JP6004430B2 - Biological light measurement device - Google Patents

Description

  The present invention relates to a biological light measurement device that measures brain function using light.

Near-infrared brain function measurement (fNIRS) is already known as a biological light measurement device, and in order to perform multipoint measurement in near-infrared brain function measurement (fNIRS), a large number of incident probes and detection probes are listed on the head. When the near-infrared light is irradiated into the brain from the incident probe and the brightness of the light emitted from the detection probe is measured, the near-infrared light is absorbed by hemoglobin in the living body. Changes can be estimated. In addition, since the absorption coefficient for light wavelength differs between oxyhemoglobin and deoxyhemoglobin, changes in oxy and deoxyhemoglobin concentrations can be determined by measuring with near-infrared light of different wavelengths, which allows the state of brain function to be determined. It is to be measured.
In order to measure accurately, it is important that the contact between each probe and the head surface is stably maintained for a large number of incident probes and detection probes arranged on the head surface. In order to detect unstable contacts, conventionally, the amount of change in total hemoglobin at a certain time interval (see Patent Document 1) and the residual that occurs when oxy and deoxyhemoglobin changes are determined by the least squares method in three-wavelength measurement ( (See Patent Document 2) and the like, and when these were observed larger than a certain standard, the channel was determined to be unstable.

JP 2004-261265 A JP 2006-109964 A

In the conventional detection of defective contacts, attention is paid to the magnitude of noise for each channel. For this reason, it is possible to detect which channel has the problem, but it is not possible to identify which of the incident probe and the detection probe constituting the channel is the problem, and narrowing down bad contacts. Could not do enough. Also, if there is a problem with a channel, trial and error will be done to remove the hair directly under the probe and to devise a method for binding the probe. In this case, adjustment of both probes that make up the channel Therefore, in some cases, the probe on the side having no problem may be tampered with, and conversely, the contact may be deteriorated.
In addition, noise caused by defective contacts can be motion artifact (noise due to body movement) and measurement noise (white noise). The probe adjustment method cannot be changed accordingly. In general, motion artifacts cause more damage to data than measurement noise, but it is not possible to make a flexible response such that only probes with large motion artifacts are adjusted and other probes are not tampered with.

As the evaluation of the probe contact, the steady magnitude of the measurement light attenuation and the temporal fluctuation of the attenuation degree are considered. It can be shown that the magnitude of the measurement noise is determined from the steady magnitude of the attenuation, and the magnitude of the motion artifact is determined from the magnitude of the temporal fluctuation of the attenuation. For this reason, two kinds of noise magnitudes in each channel are obtained from the measurement data, and on the contrary, the steady magnitude of attenuation of the measurement light at the contact and the temporal fluctuation of the attenuation degree are calculated, and these are calculated. Used as an evaluation value for measurement noise and motion artifact of the contact. However, since the channel is composed of the incident probe and the detection probe, the magnitude of noise in each channel is determined by the product of the evaluation of the two probes. In multipoint measurement, since probes are used across several channels, the evaluation value of each probe is obtained by solving the inverse problem using information on which probe is used by each channel.
The measurement data (absorbance data) of each channel includes a signal resulting from a change in the blood flow of the living body, and this signal is a hindrance from the viewpoint of detecting a defective contact. According to the Modified Beer-Lambert law, which is a basic law of near-infrared brain function measurement, this signal is distributed on a plane (referred to as a BL plane) spanned by extinction coefficient vectors of oxy and deoxyhemoglobin. Since noise such as motion artifacts is not necessarily on this plane, it is possible to extract only the noise component by extracting a component orthogonal to this plane of the absorbance data (referred to as BL plane orthogonal component). In addition, since motion artifacts and measurement noise have different frequency spectra, only the respective components can be extracted by combining a low-pass filter and a high-pass filter. FIG. 9 shows a processing flow of the present invention.

That is, according to the present invention, a large number of incident probes and detection probes are arranged on the head surface, near-infrared light having three or more wavelengths is irradiated into the brain from the incident probes, and the luminance of light emitted from the detection probes is measured. From the change in absorbance obtained, the change in oxy and deoxyhemoglobin concentration is determined to measure the state of brain function, and the BL plane orthogonal component is extracted as a noise component from the change in absorbance to detect probe contact failure. In the biological light measurement device, a low-pass filter and a high-pass filter are provided, and the output obtained by low-pass filtering the extracted noise component is based only on motion artifacts, and the output obtained by performing high-pass filter processing on the extracted noise component is based only on measurement noise. It is characterized by being decomposed into two.
Further, the present invention provides a motion artifact and measurement included in a change in absorbance in the living body optical measurement device due to a steady magnitude of attenuation of measurement light at the contact of the incident probe and the detection probe and temporal fluctuation of the attenuation. Since the magnitude of noise is determined, the inverse problem is solved from the low-pass filtered output to calculate the temporal fluctuation of the measurement light attenuation at each probe contact, and the inverse problem is calculated from the high-pass filtered output. The steady-state magnitude of attenuation of the measurement light at the contact of each probe is calculated, and these are used as evaluation indexes for the motion artifact and measurement noise of the contact.
Further, the present invention is characterized in that the biological light measurement device further comprises a monitor for displaying an evaluation index for the calculated motion artifact and an evaluation index for measurement noise.
The present invention is also characterized in that, in the biological light measurement device, a display device for displaying the evaluation index for the calculated motion artifact and the evaluation index for measurement noise for each probe is installed in each probe.

