitrarily selected.

[0371] Comparative data can also be further analyzed using Sigmoid function fitting. Using brain wave feature quantities, in order to match the conditions and divide them into two or three or more categories and make a discrimination / estimation model, one of the methods is to make a plot and fit (make it fit) to a suitable fitting function such as a Sigmoid function pattern. Fitting can be performed using any method known in the art. As such specific fitting functions, step functions, Boltzmann functions, double Boltzmann functions, Hill functions, logistic dose-response functions, Sigmoid Richards functions, Sigmoid Weibull functions, etc. can be cited, but are not limited to these functions. Among these, the standard logistic function is called the Sigmoid function, and the standard function or its deformation is general and preferred. When the regression coefficient of the fitting to a suitable function pattern such as the Sigmoid function pattern is above the specified value as needed, the threshold value for pain determination can be set based on the Sigmoid curve, etc. Here, in the case of an S-shaped curve, it can be generated based on the inflection point, but is not limited to this. If necessary, the pain classification value may be corrected (calibrated) so that the pain type and level classification are maximized. The threshold value can be used to calculate or classify the pain type and level and can be used to determine the treatment effect.

[0372] Therefore, in a specific embodiment, the association includes setting a difference in pain based on conditions, such as the environment and the stimulus, and finding a feature associated with the difference. Furthermore, generating the pain determination device includes using the feature to attach a label identifying the difference in the stimulus. In a specific embodiment, the pain determination device is generated through sigmoid function fitting or machine learning.

[0373] Step S10400: Determine whether there is a persistent characteristic and analyze and distinguish the pain

[0374] This is an arbitrary step, but it determines the presence or absence of a persistent characteristic and analyzes pain. The ability to determine pain by measuring and analyzing EEG data or its analysis data from the mid-term and subsequent time periods was unexpected. Furthermore, the discovery that the persistence characteristic observed in this range is associated with specific pain conditions was also unexpected.

[0375] Furthermore, the brainwave characteristic quantities used have complex characteristics, including amplitude, latency, duration of effect, distribution, frequency power, etc., including temporal, spatial, or both interactions. Therefore, the relationship between specific stimuli and / or environmental conditions and the characteristic quantities can be investigated by statistically comparing the characteristic quantities (t-test and / or analysis of variance (ANOVA)) or investigating continuous relationships (correlation and / or regression).

[0376] The obtained EEG data can be subjected to basic signal processing such as filtering, eye movement correction, and spurious signal removal as needed, and then the corresponding portion of the signal can be extracted in association with conditional parameters to generate EEG feature quantities. Examples include mean values ​​(arithmetic mean, geometric mean), other representative values ​​(median, mode), entropy, frequency power, wavelet, average, and components of potentials associated with single events.

[0377] When features such as persistence are discovered, they can be associated with pain, creating a means for determining the subject's pain. Here, as needed, the associated EEG feature values ​​can be used to set a threshold and / or determination index within the model curve obtained through fitting. The threshold can be set using numerical values ​​such as threshold potential, amount of positive potential, or level, and used as the determination index.

[0378] As needed, associated and specific feature quantities can be used to create a model for discriminating / estimating pain relative to existing or unknown stimuli and / or environments.

[0379] It is also possible to use brainwave feature quantities, in order to cooperate with conditional parameters to carry out the classification of pain into two or three categories or more and make a discrimination / estimation model, for example, one of the methods is to make a plot and fit (make it fit) to suitable fitting functions such as Sigmoid function patterns. Fitting can be carried out using any method known in the art. As such specific fitting function, step function, Boltzmann function, double Boltzmann function, Hill (Hill) function, logical dose response, Sigmoid Richards function, Sigmoid Weibull function etc. can be enumerated, but are not limited to these functions. Among these, the standard logistic function is called the Sigmoid function, and the standard function or deformation is general and preferred.

[0380] If the regression coefficient of the appropriate function pattern such as the Sigmoid function pattern is greater than a predetermined value as needed, a threshold value for determining whether or not discomfort is present can be set based on the Sigmoid curve or the like. Here, in the case of a Sigmoid curve, it can be generated based on an inflection point, but is not limited thereto. If needed, the pain classification value can also be corrected (calibrated) in such a way that the classification of the type of pain or the level of discomfort is maximized. The threshold value can be applied to the calculation or classification of the type and level of pain and can be used to determine the effectiveness of treatment.

[0381] Therefore, in a specific embodiment, the association includes setting a pain classification or difference based on a condition, such as a certain stimulus, and finding a feature quantity related to the classification or difference. Furthermore, generating the pain determination device includes using the feature quantity to attach a label identifying the difference in the stimulus. In a specific embodiment, the generation of the pain determination device can be achieved through sigmoid function fitting or machine learning.

[0382] In actual medical devices, one or more of steps S10100 to S10400 can be executed. However, the determiner or determination value can also be pre-set. In this case, S10300 uses an existing dataset of reference stimuli to set the benchmark feature value. This method calculates the z value of the feature value associated with the reference stimulus based on the activity level in the rest area where the reference stimulus is not presented, and stores this value as pre-stored data. When data for the target stimulus is newly recorded, the z value for the target stimulus is similarly calculated and compared with the existing reference stimulus value.

[0383] When the same subject becomes the subject, a step of using the previous pain assessment device (value, etc.) to inherit or update the assessor and / or assessment value may be included.

[0384] Alternatively, experimental electroencephalogram data or analysis data related to an unknown condition may be obtained from the subject and applied to the pain determination means to determine pain in the subject. In this case, a numerical value corresponding to the determinator and / or threshold value may be calculated based on actual measured values ​​related to the subject's unknown condition, such as electroencephalogram data or analysis data, and compared with the determinator and / or threshold value to determine the presence, type, or level of pain.

[0385] In the step of obtaining the brain wave data (e.g., amplitude data) of the object, brain wave data of an unknown state of the object is obtained from the object to be measured, regardless of whether it is subjected to some kind of stimulation or treatment. Any method can be used as long as it is a method that can obtain brain wave data. The same method as that used in the present invention to obtain brain wave data can be used, and the same method is usually used. In addition, a pain determiner and / or judgment value is applied to determine the pain of the object. The prescribed pain determination device or value is associated with the level of discrimination / estimation for the object, and is called a "pain determiner" or a "no pain determination predictor." Regarding the numerical value on the side where the pain is stronger than the threshold, it is determined or predicted that there is pain (or a specific type of pain), and regarding the numerical value on the side where the pain is weaker than the threshold, it is determined or predicted that there is no pain (or a specific type of pain).

