JP2016123823A - Mental disease classification method, discrimination method, and discrimination criteria creation and learning system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a classification/discrimination criteria generation method and a discrimination criteria learning system capable of improving the accuracy of disease classification using biological optical measurement.SOLUTION: A composite waveform obtained by combining several representative basic waveforms A to D extracted from an actual measurement signal 501 for biological optical measurement by arbitrary weighting 515 is presented as a pseudo signal. Results of deciphering composite waveforms by experts are accumulated in a database in association with the weighting 515 of the composite waveforms with the basic waveforms A to D. Using the database makes it possible to create and learn discrimination criterion for signals including no ambiguity.SELECTED DRAWING: Figure 5

Description

  The present invention relates to biological light measurement for obtaining in-vivo information, and more particularly to a technique for classifying biological light measurement signals for evaluating brain activity based on waveform characteristics.

  Psychiatric disorders are characterized by not only the highest number of patients, the second largest number of patients in the country after hypertension, but also the long-term morbidity. Therefore, improving the quality of psychiatric practice is regarded as a social request. In particular, the diagnosis that corresponds to the first stage of medical care has often been based only on conventional interviews. Interrogation is a diagnostic method that is likely to cause differences in the skills of the doctors in charge, and it is desirable to incorporate objective means to improve the quality of medical care. The introduction of objective means has the effect of helping patients understand the medical condition. In research using biophotometry technology in psychiatry, changes in biophotometric signals (hemoglobin signals) in the forehead associated with verbal fluency tasks are measured and psychiatric disorders are classified based on their waveform patterns. In this regard, it was shown that, for example, depression and bipolar disorder can be correctly classified with a probability of about 70% by using one index (centroid index), which is the center of gravity time of the amplitude of the biological light measurement signal. (Non-Patent Document 1).

Takizawa, R., et al., Neuroimaging-aided differential diagnosis of the depressive state, NeuroImage (2013) Katura, T., et al., “Extracting task-related activation components from optical topography measurement using independent components analusis, Journal of Biomedical Opitcs 13 (5), September / October 2008

  In the psychiatric application of the above-mentioned biological light measurement technology, further improvement in determination accuracy is required. There are two main reasons why a certain error is included in the determination result. One is the case where there is room for improvement in the automatic discrimination index adopted, and the other is the measurement accuracy of the brain activity signal. In the test, a psychological experiment task called “word fluency task” is generally used. However, since the internal process of executing the psychological experiment task is difficult to control from the outside, the measurement results vary. That is, it becomes an error factor of the inspection result. These two issues need to be considered separately, but it is impossible to distinguish between the two when measured biological light measurement signals are used. On the other hand, in learning to acquire a discrimination technique, it is important to use a biological optical measurement signal that does not include ambiguity as described above. However, avoid ambiguity in the measured signal. Not. Even if the signals are artificially synthesized, the parameters necessary to synthesize the waveforms of depression and schizophrenia are unknown.

  An object of the present invention is to solve the above-described problems and provide a biological light measurement device, a signal processing device, and a method thereof that can determine a biological light measurement signal with high accuracy.

  In order to achieve the above object, in the present invention, a biological light measurement signal is modeled based on a measurement unit that measures a biological light measurement signal, two or more basic waveforms of the biological light measurement signal, and the basic waveform. Measured by the measurement unit based on the weighting factor for each basic waveform, the database that accumulates the results of waveform interpretation of the biological light measurement signal, and the criteria for classification of the biological light measurement signal that are created using the database Provided is a biological light measurement device having a processing unit for determining a biological light measurement signal to be performed.

In order to achieve the above object, the present invention includes a processing unit, and the processing unit weights the biological light measurement signal for each of the plurality of basic waveforms based on two or more basic waveforms of the biological light measurement signal. Provided is a signal processing apparatus configured to model using coefficients, and to store waveform interpretation results of signals obtained by the modeling in a database.
A processing unit and a storage unit, wherein the processing unit models the biological light measurement signal using a weighting factor for each of the plurality of basic waveforms based on two or more basic waveforms of the biological light measurement signal; Provided is a signal processing device configured to store a basic waveform, a weighting factor, and a waveform interpretation result of a signal obtained by modeling as a database.

  Furthermore, in order to achieve the above object, according to the present invention, there is provided a signal processing method in a signal processing device, wherein the signal processing device outputs a biological light measurement signal based on two or more basic waveforms of the biological light measurement signal. Provided is a signal processing method for modeling using a weighting factor for each of a plurality of basic waveforms and storing the basic waveform, the weighting factor, and the waveform interpretation result of the signal obtained by the modeling as a database.

  According to the present invention, by providing a signal waveform that does not include the ambiguity of brain activity, it becomes possible to extract the criteria for discriminating the expert, and to realize highly accurate classification judgment of the biological light measurement signal.

