JP5909388B2 - Biological light measurement device - Google Patents

Description

  The present invention relates to biological measurement, and more particularly to a biological optical measurement device that removes a noise signal superimposed on a desired biological signal and obtains the desired biological signal with high accuracy.

A measurement signal obtained by measurement of internal biological information using light (for example, Patent Document 1) includes a noise signal having several causes in addition to a target signal. In order to reduce the influence of the noise signal, an averaging or a band pass filter is used. The method based on addition averaging is a noise signal reduction method based on the premise that brain activity shows the same response to the same task. The band-pass filter method is a noise signal reduction method on the premise that the brain activity signal and the noise signal exist in different frequency bands.
These are processing for a plurality of noise signals, but processing specialized for noise signals due to a specific cause is also being studied.

  For example, a noise signal (pulse noise signal) due to the influence of a pulse can be reduced by a band cut filter because the frequency can be easily specified. In addition, the pulse noise signal included in the optical measurement signal is reduced based on the pulse signal measured at a part such as an ear (Patent Document 2), the pulse noise signal is extracted from the optical measurement signal itself, and the influence is extracted. A technique for reducing this has been devised (Non-Patent Document 1).

  In addition, there is a method in which optical measurement signals measured at multiple points are decomposed into a plurality of signals by signal analysis to remove noise and extract a target signal (Patent Document 3).

  The independent component analysis method is described in Non-Patent Document 2 and the like.

JP-A-9-135825 JP 2004-173751 A JP-A-2005-143609

Maria Angela Franceschini et al., NeuroImage 21 (2004) 372-386 Aisaka et al., Physica D: Nonlinear Phenomena Volume 194, Issues 3-4, 15 July 2004, Pages 320-332 Katura et al., J Biomed Opt. 2008 Sep-Oct; 13 (5): 054008. Sato et al., Neuro Image. 2004 Apr; 21 (4): 1554-62

  There is a method of imposing a task on a subject in order to obtain a desired biological signal. For example, when a subject performs a simple exercise at a predetermined timing in order to obtain a desired brain activity signal, a systemic hemodynamic change due to an increase in heart rate associated with the exercise can be a noise. At this time, there is a high possibility that the noise signal superimposed on the desired biological signal changes at the same timing as the desired biological signal. In such a case, there is a risk of misinterpreting the obtained signal.

  When the target signal is brain activity that is a response to some experimental stimulus such as imposing a task, systemic hemodynamic changes that occur in synchronization with the experimental stimulus occur in synchronization with the brain activity that is the target signal Therefore, it is very difficult to separate them. However, discrimination between the two is essential.

  The inventors have so far proposed a method of separating these signals by a signal separation method such as an independent component analysis method (Non-Patent Document 3). However, this method requires a sufficient number of data groups to perform clustering analysis, and is difficult to apply to a small number of data groups.

  Accordingly, one object of the present invention is to obtain a target signal from a measurement signal of light transmitted through an optical path in the head reflecting a change in oxygenated hemoglobin concentration or deoxygenated hemoglobin concentration indicating brain hemodynamic changes. The object is to discriminate between a waveform and a noise component, which is a systemic hemodynamic change mixed therein, and to effectively obtain a brain activity signal from which the noise component has been removed.

  A representative biological light measurement device according to the present invention includes, for example, a light irradiation unit that irradiates light to a plurality of irradiation points on a subject's head in order to measure hemodynamics of the subject's head, and Obtained by one or a plurality of light detection units that detect light that has passed through or reflected in the living body, a separation calculation unit that separates an optical signal obtained by the light detection unit, and a separation calculation unit A selection unit that selects a signal from the separation signal according to a selection criterion, a reconstruction calculation unit that reconstructs the signal from the separation signal selected by the selection unit, and parameter input in the separation calculation unit, the selection unit, and the reconstruction calculation unit It is comprised by the display part which displays a result.

  More specifically, measurement signals indicating changes in oxygenated hemoglobin concentration and changes in deoxygenated hemoglobin concentration by using a light source having a plurality of wavelengths from the visible to the near-infrared region as the measurement light source of the light irradiation unit. (Non-patent Document 3 etc.). For the measurement signal of one or both elements obtained in this way, the separation calculation unit performs, for example, independent component analysis, and separates the measurement signal into a plurality of independent components (separated signals). The selection unit selects the separation signal based on the similarity between the two model wave forms.