[0386] In one embodiment, the data recording locations for the EEG data or its analysis data may include locations on the scalp from the frontal lobe to the parietal lobe and then to the occipital region, such as F3, F4, C3, C4, P3, and P4, based on the international 10-20 standard or its extended standard, as electrode locations. Alternatively, locations separated by a specific uniform distance (e.g., 2.5 cm) may be included, and the EEG data or its analysis data may include at least one EEG feature selected from a combination of these.

[0387] In another embodiment, the brain wave feature includes at least one selected from Fp1, Fp2, Fpz, F3, F4, Fz, C3, C4, Cz, P3, P4, and Pz, for example, including average amplitudes Fz, C3, and C4, and frequencies Fz(δ), Fz(β), Cz(δ), C3(θ), and C4(β). Although preferably including Cz(amplitude), C3(α), Cz(β), Fz(δ), and Cz(γ), it is not limited to this.

[0388] Figure 28 A schematic diagram of the device of the present invention is described in . In the case where this embodiment is to generate a pain determiner (device), 11000 to 13000 are involved. The stimulation presentation unit 11000 corresponds to A), in which information related to the environment and / or stimulation type in which the stimulation is presented is transmitted to the electroencephalogram data acquisition unit 12000 and the pain judgment value generation unit 13000. The electroencephalogram data acquisition unit 12000 is configured to have an electroencephalogram or be connected to the electroencephalogram (12500), which is connected to or can be connected to the object (11500) in a manner that can obtain electroencephalogram data synchronized with the stimulation emitted from the stimulation presentation unit to the object (11500).

[0389] Figure 29This is a block diagram illustrating the functional configuration of a pain assessment system 15100 according to one embodiment (note that several components in this configuration diagram are arbitrary and may be omitted). System 15100 includes an electroencephalogram measurement unit 15200, which internally includes or externally connects an electroencephalogram recording sensor 15250 and, if necessary, includes or externally connects an electroencephalogram amplifier 15270. Pain assessment device 15300 performs signal processing and identification / estimation of pain. In pain assessment device 15300, electroencephalogram signal processing unit 15400 processes electroencephalogram signals (and, if necessary, electroencephalogram feature extraction unit 15500 extracts electroencephalogram feature quantities). Pain assessment unit 15600 identifies / estimates pain or the level or type of pain, and (if necessary) visualization of discomfort is performed by identification level visualization unit 15800. In addition, a stimulation device unit 15900 is provided inside or outside the device 15300. The stimulation device unit 15900 transmits stimulation information (stimulation type, environmental information, etc.) for the purpose of generating the subject's pain and pain discriminator and discriminating the level of the actual unknown pain. In addition to the stimulation presentation unit 15920, the stimulation device unit 15900 may also include a stimulation information visualization unit 15960 as needed to display images, numbers, and other information related to the stimulation and / or environment. In addition, the pain determination system may also include a generation unit 15700 for generating a discriminator and / or a determination value outside or inside the device 15300.

[0390] Thus, the pain assessment system 15100 includes an electroencephalogram measurement unit 15200 and a pain assessment device 15300, and, if necessary, a stimulation device unit 15900. The pain assessment device 15300 is implemented, for example, by a computer having a processor and memory. In this case, when the processor executes a program stored in the memory, the pain assessment device 15300 causes the processor to function as an electroencephalogram amplification unit 15270, an electroencephalogram signal processing unit 15400, a pain assessment unit 15600 (if necessary), and a discrimination level visualization unit 15800 (if necessary). Stimulus and / or environmental information can also be visualized as needed. Furthermore, the system 15100 or pain assessment device 15300 of the present invention can be implemented, for example, by a dedicated electronic circuit. The dedicated electronic circuit can be a single integrated circuit or multiple electronic circuits. The electroencephalogram data acquisition unit and the pain assessment value generation unit can also employ the same configuration as the pain estimation device.

[0391] The electroencephalogram measurement unit 15200 obtains a plurality of electroencephalogram data from an estimation target by performing multiple electroencephalogram measurements via an electroencephalogram meter (electroencephalogram recording sensor 15250). The estimation target is a living body whose electroencephalogram changes due to stimulation and / or environment, and is not limited to humans.

[0392] The pain determination unit 15600 uses the determination value to discriminate / estimate pain. Even if the discriminator and / or determination value have not been pre-generated externally or internally, they are generated. The unit generating the discriminator and / or determination value can be provided externally or internally to the apparatus 15300 as the pain determination value generation unit 15700. The pain determination value is used to estimate or classify pain based on the amplitude of multiple sets of electroencephalogram data. In other words, the pain determination unit 15600 or the pain determination value generation unit 15700 can generate a determination value for estimating or classifying the subject's pain based on the electroencephalogram data.

[0393] Electroencephalogram recording sensor 15250 uses electrodes on the scalp to measure electrical activity within the subject's brain. Electroencephalogram recording sensor 15250 then outputs electroencephalogram data as the measurement result. The electroencephalogram data can be amplified as needed.

[0394] based on Figure 28 Further explanation is given. The scheme including the judgment unit is explained. Figure 28 In addition to the pain determination unit 14000, the electroencephalogram data acquisition unit 12000 is also referenced. The dotted line shows the step of creating a discrimination model, and the solid line shows the step of performing discrimination / estimation of the actual pain level. In this case, as described in the section (Generating Pain Judgment Values), electroencephalogram data can be obtained from the subject 11500 via an electroencephalogram. That is, the electroencephalogram data acquisition unit 12000 is configured to be connected to the subject 11500, and the electroencephalogram data acquisition unit 12000 is configured to have an electroencephalogram or to be connected to the electroencephalogram (12500), which is connected to or can be connected to the subject (11500) in a manner that obtains electroencephalogram data obtained from the subject (11500). The pain determination unit 14000 pre-stores the pain judgment value, or is configured to receive separately generated data and to be able to refer to it as needed. Such a connection structure can be either wired or wireless. The pain assessment value stored in advance is generated in the pain assessment value generating unit 13000 by, for example, a discriminator (Sigmoid function fitting, etc.) based on feature amounts.