It is a figure which shows an example of a structure of the biological light measuring device based on Example 1. FIG. It is a figure which shows an example of a structure of the signal processing apparatus based on Example 1. FIG. It is a figure which shows an example of the content of the basic waveform and database based on Example 1. FIG. It is a figure which shows the flow of the data analysis and classification method based on Example 1. FIG. It is a figure which shows an example of the discrimination tree for discrimination label output based on Example 1. FIG. It is a figure explaining an example of modeling of the actual measurement waveform based on Example 1. FIG. It is a figure which shows an example of the content of the basic waveform and database based on Example 2. FIG. It is a scatter diagram which shows an example of the waveform interpretation result based on Example 2, and a discrimination criterion. It is a figure explaining an example of modeling of the actual measurement waveform based on Example 2. FIG. It is a figure which shows the creation flow of a database based on Example 3. FIG. It is a figure which shows the learning flow using the database based on Example 3. FIG. FIG. 10 is a diagram illustrating an example of a composite waveform screen display according to the third embodiment. FIG. 10 is a diagram illustrating an example of a screen display of a waveform determination window according to the third embodiment. It is a figure which shows an example of the database reconstruction flow based on Example 4. FIG.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In this specification, a synthetic parameter when generating a pseudo signal that is a synthetic waveform by modeling with at least two or more basic waveforms extracted from the measured biological light measurement signal is referred to as “weighting coefficient” or “weighting”. Called. A signal component extracted by independent component analysis, principal component analysis, or the like is referred to as a “basic waveform”.

  In various embodiments of the present invention, which will be described in order below, a function of presenting a signal waveform obtained by combining basic waveforms, which are representative signal components extracted from a biological light measurement signal, with arbitrary weighting. provide. In addition, by providing a function that accumulates the results of interpretation of synthesized waveforms by experts as a database in association with the weighting coefficient that is the synthesized parameter of the synthesized waveform, it is possible to create discrimination criteria for signal groups that do not include ambiguity . Furthermore, based on the discrimination criterion created in this way, a discrimination method is automatically generated, and a function for discriminating an arbitrary biological light measurement signal is provided. Furthermore, a function is provided for a person who wants to learn a reading method to learn a synthesized waveform read by the accumulated expert based on the reading result.

  Signal component extraction processing such as independent component analysis and principal component analysis is performed on biological optical measurement signals measured from a plurality of persons, for example, about 50 subjects, and several signal components, for example, about three components, are used as basic waveforms. Extract. By using the basic waveform which is the signal component extracted in this way, the signal of each subject can be restored to some extent. That is, the extraction of signal components means the reduction of the signals of multiple subjects. For example, as such a signal component, two components of a brain activity component and a systemic component described in Non-Patent Document 2 by the present inventors are extracted, and a plurality of extracted signal components are recorded.

  Next, each of the plurality of signal components is arbitrarily weighted, and a combined signal is generated as a pseudo signal. At this time, the weighting as the synthesis parameter is preferably determined to be similar to the actually measured amplitude of the biological light measurement signal. The expert interprets the synthesized signal generated in this way and records the labeled waveform interpretation result together with the synthesized signal. This is performed a plurality of times for different synthesized signals, for example, 100 times or more, and the result is accumulated as a database. That is, the accumulated information is a data set of weights for each signal component and a label indicating the interpretation result. Based on this database, for example, by applying a data identification method such as a support vector machine, it is possible to define an expert discrimination standard as an objective discrimination method. By applying this discrimination method to any actually measured biological light measurement signal, the signal can be discriminated.

  The database is also useful for learners of discrimination methods. For example, a computer selects one data set from the database, synthesizes a waveform based on the weight of the data set, and presents it to the learner. The learner labels the presented waveform. If the result is the same as that of an expert, it is possible to assist the learner in reading the waveform efficiently by providing a system that makes the correct answer.

  Embodiment 1 is an embodiment relating to a biological light measurement device, a signal processing device for biological light measurement waveform classification, and a biological light measurement waveform classification method in these devices. That is, a processing unit and a storage unit are provided, and the processing unit models the biological light measurement signal based on two or more basic waveforms of the biological light measurement signal using a weighting factor for each of the plurality of basic waveforms. These are embodiments of a signal processing apparatus and method, and a biological light measurement apparatus configured to store a basic waveform, a weighting coefficient, and a waveform interpretation result of a signal obtained by modeling as a database.