  For example, if one of the model waveforms is a brain activity model waveform and the other is a synchronous noise model waveform, the correlation coefficient with the brain activity model waveform is greater than a certain value and the correlation with the brain activity model waveform A separated signal whose number is larger than the correlation coefficient with the synchronous noise model waveform is selected as a brain activity signal. Thereby, noise removal is realized.

  It enables separation / removal of noise components generated in synchronization with the experimental task and mixed in the signal, and provides a signal with less noise.

It is a block diagram which shows the apparatus structure of the Example of this invention. It is a wave form diagram which shows the example of the measurement signal in the said Example. It is an example of the isolation | separation signal calculated by the isolation | separation calculating part of the said Example. It is a model waveform currently recorded on the recording part of the said Example. It is a conceptual diagram which shows the concept of the discriminating method which discriminate | determines a separated signal by the division | segmentation on the two-dimensional space which the correlation coefficient of each separated signal and two model waveforms produces. It is another example of the model waveform corresponding to the synchronous noise of the said Example.

  FIG. 1 shows a specific configuration diagram as an embodiment.

  The biological light measurement interface unit 110 is attached to a part or the whole of the subject's head. Mixed light having a wavelength of 690 nm and 830 nm emitted from the light source of the light irradiation unit 101 is guided to the interface unit 110 by the irradiation optical fiber 111. As a result, the living body is irradiated with light from a plurality of irradiation points determined when the interface unit 110 is mounted. An optical fiber 114 for detection is attached to the interface unit 110, and each end is connected to a photodetector of the photodetector 102. Thereby, each light detector that has passed through the living body and reached each detection point is detected.

In the embodiment, in order to obtain a brain activity signal, the subject is caused to periodically execute a predetermined task, and a measurement signal waveform based on the optical signal measured by the above configuration is recorded in the storage unit 103 during that period. Imposing a periodic task serves as a stimulus to the brain activity of the subject, and the purpose of measurement in the embodiment is to acquire a hemodynamic change in the brain that responds to the stimulus. In some cases, a visual stimulus or the like is periodically given to the subject instead of the periodic task. These are hereinafter referred to as experimental stimuli. Here, the wavelength of the irradiated light may be other. Three or more combinations of wavelengths may be used. Further, a measurement signal waveform obtained by performing some arithmetic processing on the optical signal obtained by the light detection unit 102 may be recorded in the storage unit 103.

  Next, the recorded measurement signal waveform is read by the separation calculation unit 121, and a separation calculation based on the method and setting selected by the user is performed. A signal separated by the separation calculation unit (hereinafter referred to as a separation signal) is manually or automatically selected by the selection unit 122 depending on the method and setting selected by the user, and the signal selected by the selection unit 122 (hereinafter referred to as selection). The signal is reconfigured by the reconstruction calculation unit 123 based on the method and setting selected by the user. Parameter input and result display in the separation calculation unit, selection unit, and reconstruction calculation unit are performed on the display unit 124. Hereinafter, details will be described.

  First, the light detection signal obtained by each detector must be separated and identified by the path through which light has passed. A typical method for identifying the path is a method of irradiating light from a plurality of irradiation points close to one detection point, not simultaneously but sequentially. The detection signal from the detector is distributed by a demultiplexer corresponding to the sequential switching on the irradiation side, and used as a measurement signal for each path. As another method of path identification, it is also effective to modulate the intensity of irradiation light at a plurality of irradiation points at different frequencies. In this case, the photodetection unit synchronously detects the detection signal from the detector using the reference signal of each frequency, thereby obtaining each measurement signal that identifies which irradiation point is the starting light. Alternatively, a technique may be employed in which the digital amplitude modulation based on a unique code is applied to the irradiation light at each of the plurality of irradiation points. The light irradiation unit 101 and the light detection unit 102 are configured to realize any one of the path identification methods described above.

  The light irradiation unit 101 and the light detection unit 102 also have means for separating and identifying the measurement signals whose paths are identified as described above into measurement signals of each wavelength component. For this wavelength identification as well, for example, a mixed light of multiple wavelengths with different amplitudes, that is, intensity modulation with different modulation frequencies, is used for each wavelength light source drive signal. A technique for separating and identifying each wavelength component by frequency can be employed. Patent Document 1 described above describes an optical measurement device that uses intensity modulation with different modulation frequencies for both identification of a path, that is, a measurement point and identification of a wavelength component. Of course, a combination in which intensity modulation having a different modulation frequency is employed in only one of the path, that is, the measurement point identification and the wavelength component identification, and the other is another technique is also effective.