[0395] Figure 29This is a block diagram illustrating the functional configuration of a pain assessment system 15100 according to one embodiment. System 15100 includes an electroencephalogram measurement unit 15200, which internally or externally includes an electroencephalogram recording sensor 15250 and, if necessary, an electroencephalogram amplifier 15270. Pain assessment device 15300 processes and discriminates / estimates pain signals. In pain assessment device 15300, electroencephalogram signal processing unit 15400 processes electroencephalogram signals, pain assessment unit 15600 estimates / discriminates pain (if necessary), and pain is visualized (if necessary) by assessment level visualization unit 15800. Furthermore, a stimulation device 15900 is internally or externally included to facilitate the creation of a pain discriminator for the subject. A determination value may also be pre-created by pain determination value generation unit 15700.

[0396] As shown above, the pain determination system 15100 has an electroencephalogram measuring unit 15200 and a pain determination device 15300. The pain determination device 15300 is implemented, for example, by a computer having a processor and a memory. In this case, when the program stored in the memory is executed by the processor, the pain determination device 15300 causes the processor to function as an electroencephalogram amplifying unit 15270, an electroencephalogram signal processing unit 15400, (as needed) a pain determination unit 15600, (as needed) a discrimination level visualization unit 15800, etc. as needed. The processor can also generate and visualize reference stimuli as needed. In addition, the system 15100 or device 15300 of the present invention can also be implemented, for example, by a dedicated electronic circuit. The dedicated electronic circuit can be either an integrated circuit or a plurality of electronic circuits. The electroencephalogram data acquisition unit and the pain classification value generation unit 13000 (refer to Figure 28 ) can adopt the same structure as the pain estimation device or can be constructed externally.

[0397] The electroencephalogram measurement unit 15200 obtains a plurality of electroencephalogram data from an estimation subject by performing multiple electroencephalogram measurements via an electroencephalogram meter (electroencephalogram recording sensor 15250). The estimation subject is a living body whose electroencephalogram changes due to pain, and is not limited to humans.

[0398] The pain determination unit 15600 generates the pain determination value based on the pain determination value generated by the pain determination value generation unit 13000 (see Figure 28 ) creates a pain classification value and estimates or classifies the magnitude of the pain based on the amplitude of the plurality of electroencephalogram data. That is, the pain determination unit 15600 estimates or classifies the subject's pain based on the electroencephalogram data based on the determination value.

[0399] Electroencephalogram recording sensor 15250 uses electrodes on the scalp to measure electrical activity within the subject's brain. Electroencephalogram recording sensor 15250 then outputs electroencephalogram data as the measurement result. The electroencephalogram data can be amplified as needed.

[0400] Next, the processing or method of the device configured as above will be described. Figure 27 Flowchart 10100 to 1040 are steps in this process.

[0401] The pain judgment value can be created and stored in the pain judgment unit 14000 (see Figure 28 ), the pain assessment unit 14000 may also be configured to receive value data. Alternatively, if a pain assessment value generating unit 13000 is provided, the pain assessment value may be stored in the generating unit, or a separate recording medium may be provided. Alternatively, the value may be received via communication.

[0402] Next, brain wave data is obtained from the subject (S10200) (see Figure 27 ). The brain wave data can be obtained using the same technology and the same implementation method as the reference data, but it is not necessary to always use the same device or component, and can be different or the same.

[0403] Next, the brain wave data (eg, amplitude data) obtained in S10200 is fitted to the pain judgment value, and pain discrimination / estimation corresponding to the brain wave data is performed (S10300) (see Figure 27 Such pain assessment may also be configured to display specific text (comfortable pain, uncomfortable pain, etc.) or make a sound when a predetermined value is output in advance, or to display the actual value and the pain assessment value side by side so that the user (clinician) can discuss.

[0404] Figure 30 is Figure 29The block diagram of the pain assessment system 15100 is expanded to include the process of generating a discriminator (e.g., determining a discriminant value by fitting a Sigmoid function) and includes the work content. System 15100 includes an electroencephalogram measurement unit 15200 connected to an electroencephalogram 15220. Electroencephalogram features such as average values ​​can be obtained from collected electroencephalogram data as needed by a feature extraction unit 15500. When a pain assessment value is generated in advance, a pain assessment value generation unit 13000 located outside or inside the pain assessment device 15300 generates a pain assessment value by fitting a discriminator, such as a Sigmoid function, and / or generating parameters for determining persistence characteristics. The pain assessment value is then sent to the pain assessment unit 15600 and stored. In actual pain determination of an unknown stimulus type or environment, after the brainwave data synchronized with the stimulus presentation or display of the stimulus presentation unit 15920 is sent from the electroencephalogram 15220 to the measurement unit 15200, it is converted into brainwave features in the feature extraction unit 15500 and sent to the pain determination unit 15600, and the pain determination value is used to discriminate / estimate the pain of the unknown stimulus and / or environment. (As needed) The pain is visualized by the discrimination level visualization unit 15800. Such a series of processes can be implemented by a computer and / or portable terminal having a processor and a memory, or by a dedicated electronic circuit. The dedicated electronic circuit can be either an integrated circuit or a plurality of electronic circuits. In addition, it can be implemented by software or by controlling the necessary hardware.

[0405] (Other embodiments)

[0406] While the pain estimation device according to one or more embodiments of the present invention has been described above based on an embodiment, the present invention is not limited to such embodiment. Various modifications to the embodiment that would be conceivable to a person skilled in the art, and / or configurations combining components from different embodiments, may also be included within the scope of one or more embodiments of the present invention, provided that they do not depart from the spirit of the present invention.

[0407] For example, in each of the above embodiments, the peak-to-peak value may be used as the amplitude value of the electroencephalogram data, but the present invention is not limited thereto. For example, a single peak value may be used as the amplitude value.

[0408] Furthermore, according to the above embodiment, it is assumed that the pain severity value Pmax corresponding to the upper limit Amax of the EEG amplitude is 1, and the pain severity value Pmin corresponding to the lower limit Amin of the EEG amplitude is 0, and the visualization unit 15800 displays the assessment level. However, this is not limiting. For example, the pain severity can also be expressed on a scale of 0 to 100. In this case, the pain assessment unit 5600 can estimate the pain severity value Px using the following equation.

[0409] Px=Pmax×(Ax-Amin) / (Amax-Amin)

[0410] In addition, in the above description, curve fitting is described as an example of generating a judgment value for uncomfortable pain by analyzing multiple brain wave data, but the present invention is not limited to this. In addition, a predetermined value can also be used as the upper limit of the brain wave amplitude. In order to remove external and biological false signals in the process of calculating the brain wave feature value related to the event related to the instantaneous pain, the predetermined exclusion value is, for example, 50μV to 100μV, which can be set through experiments or experience. In this way, in conventional analysis, as a false signal removal method, data from approximately plus or minus 50μV to 100μV is excluded, but such false signal removal can also be implemented in the present invention as needed.