  First, FIG. 1A is used to show a configuration example of a biological light measurement device that acquires, analyzes, and classifies a biological light measurement waveform of the present embodiment. In a biological light measurement device that is a biological light measurement device capable of detecting light that has entered a living body and can detect light that has been scattered, absorbed, and propagated in the living body, from one or more light sources 101 included in the device body 20 The irradiated light 30 is incident on the subject 10 through the waveguide 40. The light 30 enters the subject 10 from the irradiation point 12, passes through and propagates through the subject 10, and then passes through the waveguide 40 from the detection point 13 at a position away from the irradiation point 12. , Detected by one or more photodetectors 102. The irradiation-detector distance (SD distance) is defined by the distance between the irradiation point 12 and the detection point 13.

  Here, the one or more light sources 101 are a semiconductor laser (LD), a light emitting diode (LED) or the like, and the one or more photodetectors are an avalanche photodiode (APD), a photodiode (PD), a photoelectron amplifier, or the like. A double tube (PMT) or the like may be used. The waveguide 40 may be an optical fiber, glass, light guide, or the like.

  The light source 101 is driven by a light source driving device 103, and the gain of one or more photodetectors 102 is controlled by a control / analysis unit 106. The control / analysis unit 106 also controls the light source driving device 103 and receives an input of conditions and the like from the input unit 107. The electrical signal photoelectrically converted by the photodetector 102 is amplified by the amplifier 104, converted from analog to digital by the analog-digital converter 105, and sent to the control / analysis unit 106 for processing. In the configuration of the biological light measurement device in FIG. 1, the block configuration illustrated on the right side of the control / analysis unit 106 in FIG. 1 may be collectively referred to as a measurement unit that measures a biological light measurement signal.

  The control / analysis unit 106 is a processing unit, and executes processing such as analysis / classification based on the signal detected by the photodetector 102 of the measurement unit. Specifically, a digital signal obtained by conversion by the analog-digital converter 105 is received, and based on the digital signal, for example, a change in the oxygenated / deoxygenated hemoglobin concentration length is detected from a change in detected light amount or a change in absorbance. Calculate (oxy-Hb, deoxy-Hb). Here, the density length change is a change amount of the product of the density and the optical path length. This digital signal is also stored in a data analysis / classification database to be described later.

  Here, the control / analysis unit 106 is described assuming that it is a processing unit that performs all of the driving of the light source 101, the gain control of the photodetector 102, and the signal processing from the analog-digital converter 105. This function can also be realized by having a control unit and a means for integrating them.

  The measurement data and the hemoglobin concentration length change calculation result are stored in the storage unit 108, and the measurement result can be displayed on the display unit 109 based on the analysis result and / or the stored data.

  Although the light transmitter 50 and the light receiver 60 are not shown in FIG. 1A, the light transmitter 50 includes, for example, the waveguide 40 on the light source 101 side, and is installed in a state in contact with or close to the subject 10. The light receiver 60 includes, for example, the waveguide 40 on the photodetector 102 side, and is placed in contact with or close to contact with the subject 10. At this time, the light transmitter 50 and the light receiver 60 are arranged on the subject 10 so that the light received by each light receiver propagates through both the gray matter and the scalp. This is because, in the classification method described below, it is assumed that the brain-derived signal included in each received light signal increases approximately and linearly according to the SD distance. This is because it is necessary to include a brain-derived signal. When the SD distance is very short and the gray matter has a small average optical path length, the inclination of the brain activity component with respect to SD cannot be obtained with high accuracy.

  Next, a method for separating and extracting brain activity components and systemic components using measurement data and hemoglobin concentration length change calculation results will be described. This method uses independent component analysis (ICA) to extract multiple independent components from near-infrared spectroscopy (NIRS) signals obtained by measurement and extract them as brain activity components or It is a method of classifying into systemic components. Independent component analysis is one of the signal separation methods, and is an analysis method that can separate linearly mixed signals without prior information. There are multiple signal sources, and it is effective for analyzing data measured at multiple points. Regarding the brain activity component and the systemic component, Non-Patent Document 2 by the present inventors described above can be referred to.

  Next, FIG. 1B shows a configuration example of a signal processing apparatus that performs biological light measurement waveform classification according to the first embodiment. Unlike the configuration of FIG. 1A, the apparatus shown in FIG. 1 does not include a measurement unit that performs biological light measurement, and can be configured by a normal personal computer (PC) or the like. The measured biological light measurement signal is stored in the database 111. The database 111 corresponds to the storage unit 108 in FIG. 1A, and the display screen 112, the central processing unit (CPU) 113, and the input device 114 correspond to the display unit 109, the control / analysis unit 106, and the input unit 107, respectively. The database 111 is stored in a data center or the like connected to a network via a storage unit of a PC, an external external storage device, or a network interface unit of a PC (not shown).