  The waveform of the measurement signal at each measurement point separated and identified for each path and wavelength component as described above is output from the light detection unit 102 to the storage unit 103 and recorded. Next, the sampling values of the measurement signal with the wavelength of 690 nm and the measurement signal with the wavelength of 830 nm at each measurement point are read out, and the oxygenated hemoglobin concentration at the measurement point is calculated from both values and stored in the storage unit 103 again. In this way, the waveform of the oxygenated hemoglobin concentration change at each measurement point is calculated, and the result is stored in the storage unit 103. Not only oxygenated hemoglobin but also deoxygenated hemoglobin concentration change may be calculated and used. The details of the technique for calculating the relative concentration values of oxygenated hemoglobin and deoxygenated hemoglobin from the amount of transmitted light having two wavelengths transmitted through the body are described in Non-Patent Document 4.

  FIG. 2 shows an example of oxygenated hemoglobin concentration change at each measurement point calculated. Waveforms 201, 202, 203, and 204 are oxygenated hemoglobin concentration change waveforms at different measurement points, with the horizontal axis representing time (seconds) and the vertical axis representing hemoglobin concentration change (mM · mm). As described above, how to respond to brain activity (hemodynamic change), the degree of mixing of systemic hemodynamic changes that should be distinguished from brain activity, the degree of mixing of sensor noise, and the like vary depending on the measurement point.

  Next, a specific calculation method performed by the separation calculation unit 121 will be described.

  The separation calculation unit 121 separates the measurement signal at the measurement point, here the signal waveform indicating the oxygenated hemoglobin concentration change, into a plurality of component waveforms that compose it additively, and separated into separate signal waveforms that are statistically independent of each other To do. As one of separation methods, the TDDICA method (Non-patent Document 2) can be used. This method is one of the methods called an independent component analysis method.

  A measurement signal X represented by (Equation 1) consisting of sampling points T at four measurement points is a linear mixed signal X = a mixture coefficient matrix A of the original signal S consisting of n independent components represented by (Equation 2). Assuming that AS is represented,

The mixing matrix A and the original signal S are estimated from only the independent condition of the original signal component and the measurement signal X. Estimation is performed by whitening and rotation.

  Whitening is changing to a linearly independent signal by mixing measurement signals. It is given as (Equation 3) using the variance-covariance matrix C of the measurement signal X.

The estimated value U of the original signal S composed of independent components is given by (Expression 4) using an appropriate rotation matrix R.

It is the above independent condition that determines this rotation matrix. Here, it is defined as independent that the delay correlation becomes zero at K delay times τ _k (k = 1, 2,..., K). In reality, it is difficult to realize this, but instead, the condition is to minimize the magnitude of the correlation. This is realized by the rotation matrix R that minimizes the loss function expressed by the following equation (Equation 5).

Here, x ′ represents a transpose and <•> _t represents a time average. The estimated value of the mixing coefficient matrix is given by (Equation 6).

  Noise applied to each sensor can be handled explicitly. In that case, instead of X = AS, (Equation 7) in which the observation noise N is added to the right side is used.

At this time, not whitening but pseudo whitening is performed, and the noise component covariance matrix G is used as shown in (Expression 8).

The subsequent steps are the same as when sensor noise is not considered. The matrix G needs to be estimated separately. It is possible to fit a section of the rest state with a polynomial and estimate the matrix G from the residual at that time.

  Although the case where the number of measurement points is four has been described, separation can be performed by a similar procedure with a general number. The separation signal U and the mixing coefficient sequence R are obtained by the above calculation.

  Here, in order to perform signal separation with higher accuracy, a filtering process may be performed before performing the separation operation. When the frequency band of the target signal is known, a filtering process for reducing a band different from the frequency band is effective. In other words, in this embodiment, a filter that suppresses signal components in a band that passes through and deviates from the frequency band of the response of the brain activity to the periodic task or stimulus imposed on the subject, and separates the measurement signal that has passed through the filter. It is desirable to input to the department.