[0411] In addition, regarding the stimulus presentation unit 15920 (see Figure 30 ) is applied to subject 15099. Any type of stimulus may be applied as long as the magnitude of pain felt by subject 15099 varies depending on the stimulus type and / or presentation environment. However, for the extraction of EEG feature quantities of instantaneous pain in the present invention, the following examination paradigm (pain oddball paradigm) is used, namely, a reference stimulus serving as a benchmark or background stimulus is presented with a strong pain stimulus or an uncomfortable pain stimulus in a ratio of, for example, 7 to 3.

[0412] In addition, part or all of the components of the pain determination device in each of the above embodiments may be constituted by a system LSI (Large Scale Integration). Figure 30 As shown, the pain determination device 15300 may be configured by a system LSI including the measurement unit 15200 and the stimulation presentation unit 15920 as needed.

[0413] A system LSI is a highly versatile LSI manufactured by integrating multiple components onto a single chip. Specifically, it is a computer system consisting of a microprocessor, ROM (Read Only Memory), and RAM (Random Access Memory). The ROM stores a computer program. The microprocessor operates according to the program, enabling the system LSI to achieve its functions.

[0414] It should be noted that although the term "system LSI" is used here, it is sometimes referred to as IC, LSI, super LSI, or large-scale LSI based on differences in integration. In addition, the method of integrated circuitization is not limited to LSI, and can also be achieved through dedicated circuits or general-purpose processors. FPGAs (Field Programmable Gate Arrays) that can be programmed after LSI manufacturing, or reconfigurable processors that can reconfigure the connections and / or settings of circuit units within the LSI can also be used.

[0415] Furthermore, if advances in semiconductor technology or other derivative technologies lead to the emergence of integrated circuit technology that can replace LSIs, then this technology can naturally be used to integrate functional blocks. Applications to biotechnology are also possible.

[0416] Furthermore, one embodiment of the present invention may not only be such a pain assessment value generation and pain assessment device, but also a pain classification value generation and pain classification method that incorporates the characteristic components included in the pain estimation device as steps. Furthermore, one embodiment of the present invention may be a computer program that causes a computer to execute the characteristic steps included in the pain assessment value generation and pain assessment method. Furthermore, one embodiment of the present invention may be a computer-readable nonvolatile recording medium that stores such a computer program.

[0417] Furthermore, in each of the above-described embodiments, each component may be implemented using dedicated hardware or by executing a software program suitable for each component. Alternatively, each component may be implemented by a program execution unit, such as a CPU or processor, reading and executing a software program stored on a recording medium, such as a hard disk or semiconductor memory. The software implementing the pain estimation device and the like in each of the above-described embodiments may be referred to as the aforementioned program in this specification.

[0418] Therefore, the present invention provides a program that enables a computer to execute a method for distinguishing or evaluating pain, the method comprising the following steps: comparing all or part of the brain wave data or its analysis data during the period of 2000 milliseconds from the earliest time point among the induced brain wave component, the initial event-related potential component and 250 milliseconds after the application of the object stimulus, with the brain wave data or its analysis data after the same time period after the application of the reference stimulus.

[0419] In another embodiment, the present invention provides a recording medium storing a program for causing a computer to execute a method for distinguishing or evaluating pain, the method comprising the steps of comparing all or part of the brain wave data or its analysis data during the period of 2000 milliseconds from the earliest time point among the induced brain wave component, the initial event-related potential component and 250 milliseconds after the application of the object stimulus with the brain wave data or its analysis data after the same period of time after the application of the reference stimulus.

[0420] In another embodiment, the present invention provides a pain identification or assessment system comprising an electroencephalogram (EGW) data input unit and an analysis unit. The EGW data input unit inputs EGW data or its analysis data, and the analysis unit compares all or part of the EGW data or its analysis data from the earliest time point of 2000 milliseconds after application of a target stimulus, including an evoked EGW component, an initial event-related potential component, and 250 milliseconds, with EGW data or its analysis data from the same time period after application of a reference stimulus. This system can function as a medical device or as a portion thereof.

[0421] (Note)

[0422] In this specification, "or" is used when "at least one or more" of the items listed in the text can be used. The same applies to "or". In this specification, when it is clearly stated that "within the range between two values", the range also includes the two values ​​themselves.

[0423] References such as scientific literature, patents, and patent applications cited in this specification are incorporated herein by reference in their entirety to the same extent as if each were specifically described.

[0424] The present invention has been described above by way of preferred embodiments for ease of understanding. The present invention will be described below based on examples. However, the above description and the following examples are provided for illustrative purposes only and are not intended to limit the present invention. Therefore, the scope of the present invention is not limited to the specific embodiments described in this specification or the examples, but is defined solely by the claims.

[0425] Example

[0426] The following examples describe the treatment of the subjects used in the following examples, where necessary, in accordance with the standards established by Osaka University, and in accordance with the Declaration of Helsinki and ICH-GCP in cases related to clinical research.

[0427] (Example 1: Sparse Model Analysis under Thermal Pain Stimulation)

[0428] In this example, sparse model analysis of thermal pain stimulation experimental data was performed.

[0429] (Participant)

[0430] Forty healthy adults ranging in age from their 20s to their 70s participated in this study. Participants signed informed consent forms before the experiment. All participants self-reported having no history of neurological and / or psychiatric disorders or acute and / or chronic pain treated with clinical medications. This study was conducted under the approval of the Ethics Committee of Osaka University Hospital and the Declaration of Helsinki.

[0431] (step)

[0432] The inventors of the present invention used a thermal stimulus presentation paradigm. The paradigm is: using thermal stimulation, the baseline temperature is 35°C, and the temperature is increased by 2°C per level from 40°C at level 1 to 50°C at level 6. The trial block of each stimulation level includes 3 stimulations, each stimulation has a standby time of 5 seconds for rise and fall, and a plateau period of 5 seconds. There is an inter-stimulus interval of 5 seconds between each stimulation. The rest period between blocks is fixed at 100 seconds. Participants wear a thermal stimulation probe on the inner side of their left forearm and lie in an easy chair to receive thermal stimulation. In addition, participants continuously evaluate the pain intensity on a scale from 0 to 100 (0 = "no pain"; 100 = "unbearable pain") based on a computerized visual analog scale (COVAS). COVAS data are recorded while the stimulation intensity changes.