  The contents of the database 111 in FIG. 1 will be described with reference to FIG. As described above, the database includes several signal components that are fundamental for expressing biological optical measurement signal analysis. Here, description will be made with four signal components 201, 202, 203, and 204. The signal component 201 corresponds to a brain activity component. In each of the signal components 201 to 204 in the left column of the figure, the horizontal axis indicates time, the vertical axis indicates amplitude, and the gray square 205 in the figure indicates the period during which the experiment subject is presented. This task period 205 is necessary when matching the task period of the actual measurement signal with the task period set in the database when discriminating the actual measurement signal.

  The tabular data 206, which is the contents of the database 111 in the right column of the figure, stores a data number, a weighting value of each signal component, a discrimination label for a synthesized waveform synthesized according to the weighting, and the like as one set. This discrimination label indicates the result of waveform interpretation by an expert. In the example, the label is N or P, indicating that 66 cases of data are stored in the database 111.

  FIG. 3 is a flowchart for explaining the flow of data analysis / classification of biological light measurement signals in the apparatus of this embodiment. In the apparatus of FIGS. 1A and 1B, this analysis / classification flow is realized by executing a program of a processing unit such as the control / analysis unit 106 or the CPU 113 under the operation of the operator.

  First, a database to be analyzed is selected (301). This is because the determination method can be determined flexibly according to conditions such as age group, sex, or measurement location. Next, a method for creating a discrimination criterion is selected (302). Several methods for discriminating labels based on the selected database have been proposed, and it is preferable to use a method suitable for the purpose. For example, a well-known decision tree or support vector machine (SVM) can be used. In discrimination criterion creation (303), a discrimination criterion is generated by the selected discrimination criterion creation method.

  Here, a discrimination algorithm generated by the decision tree technique as shown in FIG. 4 will be described as an example. In modeling of the actually measured waveform (304) in FIG. 3, modeling of the actually measured biological light measurement signal, that is, generation of a pseudo signal is performed. Although not shown in FIG. 2, measured waveform data is also stored in the database 111 and the storage unit 108. In the actual waveform modeling (304), the measured and accumulated signals are modeled by regression analysis using the waveforms A201, B202, C203, and D204 of FIG. That is, the measured / accumulated biological light measurement signal is expressed by a pseudo signal that is the sum of signal groups obtained by multiplying the basic waveforms A, B, C, and D by appropriate weighting and weighting factors.

  FIG. 5 schematically shows a specific example of modeling the measured waveform (304). The actual measurement signal 501 is modeled by four basic waveforms A to D, and the amplitudes of the four basic waveforms A to D are adjusted and adjusted by discriminating (305) the weighting coefficient most applicable to the actual measurement waveform 501. The basic waveforms are A511 to D514. Finally, from the weighting 515 of each waveform component derived by discriminating (305) the weighting factor of the actually measured waveform, the waveform interpretation result for the biological optical measurement signal of the actually measured waveform 501 based on the discrimination criteria shown in the decision tree of FIG. (Label) is given and label output (306) is performed.

  That is, the determination (402 to 406) as to whether or not the weighting of each of the basic waveforms A to D is greater than or equal to the predetermined value is executed, starting with the determination (401) that the weighting of the basic waveform C is 0.02 or more shown in FIG. , The discrimination label that is the result of waveform interpretation is P or N. In the example of the actually measured waveform 501 in FIG. 5, since the weight of the basic waveform C is 0.02 or more, the label is P in the determination (401). In the discrimination label output (504), for example, “P” is displayed as the label name on the screen 112 of FIG. 1B.

  In the biological light measurement device and biological light measurement waveform classification device of the present embodiment described above, a database composed of two or more basic waveforms, weights (coefficients), and labels that are waveform interpretation results can be sequentially stored. . This database can be effectively used in an automatic classification method, a classification learning method, and the like in a biological light measurement device and a waveform automatic classification device described in the following Example 2-4.

  In the second embodiment, a biometric light for automatically discriminating a biological light measurement signal measured by a measurement unit of a biological light measurement device is generated from a database of a large number of biological light measurement signal waveforms already created. It is an Example of a measuring device, a signal processing device, etc. That is, a measurement unit that measures a biological light measurement signal, two or more basic waveforms of the biological light measurement signal, a weighting factor for each basic waveform when the biological light measurement signal is modeled based on the basic waveform, A process for discriminating biological optical measurement signals measured by the measurement unit based on a database for accumulating the results of waveform interpretation of optical measurement signals and a discrimination criterion for classification of biological optical measurement signals created using the database It is an Example of the biological light measuring device etc. of the structure provided with a part. The apparatus configuration used in the present embodiment has the same apparatus configuration as that of the biological light measurement apparatus and signal processing apparatus shown in FIGS. 1A and 1B.