  An example of the separated signal waveform obtained by the separation operation as described above is shown in FIG. Reference numerals 301, 302, 303, and 304 in FIG. 3 denote separated signal waveforms obtained by separation calculation from the oxygenated hemoglobin concentration change waveform at one measurement point. The shaded portion in the figure indicates the period during which the subject has executed the task. The measurement purpose of this embodiment is to obtain a brain activity signal waveform. However, there are various types of component waveforms that additively constitute the oxygenated hemoglobin concentration change waveform. In other words, in addition to the separation waveform that shows the noise signal that can be clearly excluded as the possibility of reflecting the brain activity associated with the task is low, such as not being synchronized with the executed task, it changes in synchronization with the task execution, A separated waveform having the same characteristics as the response signal of the brain activity appears in that there is a waveform reproducibility between repeated tasks. The latter is called synchronous noise.

  Therefore, in this embodiment, the separated signal is discriminated based on the two-dimensional space created by the correlation coefficient between each of the two model waveforms and each separated signal, and only the brain activity signal is selected. Next, selection of the separation signal by the selection unit 122 will be described.

  First, the selection unit 122 generates two model waveforms based on the model function recorded in advance in the recording unit, using information on the time timing of the experimental stimulus actually applied to the subject at the time of signal measurement. The An example of the model waveform to be created is shown in FIG. 401 is a first model waveform representing a response of typical brain activity to an experimental stimulus. Reference numeral 402 denotes a second model waveform representing a typical synchronous noise signal.

  Then, a correlation coefficient with the first and second model waveforms is calculated for each separated waveform in FIG. Then, the separated signal is discriminated by the classification in the two-dimensional space created by the correlation coefficient between each of the first and second model waveforms and each separated signal, and the brain activity signal is selected. FIG. 5 is a conceptual diagram showing the discrimination method. The horizontal axis in FIG. 5 is the axis of the correlation coefficient between the first model waveform and each separated signal, and the right end is the maximum value of the correlation coefficient. The vertical axis in FIG. 5 is the axis of the correlation coefficient between the second model waveform and each separated signal, and the upper end is the correlation coefficient maximum value. In the selection unit 122 of the present embodiment, the value of the correlation coefficient with the first model waveform is larger than a predetermined threshold in the two-dimensional space created by these two axes, and the correlation with the second model waveform A separation waveform belonging to a region larger than the number value is selected as a separation waveform reflecting brain activity. In FIG. 5, a boundary line 501 is a line whose correlation coefficient value with the first model waveform is the predetermined threshold value. That is, the separated waveforms corresponding to the points 513 and 514 plotted in the region on the left of the boundary line 501 are excluded because the value of the correlation coefficient with the first model waveform is smaller than the predetermined threshold value. The boundary line 502 is a line in which the correlation coefficient with the second model waveform is equal to the correlation coefficient with the first model waveform. Therefore, the separated waveform corresponding to, for example, the point 512 on the upper left side of the boundary line 502 is Since the correlation coefficient with the first model waveform is smaller than the correlation coefficient with the second model waveform, it is excluded. Eventually, a separation waveform corresponding to, for example, the point 511 plotted on the right side of the boundary line 501 and the lower right side of the boundary line 502 is selected as a separation waveform reflecting brain activity. In the example shown in FIG. 5, only one separated waveform is selected. However, it is not necessarily one separated waveform that is selected as a separated waveform reflecting brain activity, and there may be a plurality of separated waveforms.

  Next, the reconstruction calculation unit (123) performs signal reconstruction based on the signal selection by the selection unit. This signal reconstruction processing is reconstructed by a reconstruction method of an independent component analysis method. Specifically, when there are a plurality of selected brain activity signals, they are multiplied by the mixing coefficient corresponding to the brain activity signal selected by the above procedure (however, the mixing coefficient is obtained by independent component analysis). Add together to obtain a reconstructed waveform.

  In the above-described embodiment, the first and second model waveforms are generated based on the model function recorded in advance in the recording unit. The model function used to generate the model waveform is preferably prepared and recorded for each type of experimental stimulus that is repeatedly applied to the subject during signal measurement. That is, it is preferable that the first model waveform generation function and the second model waveform generation model function can be selected from a plurality of recorded model functions in accordance with the actually applied experimental stimulus. These model functions are derived from measurement results to which experimental stimuli are applied. It is more preferable that the model function can be selected in accordance with the distinction when the measurement results for each age, or for the healthy person and the non-healthy person are accumulated.