[0433] (EEG data recording)

[0434] The interval between rest blocks was fixed at 100 seconds. In the two-way paradigm, EEG was recorded from five scalp Ag / AgCl scalp electrodes (Fp1, Fz, Cz, C3, C4) using a commercially available bioamplifier (EEG 1200; Nihon Koden). Fp1, the most anterior electrode, was used to record EOG activity. Reference electrodes were worn on the earlobes of both ears, and the outer electrodes were placed on the center of the forehead. The sampling rate was 1,000 Hz, and amplification was performed using a bandpass filter in the range of 0.3 to 120 Hz. The impedance of all electrodes was less than 15 kΩ.

[0435] (EEG Analysis)

[0436] The four electrodes on the scalp (Fz, Cz, C3, and C4) were selected for EEG analysis. The frontal electrode Fp1 was used for eye movement processing. Before conditional clipping of the EEG data, the following regression filter was used to remove EOG from the analysis electrodes. Since the Fp1 data is closest to the left eye and is significantly affected by eye movements, it was used as the EOG data.

[0437] Original EEG = β × EOG + C

[0438] Estimated EEG = original EEG - β × EOG

[0439] β: partial regression coefficient

[0440] C: intercept

[0441] Estimated EEG: Estimated EEG

[0442] After the attenuation of VEOG, a 60Hz notch filter was applied to reduce external electrical noise. In addition, time periods in which potentials exceeding 100μV were mixed were removed from the data. Afterwards, the brain wave data from 5 seconds before each stimulus presentation to 15 seconds after the stimulus presentation (18 time periods (18 epochs)) were intercepted, and baseline correction was performed using the potential before the stimulus presentation. After baseline correction, the amplitude was made absolute, and the potential was standardized using the maximum value of all conditions and all stimuli for each electrode. Finally, the average amplitude during the 15 seconds after the stimulus presentation was calculated, and the average potentials of levels 1 and 2 were extracted as the characteristic quantity of weak pain levels, and the average potentials of levels 5 and 6 were extracted as the characteristic quantity of strong pain levels.

[0443] (Frequency Analysis)

[0444] Four electrodes on the scalp (Fz, Cz, C3, and C4) were used for EEG analysis. Initially, a 60Hz notch filter was applied to attenuate external electrical noise. EEG data was captured for 15 seconds after stimulus presentation (18 epochs) and Fourier transformed to calculate the frequency power (log value of the real part) of δ (1-3Hz), θ (4-7Hz), α (8-13Hz), β (14-30Hz), and γ (31-100Hz).

[0445] [Formula 6]

[0446] Frequency Power = log(abs(FFT))

[0447] log:log10

[0448] abs: absolute value

[0449] FFT: Fourier transform data of each frequency band

[0450] Normalization was performed based on the maximum value of the frequency power across all electrode usage conditions and all stimulations. Frequency powers of levels 1 and 2 were extracted as features indicating weak pain levels, while frequency powers of levels 5 and 6 were extracted as features indicating strong pain levels. Through these EEG and frequency analysis, a total of 24 features for modeling were obtained using Fz, Cz, C3, and C4.

[0451] (Analysis of a sparse model for distinguishing the intensity of thermal pain stimulation)

[0452] The inventors of the present invention used 24 feature quantities (4 time domain data, 20 frequency domain data) and two levels of thermal pain stimulation labels (1 = "weak"; 2 = "strong"; n = 160) and performed sparse model analysis (multiple regression analysis) using the LASSO algorithm. Figure 4 As shown, the model-making data and test data were divided into an 8:2 ratio, and a ten-fold cross-validation was performed using the model data to determine the characteristic quantity coefficient (partial regression coefficient) and intercept of the regression equation and calculate the optimal λ value. After determining the λ value, partial regression coefficient, and intercept, the estimated value of the pain level was calculated using the characteristic quantity of the test data. Regarding the estimated value, the estimated value below 50% of the total was set to "1", that is, it was set to "weak pain", and the estimated value greater than 50% of the total was set to "2" (strong pain). The discrimination accuracy was compared with the actual pain label to obtain the discrimination accuracy. The process of randomly selecting learning data and test data was performed 1000 times, and the characteristic quantity coefficient, the average value of the discrimination accuracy, and the distribution were calculated. In addition, the random level of discrimination accuracy was calculated by randomly changing the label of the test data, and it was compared with the actual discrimination accuracy.

[0453] (result)

[0454] Figure 5 It is the partial regression coefficient of the feature quantity used in the regression formula, the change in 1000 discrimination accuracy verifications, the average value of the coefficient, and the standard deviation. The feature quantities with high contribution are the frequency feature quantities such as C4's α power (-2.38), Fz's δ power (-2.17), C3's β power (2.10), Fz's β power (-2.03), Cz's δ power (1.88), C4's θ power (1.86), Cz's α power (-1.86), and Fz's θ power (1.57). Figure 6 As shown, the distribution of the 1,000 discrimination accuracy results showed that the percentage exceeded 80% in approximately 350 cases, exceeding 30% of the total. The average discrimination accuracy was 78.2±7.6%, which is approximately 30% higher than the chance level discrimination accuracy (50±8.7%) when the pain level of the test data was randomized.

[0455] (Example 2: Utilization in a Pre-generated Regression Model (Discriminant Model))

[0456] In this embodiment, an actual example of a pain level determination device is shown.

[0457] In practice, we hope Figure 16 As shown, the discriminant model generation unit is connected to the pain discrimination / estimation device or pre-connected in an accessible manner. Such a pain discrimination / estimation device can be provided using the discriminant model based on sparse modeling obtained in the present invention.

[0458] According to Example 1, a method for discriminating / estimating the pain level of an unknown estimation subject is demonstrated, using a pre-created pain discrimination / estimation model. This pre-created discrimination model is stored in the pain discrimination / estimation unit or is pre-accessible. The subjects and data analysis method are the same as in Example 1, but the model creation data is generated by randomly excluding one subject (four samples). This excluded subject 1 is then substituted into the discrimination / estimation model as the unknown estimation subject.