  First, an example of a database accumulated in a storage unit used in this embodiment will be described with reference to FIG. In this embodiment, a case where two types A and B are used as the basic signal components from the four types A to D shown in FIG. These two types of waveforms have the same shape as the signal components extracted as two main components of the biological light measurement signal in Non-Patent Document 2 by the present inventors. However, information in the time direction such as task period is different. The component A601 in FIG. 6 is a brain activity component, and the signal component B602 is a systemic component. The biological optical measurement signal contains information about hemodynamic changes in the measurement target area. When the measurement area is the cerebral cortex, the hemodynamic changes are caused by brain activity and systemic hemodynamic changes. Is included at least. Furthermore, since the biometric measurement is performed from the scalp, signal components other than the cerebral cortex are also included. In Non-Patent Document 2, since the brain activity component indicated by component A601 and the systemic component indicated by component B602 are the main signal components, these two types are suitable as basic signal components.

  The tabular data 604 indicating the database in the right column of FIG. 6 is the same as the tabular data 206 described in FIG. The table shows the labels that are the results of waveform interpretation. However, in modeling, it is desirable to remove noise components by pre-processing, for example, by a band pass filter or the like. The tabular data 604 is a database of 100 examples.

  In the apparatus of this embodiment, a waveform discrimination reference is created using this database. FIG. 7 shows an example in which the weights and labels, which are data in the database of FIG. 6, are illustrated in a two-dimensional plot, that is, a two-dimensional scatter diagram. In each of the left and right scatter plots in the figure, the horizontal axis represents the weighting of the waveform component A, and the vertical axis represents the weighting of the waveform component B. Label N is indicated by □ and label P is indicated by ○. As a method for discriminating a label based on the database, SVM, decision tree, etc. are known as described in the first embodiment, and it is preferable to use a method suitable for the purpose.

  An example in which a two-dimensional scatter diagram is displayed to the user and the user manually determines the boundary line is shown. That is, the user sets three boundary lines 701 to 703 that are considered to be optimal visually from the scatter diagram, and determines the discrimination region. As can be seen from the figure, the two labels N and P can be discriminated with high accuracy by these boundary lines 701 to 703. Needless to say, the setting of the discrimination criterion and the discrimination area of the waveform can be automatically set by the control / analysis unit 106 or the CPU 113, which is a processing unit, using the weighting and label data which are data in the database.

  That is, in this embodiment, the display unit displays the weighting factor for each basic waveform and the waveform interpretation result as a scatter diagram on the display screen, and the user is based on the scatter diagram displayed on the display screen. The configuration in which the discrimination criteria can be input using the input unit, or the processing unit can automatically create the discrimination criteria using the database.

  Next, a process for discriminating a newly measured biological light measurement signal by the apparatus according to the present embodiment using the created waveform discrimination reference and discrimination area will be described. First, the control / analysis unit 106 or CPU 113, which is a processing unit, uses the basic waveforms A601 and B602 recorded in the database and models the measured signal by regression analysis, that is, sets it as a pseudo signal.

  FIG. 8 shows an example of modeling. The actual measurement signal 801 is modeled by two basic waveforms, and the amplitude of each of the two basic waveforms is adjusted by weighting that is most applicable to the actual measurement waveform (811, 812). Finally, from the weighting of each waveform component derived by modeling, a label as a waveform interpretation result is given to the actual measurement signal 801 based on the waveform discrimination criteria set as described above. In the example of FIG. 8, the weights 813 of the waveforms A and B are 0.69 and 0.87, respectively. Therefore, as is apparent from the boundary lines of the waveform determination criteria in the scatter diagram of FIG. Therefore, the label based on the waveform discrimination criterion is N. In discrimination label output, for example, “N” is displayed as a label name on the display screen 113 in FIG. 1B.

  As described above, according to the present embodiment, a database consisting of basic waveforms A and B that are basic, weights for labeling using them, and labels that are the results of waveform interpretation are accumulated in the apparatus. In addition, a biometric optical measurement device capable of discriminating data of a biological optical measurement signal newly measured from a subject with high accuracy by generating a discrimination criterion and a discriminating region of a waveform using this database. can do. In addition, even in a signal processing device not provided with a measuring unit, an automatic discrimination method that can receive data of a biological light measurement signal measured by biological light measurement and can discriminate with high accuracy can be provided.

  The third embodiment is an embodiment relating to the generation of the database and the creation of the discrimination criterion described in the first and second embodiments, and the learning method for the learner to learn the created discrimination criterion using the generated database. is there. That is, it further includes an input unit, and the processing unit selects an arbitrary basic waveform and weighting factor stored in the database, performs waveform synthesis, displays the obtained synthesized waveform on the display unit, and the learner It is an Example of the biological light measuring device etc. of the structure which compares the judgment result with respect to the synthetic | combination waveform input from an input part with the waveform interpretation result memorize | stored in the database. Also in the present embodiment, the apparatus configuration described in FIG. 1A and FIG. 1B of the first embodiment can be used.