  Further, the first model waveform representing the response of typical brain activity to the experimental stimulus is generated from the model function derived from the measurement result as described above, and the second model waveform representing the synchronous noise is generated from the first model waveform. It is also possible to use a differential waveform of the model waveform. Further, the second model waveform may be used as a model waveform (an N-shaped pattern model waveform) defined by a descending pattern followed by a rising pattern followed by a rising pattern as shown in FIG.

  The present invention is expected to increase the accuracy of measurement signals of biological optical measurement devices and optical topography devices, widen the range of use, and further increase the contribution of this type of device to the medical field and medical analysis field. .

101 Light irradiation unit
102 Photodetector
103 Memory
110 Interface section
111 optical fiber
121 Separation calculation unit
122 Selector
123 Reconstruction calculation unit
124 Display
201 Changes in oxygenated hemoglobin concentration at measurement point 1
202 Changes in oxygenated hemoglobin concentration at measurement point 2
203 Change in oxygenated hemoglobin concentration at measurement point 3
204 Changes in oxygenated hemoglobin concentration at measurement point 4
301 Separation signal 1 of oxygenated hemoglobin concentration change
302 Separation signal 2 of oxygenated hemoglobin concentration change
303 Separation signal 3 of oxygenated hemoglobin concentration change
304 Separation signal 4 of oxygenated hemoglobin concentration change
401 Model waveform 1
402 Model waveform 2
501 Border 1 by condition
Boundary line 2 based on 502 conditions
511 Plot point corresponding to one of the separated signals
Plot points corresponding to one of 512 separated signals
513 Plot points corresponding to one of the separated signals
514 Plot point corresponding to one of the separated signals
601 Model waveform corresponding to synchronous noise

Claims (6)

In order to measure the hemodynamics of the subject's head, a light irradiation unit that emits light from one or more irradiation points,
A light detection unit for detecting light passing through or reflecting the head of the subject at one or more detection points, and
A recording unit for recording the optical signal waveform of the measurement site based on the detection output of the light detection unit;
A separation operation unit that additively configures the optical signal waveform and separates the optical signal waveform into a plurality of separated signal waveforms that are statistically independent from each other;
The degree of similarity with the first model waveform and the degree of similarity with the second model waveform are calculated for each of the separated signal waveforms separated by the separation operation unit, and the plurality of separated signal waveforms are calculated based on these similarities. A selector that selects a separated signal waveform that reflects the response of the brain activity;
A biological light measurement device having a display unit for reconstructing the response of the subject from the selected separated signal waveform and displaying the result ,
The optical signal waveform to be recorded by the recording unit is an optical signal waveform measured while giving a periodic task or stimulus to the subject, and the first model waveform used by the selection unit is the periodic task or A model waveform representing a response of typical brain activity to a stimulus, and the second model waveform is a model waveform representing synchronous noise;
The selection unit has a two-dimensional structure of a first correlation coefficient between the first model waveform and each separated signal waveform and a second correlation coefficient between the second model waveform and each separated signal waveform. A biological light measuring device, wherein a plurality of separated signal waveforms are selected according to a division in space . The selection unit reflects a response of brain activity to a separated signal belonging to a category in which the value of the first correlation coefficient is greater than or equal to a predetermined threshold value and greater than the value of the second correlation coefficient The biological light measuring device according to claim 1 , wherein the biological light measuring device is selected as a separated signal waveform to be selected. The biological light measurement apparatus according to claim 1 , wherein the first and second model waveforms used in the selection unit are each generated from a model signal pattern recorded in the recording unit. The biological light measurement apparatus according to claim 1 , wherein the second model waveform is generated by differentiating the first model waveform generated from a model signal pattern recorded in the recording unit. . The biological light measurement apparatus according to claim 1 , wherein the second model waveform is a model waveform defined by an ascending wave type, a subsequent falling wave type, and a subsequent rising wave type pattern. The optical signal waveform read from the recording unit is subjected to a filtering process for suppressing a signal component in a band that passes through a frequency band of a response of brain activity and deviates from the frequency band, and is input to the separation calculation unit. The biological light measurement device according to claim 1 .