[0459] (result)

[0460] like Figure 7 As shown, through the sparse model analysis used for discriminant model creation, the characteristic quantities of C4's δ power (2.50), C3's β power (2.43), C4's α power (-2.39), Fz's β power (-2.38), Cz's δ power (2.22), Fz's δ power (-2.16), Fz's α power (-2.13), and Fz's θ power (1.51) have a high contribution to the model. Figure 8 As shown, the intercept value during model creation ranged from 0.03 to 0.05, and the median and mean were both approximately 0.043. Figure 9 The following table shows the change of the discrimination accuracy of the ten-fold cross validation. The accuracy (maximum value "1") converges to the range of 0.6 to 0.9, and the median and average values ​​are both 0.733. The optimal λ value is 0.0027. Based on the above results, the average value of the feature coefficient and the intercept is used to create Figure 10 Such a pain level discrimination / estimation model uses a 50% threshold level for the estimated value, classifying cases below 50% as "weak pain level" and cases above 50% as "strong pain level", and then compares the estimated value with the actual pain level. Figure 11 The pain level discrimination / estimation formula and output result of the unknown estimation subject who was excluded are shown. The case of this subject shows that the pain level can be accurately discriminated / estimated based on brain feature values ​​using the discrimination / estimation model.

[0461] (Example 3: Determination of instant pain)

[0462] In this example, instantaneous pain discrimination was performed using the pain oddball task. High temperature stimulation was used as the instantaneous pain stimulus. Furthermore, after the electroencephalogram test, the discomfort of the pain stimulus was investigated in subjective reports.

[0463] (method)

[0464] (Participant)

[0465] A similar group of 80 healthy adults, ranging in age from their 20s to their 70s, participated in a pain oddball paradigm experiment using high-temperature stimulation. Participants signed informed consent forms before the experiment. All participants self-reported having no history of neurological and / or psychiatric disorders or acute and / or chronic pain under clinical medication. This example was conducted with the approval of the Ethics Committee of Osaka University Hospital and the Declaration of Helsinki.

[0466] (Experimental stimulation and procedures)

[0467] Figure 20 An overview of the experimental method is shown in the figure. Using a temperature stimulation system (Pathway: Medoc Co., Ltd., Ramat Yishai, Israel), high temperature stimulation was applied to the right forearm of the participant. The basal temperature was set to 32°C, and reference stimulation (39°C: 75 trials) and highly painful deviation stimulation (52°C: 21 trials) were randomly presented. For the reference stimulation and deviation stimulation, the rise was generated in a pulsed manner at 2 seconds, and the interval between stimulations was 1 second. The subject silently counted the number of painful stimulations during the examination and reported it after the end. In addition, after the end of the EEG experiment, the reference stimulation and deviation stimulation were randomly generated multiple times, and the discomfort was evaluated based on the computerized visual analog scale (COVAS) on a range from 0 to 100 (0 = "no pain"; 100 = "unbearable pain"). COVAS data were recorded while the stimulation intensity was changed.

[0468] (EEG data collection)

[0469] EEG data were recorded using a commercially available electroencephalogram (EEG) device using Ag / AgCl7 electrodes (Fp1, Fp2, F3, F4, C3, C4, and Pz) on the scalp. The lead electrode was the ear, with each left and right electrode connected to the same ear electrode. The frequency range was 0.3–120 Hz, with a sampling frequency of 1000 Hz. Impedance was maintained below 15 kΩ.

[0470] (EEG Analysis)

[0471] (Amplitude feature extraction)

[0472] To attenuate eye movement noise (EOG), the continuously recorded EEG data were applied to the following regression filter:

[0473] [Mathematical formula 1]

[0474] Original EEG = β × EOG + C

[0475] Estimated EEG = original EEG - β × EOG

[0476] β: regression coefficient

[0477] C: intercept

[0478] Estimated EEG: Estimated EEG

[0479] As eye movement data, the Fp1 and Fp2 data closest to both eyes were used as EOG data (Fp1 + Fp2). After EOG correction, bandpass filtering (0.3-40 Hz) was applied to attenuate low-frequency and high-frequency components. For each stimulation condition, the waveform (epoch waveform) was captured from 200 milliseconds before to 2000 milliseconds after stimulus presentation. After baseline correction using the average potential before stimulus presentation, artifacts were removed at ±50 μV and averaged.

[0480] (Subjective Evaluation Analysis)

[0481] Using a subjective evaluation paradigm, subjective scores for discomfort of reference and intensely painful stimuli were calculated. Each conditional stimulus was randomly presented three times, and the maximum value before the next stimulus was calculated starting from 1 second before the onset of the next stimulus, and the average was calculated. Generally, subjective pain ratings rise or fall with a delay after stimulus presentation, so the above-mentioned timing settings were used to reliably obtain stimulus ratings.

[0482] (Statistical Analysis)

[0483] (Continuity Characteristics Analysis)

[0484] Based on the arithmetic mean waveforms of all 80 subjects, the average amplitude from 600 milliseconds to 2000 milliseconds, which is the end of the mid-ERP period and the beginning of the late period, was used to compare the reference stimulus with the highly painful deviation stimulus using a t-test. In addition, the average amplitude from 200 to 600 milliseconds was also compared between the two stimuli to assess their effectiveness using a t-test.

[0485] (Results and Investigation)

[0486] Figure 21An example of a persistent signal (characteristic) discovered in the present invention is shown. It is clear at a glance that the waveform of the deviation stimulus (52°C) continues to shift in the positive direction compared to the reference stimulus (39°C) until the end of the 2000 millisecond time period (epoch). Observing the waveform of F4 in the right frontal lobe, its persistent effect starts as early as around 400 milliseconds, and the effect lasts for a very long time. A paired t-test was performed using the average potential of the two electrodes in the frontal lobe (600-2000ms), and the results confirmed a significant difference between the two conditions (t=2.523, p=0.014). On the other hand, in the mid-term time period before 600 milliseconds (200-600 milliseconds), no significant difference was seen between the conditions (t=0.331, p=0.742). Based on the above results, it can be understood that the deviation stimulus exhibits a persistent effect starting from the mid-to-late time period.

[0487] Figure 22 The following table shows subjective evaluations of pain discomfort for the reference stimulus (39°C) and the deviation stimulus (52°C). The discomfort for the 52°C stimulus was significantly higher than for the 39°C stimulus (t=14.38, p<0.0001). Therefore, it can be understood that the persistent positive effect in the EEG analysis above reflects the increased discomfort associated with strong pain stimulation.

[0488] (Example 4: Discrimination and estimation process of two levels of instantaneous pain using continuous ERP feature quantities)

[0489] Using the continuous ERP feature quantity observed in Example 1 and the LASSO (regularization) algorithm, instantaneous pain discrimination estimation was performed.