  9A and 9B show a processing flow of the third embodiment. FIG. 9A is a processing flow when a database is created, and FIG. 9B is a processing flow when a learner learns using the created database. Basically, these processing flows are realized by program processing executed by the control / analysis unit 106, the CPU 113, and the like shown in FIGS. 1A and 1B. As is apparent from FIGS. 9A and 9B, the “waveform synthesis”, “waveform display”, and “input of discrimination results” processes are common to both flows.

  First, in the basic waveform selection 900 of the database generation flow in FIG. 9A, a basic waveform used for waveform synthesis is selected. For example, the basic waveforms 201 to 204 shown in FIG. 2 and the basic waveforms 601 and 602 shown in FIG. 6 are selected. Next, in the waveform synthesis 901, an arbitrary weighting factor is randomly selected for each of the selected basic waveforms to generate a synthesized signal that is a pseudo signal. At this time, in order to prevent the combined signal from becoming extremely large or small, it is desirable to set the range of the value of the weighting factor appropriately. A value range may be set based on the actual signal amplitude. In the waveform display 902, the waveform is displayed on the display unit 109 and the screen 112 in FIGS. 1A and 1B in order to present it to a discriminator who discriminates a synthesized waveform that is a pseudo signal.

  FIG. 10 shows an example of display on the screen. The left side of the screen 1001 displays a waveform from which the slope of the entire waveform has been removed by baseline processing, and the right side of the screen 1002 displays the waveform before the slope of the entire waveform has been removed. The inclination component of the entire waveform is a component corresponding to low frequency noise in the biological light measurement signal.

  FIG. 11 shows an example of a waveform interpretation window displayed on the screen. Several buttons 1102 to 1106 are arranged in the waveform interpretation window 1101. This example shows a case where a waveform is discriminated into two labels P and N. In addition to two labels, a label P1102 and a label N1103, a label other 1105 and an unknown button 1104 are also arranged. Further, since the time from the display of the synthesized waveform to the press of the discrimination button by the learner or the like is considered to be related to the difficulty of waveform interpretation, the control / analysis unit 106 and the CPU 113 which are processing units are not shown in the figure. It is desirable to simultaneously record the time measured using the timer function or the like. Therefore, a Pause button 1106 for temporarily interrupting waveform interpretation is arranged on the waveform interpretation window 1101. That is, in the preferred embodiment of the present embodiment, the processing unit stores the time required for the learner to determine the classification of the composite waveform in the database.

  For the determination result input 903, a button displayed by an operation by the input unit 107 or the input device 114 such as a mouse is pressed. The result of the selected label is registered in the database in the database registration 904. At this time, what is registered as a series of data is the basic waveform or its name, the weighting coefficient of each basic waveform, the label name of the waveform interpretation result, and the time required for the discrimination described later. After registering in the database, returning to waveform synthesis (901), repeatedly discriminating the new synthesized waveform and registering it in the database, a database of a large number of samples can be created.

  Next, the case where the learner learns the discrimination criterion using the database created as described above according to the processing flow of FIG. 9B will be described. First, a database is selected by the database selection 905 in FIG. 9B. Select the appropriate database depending on the purpose, such as when you want to learn only for men, for example, in a specific age group, or when you want to use only a database created by a specific discriminator. It is a function to do. After selecting the database, in waveform synthesis 906, a waveform is synthesized from a data set randomly selected from the database. The synthesized waveform is displayed on the display unit 109 and the screen 112 by the waveform display unit 907. Here, if the time required for discrimination is recorded in the database for the selected data set, the time required for discrimination is interpreted as “difficulty” and a data set with an appropriate difficulty level is selected. May be.

  A display example of the obtained composite waveform is the same as the example of FIG. Next, the learner inputs the discrimination result using the screen 112, the input unit 107, and the input device 114. Here too, the display screen is the same as the example of the waveform discrimination window of FIG. The discrimination result input by the learner is compared with the label name of the waveform interpretation result stored in the database in the comparison with the database (909), and the correct or incorrect answer is displayed on the screen 112 or the like in the result display 910. Is done. At this time, the time required for the discrimination may be interpreted as the proficiency level of the discrimination technique and used for the evaluation of the discrimination learning.

  According to the present embodiment described above, it is possible to generate a database effective for a biological light measurement device, a signal processing device, and the like, and a learner can efficiently learn a discrimination criterion using the generated database. .

  Example 4 relates to a database reconstruction method that is useful when a large number of databases created and accumulated in the past are used when new knowledge is obtained about a biological light measurement signal obtained by a biological light measurement device. It is an example. That is, the processing unit performs waveform synthesis using the basic waveforms and weighting factors accumulated in the database, models the obtained synthesized waveform based on a plurality of basic waveforms different from the basic waveform, The new weighting factor obtained is re-accumulated in the database, and the biological light measurement signal measured by the measurement unit is created based on the discrimination criterion for classification of the biological light measurement signal, which is created using the re-accumulated database. This is an example of a biological light measurement device or the like configured to perform the determination.

  Here, it is assumed that a database created using the two basic waveforms 601 and 602 shown in FIG. For example, in order to extract the signal characteristics in more detail from the subsequent study of the biological light measurement signal, it is more effective to use the four basic waveforms 201 to 204 shown in FIG. According to the present embodiment, it is possible to reconstruct and reuse a large amount of accumulated database according to the present embodiment when it is found later.

  First, an existing database stored in the database selection 1201 is selected in a processing flow that can be realized by the control / analysis unit 106 as a processing unit or the program processing of the CPU 113 shown in FIG. In the selected database, two basic waveforms when the database is created are stored. Next, a basic waveform selection 1202 adds or selects a new basic waveform for database reconstruction. In the waveform synthesis 1203, a pseudo signal is generated as a synthesized waveform based on the two basic waveforms stored in the existing database and the weighting.

  Subsequently, in the waveform modeling 1204, the synthesized waveform pseudo signal generated in the waveform synthesis 1203 is modeled using the four basic waveforms newly added or selected in the basic waveform selection 1202. , A new pseudo signal is generated. This remodeling method may be the same regression analysis as in the first and third embodiments. The data remodeled in this way is registered in a new database together with the four basic waveforms as new data based on the four basic waveforms in the database registration 1205.

  According to the present embodiment, it is possible to provide a database reflecting new knowledge obtained from biological light measurement signals by effectively using a database in which a large number of examples have already been accumulated, and using a newly re-accumulated database. Thus, it is possible to classify biological light measurement signals more correctly.

  In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, in the above embodiment, the weighting and label, which are data in the database, are shown as a two-dimensional plot, that is, a two-dimensional scatter diagram. However, if there are three or more basic waveforms, Deformation such as plotting, that is, using a three-dimensional scatter diagram is performed. The above-described embodiments have been described in detail for better understanding of the present invention, and are not necessarily limited to those having all the configurations described above. Further, a part of the configuration of a certain embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of a certain embodiment. In addition, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment.

  Further, the above-described configuration, function, control analysis unit, processing unit, and the like have been described as an example of creating a program that realizes a part or all of them. Needless to say, it may be realized by hardware.

DESCRIPTION OF SYMBOLS 10 Subject 101 Light source 102 Photo detector 103 Light source drive device 104 Amplifier 105 Analog-digital converter 106 Control / analysis part 107 Input part 108 Storage part 109 Display part 111 Database 112 Screen 113 CPU
114 Input devices 201-204 Basic waveforms A, B, C, D
205 Schematic display of task period 206 Tabular display of database 501 Example of actually measured waveform 511-514 Adjusted basic waveforms A, B, C, D
515 Weighting 601 Basic waveform A
602 Basic waveform B
603 Schematic display of task period 604 Tabular display of database 701 Discrimination boundary 1
702 Discrimination boundary 2
703 Discrimination boundary 3
801 Actual measured waveform 811 Adjusted basic waveform A
812 Adjusted basic waveform B
813 Weighting 1001 Preprocessed waveform 1002 Preprocessed waveform 1101 Waveform discrimination window 1102 Label P button 1103 Label N button 1104 Unknown button 1105 Label and other buttons 1106 Pause button

Claims (20)