[0490] (Materials and Methods)

[0491] The experimental paradigm, EEG data collection, and analysis methods followed those of Example 1. In this example, the LASSO (Least Absolute Shrinkage and Selection Operator) algorithm was newly used for sparse model analysis to calculate the optimal coefficients and model intercept of the multiple regression model and to perform discriminant estimation of two levels of instantaneous pain.

[0492] like Figure 24As shown, the collected data (160 samples = 2 levels × 80 people) are divided into "training (learning) data" and "testing data". After the optimal λ value is calculated using the LASSO algorithm and the ten-fold cross-validation using the training data, the partial regression coefficients and intercepts of the four feature quantities are determined. The four feature quantities use the average potential of the reference stimulus and the strong pain stimulus from 1000 milliseconds to 1600 milliseconds when the sustained potential in the F3, F4, C3, and C4 electrodes can be significantly observed (normalized within the individual). The λ value is a hyperparameter that plays a role in regularization, and has the function of smoothing the fitness of the model and increasing the generalization ability. L1 regularization is used in regularization, and the LASSO function of MATLAB solves the following minimization problem.

[0493] [Mathematical formula 1]

[0494] Min(Dev(β0,β)+λΣ|β j |)

[0495] Min: Minimize

[0496] Deviation: Deviation (the deviation of the estimated value of the regression model using the intercept β0 and the regression coefficient β from the observed value)

[0497] N: number of samples

[0498] λ: Regularization parameter with positive value

[0499] In this example, the above process was repeated 1000 times to obtain the posterior distribution of discrimination accuracy. Furthermore, to demonstrate that the discrimination accuracy of observed data is higher than that of random chance, the same process was used to randomize the two pain levels, and the discrimination accuracy of the chance level was also calculated and compared.

[0500] (result)

[0501] Figure 25 The results of discriminant estimation of two levels of instantaneous pain using continuous ERP feature values ​​are shown in . The discriminant model using the coefficients and intercepts of the four feature values ​​obtained by 1000 LASSO analyses can be summarized as follows.

[0502] [Mathematical formula 2]

[0503] Y=0.1659×F3+0.171×F4-0.0079×C3-0.0841×C4+0.6671

[0504] As an example of the output results when the test data is substituted into a specific model of 1000 discriminant analysis, the instantaneous pain classification value is shown. The classification value "1.4424" is set as the threshold. When it is less than the threshold, it is considered as weak pain or no pain (reference stimulus). When it is above the threshold, it is considered as strong pain. The discrimination accuracy is calculated by comparing it with the actual pain level. The discrimination results of 1000 test data are as follows: Figure 25 The right graph shows an average discrimination accuracy of 66.7%. This accuracy is approximately 16% higher than the 50.8% accuracy obtained when the discrimination labels were randomized, indicating that persistent ERP features are effective for discriminating instantaneous pain.

[0505] (Example 5: Comparative Example Using a Discrimination Model of Non-persistent ERP Feature Quantities)

[0506] In order to further study the effectiveness of the continuous feature quantity in instantaneous pain discrimination in Example 4, the non-continuous ERP feature quantity and LASSO (regularization) algorithm in Example 3 were used to perform instantaneous pain discrimination estimation.

[0507] (Materials and Methods)

[0508] The experimental paradigm, EEG data collection, EEG analysis method, and discriminant estimation analysis followed those of Example 4. In this example, non-sustained electroencephalogram intervals from 200 milliseconds to 600 milliseconds were used as feature quantities used for discriminant estimation.

[0509] (result)

[0510] Figure 26 The discriminant estimation results for two levels of instantaneous pain using non-sustained ERP feature values ​​are shown in . The discriminant model using the coefficients and intercepts of the four feature values ​​obtained by LASSO analysis (verified with 1000 test data) can be summarized as follows.

[0511] [Mathematical formula 3]

[0512] Y=-0.0704×F3+0.1479×F4+0.2220×C3-0.2485×C4+1.4917

[0513] As an example of the output result when specific test data is substituted into a specific discriminant model, an example of instantaneous pain classification value is shown. The classification value is "1.4670", which is about "0.02" different from the example of the continuous feature value, which is similar. When it is less than the threshold, it is set as weak pain or no pain (reference stimulus), and when it is above the threshold, it is set as strong pain. The discrimination accuracy is calculated by comparing it with the actual pain level. The results of 1000 discriminant analyses are as follows: Figure 26 The right graph shows the average discrimination accuracy of 57.3%. This accuracy is higher than the chance level (50%) and approximately 8% higher than the 49.2% accuracy obtained when randomizing the discrimination labels, but approximately 10% lower than the 66.7% accuracy obtained when using persistent ERP features.

[0514] The above results demonstrate that, in this example, persistent ERP features are also effective for discriminating between the two levels of transient pain. Furthermore, even with transient pain, it is the features associated with conscious pain that emerge in the middle and later stages, rather than the brain activity in the immediate early stages, that are important. This suggests that capturing the point at which physical aggression becomes psychological aggression is effective for pain discrimination.

[0515] (Note)

[0516] As described above, the present invention is illustrated by the preferred embodiments of the present invention, but it should be understood that the scope of the present invention should be interpreted only in accordance with the claims. Regarding the patents, patent applications and documents cited in this specification, it should be understood that their contents should be cited as references to this specification, just as their contents are specifically recorded in this specification. This application claims priority to Special Application No. 2017-137723 (filed on July 14, 2017) and Special Application No. 2017-193501 (filed on October 3, 2017) filed with the Japan Patent Office, and the contents of these applications are all cited as references in this specification.

[0517] Industrial Application Possibilities

[0518] The present invention can accurately classify pain, estimate pain even without inflicting severe pain, and perform more precise diagnosis and treatment of pain.

[0519] Furthermore, the present invention can determine instantaneous pain and can perform more precise diagnosis and treatment of instantaneous pain.