A measurement unit for measuring biological light measurement signals;
A database for storing two or more basic waveforms of the biological light measurement signal, a weighting factor for each basic waveform when the biological light measurement signal is modeled based on the basic waveform, and a waveform interpretation result of the biological light measurement signal;
A processing unit for determining the biological light measurement signal measured by the measurement unit based on a determination criterion for classification of the biological light measurement signal created using the database,
A biological light measurement device characterized by that. The biological light measurement device according to claim 1,
A display unit for displaying a display screen and an input unit;
The display unit displays the weighting factor for each basic waveform and the waveform interpretation result as a scatter diagram on the display screen,
The user can input the discrimination criterion using the input unit based on the scatter diagram displayed on the display screen.
A biological light measurement device characterized by that. The biological light measurement device according to claim 1,
The processing unit creates the discrimination criterion using the database.
A biological light measurement device characterized by that. The biological light measurement device according to claim 1,
A display unit for displaying a display screen and an input unit;
The processing unit selects an arbitrary basic waveform and the weighting factor stored in the database, performs waveform synthesis, displays the obtained synthesized waveform on the display unit,
The learner compares the judgment result for the composite waveform input from the input unit with the waveform interpretation result stored in the database.
A biological light measurement device characterized by that. The biological light measurement device according to claim 4,
The processing unit stores the time required for the learner to determine the classification of the composite waveform in the database.
A biological light measurement device characterized by that. The biological light measurement device according to claim 1,
The processor is
Waveform synthesis is performed using the basic waveform stored in the database and the weighting factor, and the resultant synthesized waveform is modeled based on a plurality of basic waveforms different from the basic waveform, and the plurality of different basic waveforms and obtained Re-accumulating the new weighting factor obtained in the database,
Based on the discrimination criteria for classification of biological light measurement signals created using the re-accumulated database, the biological light measurement signals measured by the measurement unit are determined.
A biological light measurement device characterized by that. With a processing unit,
The processor is
Based on two or more basic waveforms of the biological light measurement signal, the biological light measurement signal is modeled using a weighting factor for each of the plurality of basic waveforms,
Accumulate waveform interpretation results of signals obtained by modeling in a database,
A signal processing apparatus. The signal processing device according to claim 7,
The processing unit creates a discrimination criterion for classifying the biological light measurement signal using the database, and discriminates the measured biological light measurement signal based on the created discrimination criterion.
A signal processing apparatus. The signal processing device according to claim 7,
A display unit;
The display unit can display a weighting factor for each basic waveform and the waveform interpretation result as a scatter diagram,
A signal processing apparatus. The signal processing device according to claim 7,
A display unit and an input unit;
The display unit can display a weighting factor for each basic waveform and the waveform interpretation result as a scatter diagram,
Based on the scatter diagram displayed on the display unit, the user can input a discrimination criterion for classification of biological light measurement signals using the input unit.
A signal processing apparatus. The signal processing device according to claim 7,
A display unit and an input unit;
The processing unit selects an arbitrary basic waveform and the weighting factor stored in the database, performs waveform synthesis, displays the obtained synthesized waveform on the display unit,
The learner compares the judgment result for the composite waveform input from the input unit with the waveform interpretation result stored in the database.
A signal processing apparatus. The signal processing device according to claim 7,
The processor is
Waveform synthesis is performed using the basic waveform stored in the database and the weighting factor, and the resultant synthesized waveform is modeled based on a plurality of basic waveforms different from the basic waveform, and the plurality of different basic waveforms and obtained Re-accumulating the new weighting factor obtained in the database,
Discrimination of the biological optical measurement signal is performed based on a discrimination criterion for classification of the biological optical measurement signal created using the re-accumulated database.
A signal processing apparatus. The signal processing device according to claim 7 is provided.
A biological light measurement device characterized by that. The signal processing device according to claim 7,
A storage unit;
The storage unit
Storing the basic waveform, the weighting factor, and the waveform interpretation result as the database;
A signal processing apparatus. 15. A signal processing apparatus according to claim 14, wherein
The processing unit creates a discrimination criterion for classifying a biological light measurement signal using the database stored in the storage unit, and discriminates a measured biological light measurement signal based on the created discrimination criterion I do,
A signal processing apparatus. 15. A signal processing apparatus according to claim 14, wherein
The processing unit performs waveform synthesis using the basic waveform stored in the database stored in the storage unit and the weighting factor, and the obtained synthesized waveform is based on a plurality of basic waveforms different from the basic waveform. Modeling, re-accumulating the different basic waveforms and the new weighting factor obtained in the database,
Discrimination of the biological optical measurement signal is performed based on a discrimination criterion for classification of the biological optical measurement signal created using the re-accumulated database.
A signal processing apparatus. The signal processing apparatus according to claim 14 is provided.
A biological light measurement device characterized by that. A signal processing method in a signal processing device, comprising:
The signal processing device includes:
Based on two or more basic waveforms of the biological light measurement signal, the biological light measurement signal is modeled using a weighting factor for each of the plurality of basic waveforms,
Accumulating the basic waveform, the weighting factor, and the waveform interpretation result of the signal obtained by modeling as a database,
And a signal processing method. The signal processing method according to claim 18, comprising:
The signal processing device includes:
Using the database, create a discrimination criterion for classifying the biological light measurement signal, and based on the created discrimination criterion, determine the measured biological light measurement signal,
And a signal processing method. The signal processing method according to claim 18, comprising:
The processing unit performs waveform synthesis by selecting an arbitrary basic waveform and the weighting factor stored in the database, and displays the obtained synthesized waveform on a display unit,
The learner compares the judgment result for the composite waveform input from the input unit with the waveform interpretation result stored in the database.
And a signal processing method.

Images