[0520] Description of Reference Numerals

[0521] 1000: Reference stimulus

[0522] 1500: Object

[0523] 2000: Brainwave Data Acquisition Department

[0524] 2500: Encephalograph

[0525] 2600: Brainwave feature extraction unit

[0526] 3000: Discriminant model generation unit

[0527] 3200: Pain level determination / estimation (classification)

[0528] 4000: Pain Visualization Department

[0529] 5100: Pain Level Discrimination and Estimation System

[0530] 5200: Encephalogram Measurement Department

[0531] 5250: Brainwave Recording Sensor

[0532] 5300: Pain identification / estimation device

[0533] 5400: Brainwave Signal Processing Unit

[0534] 5500: Brainwave feature extraction unit

[0535] 5600: Pain identification / estimation

[0536] 5700: Pain discrimination and calibration department

[0537] 5800: Pain Level Visualization Department

[0538] 5900: Stimulation Device Department

[0539] 5920: Reference stimulus presentation terminal

[0540] 5940: Reference stimulus generation unit

[0541] 5960: Reference stimulus level visualization unit

[0542] 11000: Stimulus presentation

[0543] 11500: Object

[0544] 12000: Brainwave Data Acquisition Department

[0545] 12500: Encephalograph

[0546] 13000: Pain judgment value generation unit

[0547] 14000: Pain Assessment Department

[0548] 15099: Object

[0549] 15100: Pain Determination System

[0550] 15200: Encephalogram Measurement Department

[0551] 15220: Encephalograph

[0552] 15250: Brainwave Recording Sensor

[0553] 15270: Brainwave Amplification

[0554] 15300: Pain Determination Device

[0555] 15400: Brainwave Signal Processing Department

[0556] 15500: Brainwave feature extraction unit

[0557] 15600: Pain Assessment Department

[0558] 15700: Pain determination value generation unit

[0559] 15800: Determination level visualization department

[0560] 15900: Stimulation Device Department

[0561] 15920: Stimulus presentation

[0562] 15960: Stimulus Information Visualization Department

Claims

1. A method for operating a system for identifying or evaluating pain, the system comprising: an electroencephalogram data input unit for inputting electroencephalogram data or analysis data thereof; and an analysis unit, which analyzes the brain wave data or its analysis data; the working method includes the following steps: The analyzing unit compares all or part of the electroencephalogram data or its analysis data for the period of 2000 milliseconds from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds after the application of the target stimulus with the electroencephalogram data or its analysis data after the same time point from the application of the reference stimulus. The electroencephalogram data or its analysis data are compared over the entire range from the earliest time point of 2000 milliseconds among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds. The analysis unit determines in the comparison whether there is a duration of time when the value of the brain wave data or its analysis data obtained by the object stimulation is different from the value of the brain wave data or its analysis data obtained by the reference stimulation. If such a duration exists, if the different state does not disappear when the baseline of the waveform is returned to the baseline through linear correction of the brain wave waveform, and disappears when the specified low-frequency band component is cut off, it is determined that the pain has a persistent characteristic and there is uncomfortable pain.

2. The method of operating the system according to claim 1, wherein: The brain wave data or the analysis data thereof includes all or part of the brain wave data or the analysis data thereof within a range of 2000 milliseconds from the mid-term time period.

3. The method of operating the system according to claim 2, wherein: The intermediate time period includes values ​​ranging from 250 milliseconds to 600 milliseconds.

4. The method of operating the system according to claim 1, wherein: The whole or part includes a range that is at least 100 milliseconds long.

5. The operating method of the system according to claim 4, wherein: In the range of at least 100 milliseconds, when a statistically significant difference is observed from the range of at least 100 milliseconds, it is determined that persistence is observed.

6. The method of operating the system according to claim 1, wherein: The brain wave data or analysis data thereof is potential, duration or a combination thereof.

7. The operating method of the system according to claim 1, wherein: The determination of pain is a determination of the discomfort of pain.

8. The operating method of the system according to any one of claims 1 to 7, wherein: The method further comprises the step of analyzing the compared data using Sigmoid function fitting.

9. The operating method of the system according to claim 1, characterized in that: The above-mentioned determination is made based on the positive component of the electroencephalogram data or its analysis data.

10. The operating method of the system according to claim 9, characterized in that: If the positive component persists even after the mid-term period, it is determined that pain is present.

11. A program product for causing a computer to execute a method for identifying or evaluating pain, the method comprising the steps of: Comparing all or part of the electroencephalogram data or analyzed data from the 2000 millisecond period starting from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component, and the 250 millisecond period after the application of the target stimulus with the electroencephalogram data or analyzed data from the same time period after the application of the reference stimulus, The electroencephalogram data or its analysis data are compared over the entire range from the earliest time point of 2000 milliseconds among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds. The method comprises the following steps: in the comparison, determining whether there is a duration in which the value of the brain wave data or its analysis data obtained by the object stimulation is different from the value of the brain wave data or its analysis data obtained by the reference stimulation; if such a duration exists, if the different state does not disappear when the baseline of the waveform is returned to the baseline by linear correction of the brain wave waveform, and disappears when the specified low-frequency band component is cut off, it is determined that the pain has a persistent characteristic and is uncomfortable.

12. A recording medium storing a program for causing a computer to execute a method for identifying or evaluating pain, the method comprising the following steps: Comparing all or part of the electroencephalogram data or analyzed data from the 2000 millisecond period starting from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component, and the 250 millisecond period after the application of the target stimulus with the electroencephalogram data or analyzed data from the same time period after the application of the reference stimulus, The electroencephalogram data or its analysis data are compared over the entire range from the earliest time point of 2000 milliseconds among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds. The method comprises the following steps: in the comparison, determining whether there is a duration in which the value of the brain wave data or its analysis data obtained by the object stimulation is different from the value of the brain wave data or its analysis data obtained by the reference stimulation; if such a duration exists, if the different state does not disappear when the baseline of the waveform is returned to the baseline by linear correction of the brain wave waveform, and disappears when the specified low-frequency band component is cut off, it is determined that the pain has a persistent characteristic and is uncomfortable.

13. A system for identifying or evaluating pain, the system comprising: an electroencephalogram data input unit for inputting electroencephalogram data or analysis data thereof; and an analyzing unit that compares all or part of the electroencephalogram data or analyzed data thereof for a period of 2000 milliseconds from the earliest time point among the evoked electroencephalogram component, the initial event-related potential component, and the 250 milliseconds after application of the target stimulus with the electroencephalogram data or analyzed data thereof after the same time period after application of the reference stimulus, The electroencephalogram data or its analysis data are compared over the entire range from the earliest time point of 2000 milliseconds among the evoked electroencephalogram component, the initial event-related potential component, and 250 milliseconds. The analysis unit determines in the comparison whether there is a duration of time when the value of the brain wave data or its analysis data obtained by the object stimulation is different from the value of the brain wave data or its analysis data obtained by the reference stimulation. If such a duration exists, if the different state does not disappear when the baseline of the waveform is returned to the baseline through linear correction of the brain wave waveform, and disappears when the specified low-frequency band component is cut off, it is determined that the pain has a persistent characteristic and there is uncomfortable pain.

Images