In the conventional detection of defective contacts, attention is paid to the magnitude of noise for each channel. For this reason, it is possible to detect which channel has a problem, but it is not possible to identify which problem is the incident probe or detection probe that constitutes the channel, and it is possible to sufficiently narrow down bad contacts. Could not do. Also, if there is a problem with a certain channel, both probes that make up that channel must be adjusted, but in some cases, the probe on the side with no problem is tampered with, and conversely the contact deteriorates. It could have happened. The method of the present invention can pinpoint which probe contact has a problem rather than a channel, and there is no such problem.
Furthermore, the method of the present invention can indicate whether the problem arising from the defective contact is a motion artifact or measurement noise. Since the causes of motion artifacts and measurement noise are different, using the proposed method, it is possible to take different approaches for probe adjustment. In addition, since the motion artifact usually gives more problems to the data than the measurement noise, it is possible to select only the probe related to the motion artifact without adjusting the probe related to the measurement noise. As a result, it is possible to avoid a situation in which a probe that has not caused a motion artifact is mistakenly tampered with and is unstable.

It is explanatory drawing of BL plane (plane P). It is a figure for demonstrating fNIRS multipoint measurement used for actual measurement. (a) shows a high-density probe arrangement, black circles are incident probes, white circles are detection probes, and lines represent each channel. There are 16 incident probes and 16 detection probes, and 44 channels are formed from these probes. (b) is the figure explaining the place which mounted | wore the high-density probe on the head. (A), (b), and (c) in FIG. 3 correspond to the subject ABC, and are measured with ch-20 marked with “*” in FIG. 2, and the left figure shows the change in absorbance (780 nm). The figure on the right side shows the BL plane orthogonal component. FIG. 4 is a diagram showing a result of processing the orthogonal component of FIG. 3 with a low-pass filter with a cutoff frequency of 0.1 Hz and a high-pass filter with 1.0 Hz. FIG. 5 shows the measurement result of the subject B. FIG. 5A is a diagram in which the BL plane orthogonal component is obtained from the obtained absorbance data of each channel, and the size (dispersion) is displayed as a map. Further, after processing this component using a low-pass filter (0.1 Hz) and a high-pass filter (1.0 Hz), the magnitude (variance) of each channel is also displayed as a map (b) and (c). FIG. 6A is a diagram showing α _{1 to} α ₃₂ obtained from the output of the low-pass filter, and FIG. 6B is a diagram showing β ₁ to β ₃₂ obtained from the output of the high-pass filter. FIG. 7 shows the measurement result of the subject A. FIG. 7A is a diagram in which the BL plane orthogonal component is obtained from the obtained absorbance data of each channel, and the magnitude (variance) is displayed as a map. Further, after processing this component using a low-pass filter (0.1 Hz) and a high-pass filter (1.0 Hz), the magnitude (variance) of each channel is also displayed as a map (b) and (c). FIG. 8 shows the measurement result of the subject C. FIG. 8A is a diagram in which the BL plane orthogonal component is obtained from the obtained absorbance data of each channel, and the magnitude (dispersion) is displayed as a map. Further, after processing this component using a low-pass filter (0.1 Hz) and a high-pass filter (1.0 Hz), the magnitude (variance) of each channel is also displayed as a map (b) and (c). It is the figure which showed the outline of the flow of a measurement data process in this invention.

  The modified Beer-Lambert rule, which is fundamental in the near-infrared brain function measurement (fNIRS), is given by the following equation when the difference in optical path length depending on the wavelength is ignored.

と

が張る平面Ｐ上にあることがわかる(図１参照)。ここでは、この平面ＰをＢＬ平面と呼ぶ。このため、計測信号中にヘモグロビン濃度変化とは異なる波長依存性を持つ何らかの吸光度変化(ノイズ、アーティファクト)がある場合、３波長以上の計測光を用いることにより、そのような吸光度変化を、Ｐに直交する成分として検出することが可能である。本論文では、このような直交成分をＢＬ平面直交成分と呼ぶこととする。一方、２波長計測の場合には、どのような吸光度変化もそれをヘモグロビン濃度の変化に帰することが可能であるため、このような検出はできない。このためｍ≧３である必要があり、本論ではｍ＝３とする。
モル吸光係数行列を

とすると、吸光度変化ａの平面Ｐに対する直交成分ｈは次のように求められる。 Where ΔA _i , ^ε HbO _{, λi} , ^ε HbR _{, λi} are absorbance changes for wavelength λ _i , molar absorption coefficient of oxy and deoxyhemoglobin, ΔHbO and ΔHbR are concentration changes of oxy and deoxyhemoglobin, and L is the optical path It is long. From this equation, the change in absorbance caused by changes in oxy and deoxyhemoglobin concentrations is the molar extinction coefficient vector of oxy and deoxyhemoglobin with respect to the wavelength used for the measurement.

When

As shown in FIG. Here, this plane P is called a BL plane. For this reason, if there is any absorbance change (noise, artifact) having a wavelength dependency different from the hemoglobin concentration change in the measurement signal, such absorbance change can be converted to P by using measurement light of three wavelengths or more. It is possible to detect as orthogonal components. In this paper, such orthogonal components are referred to as BL plane orthogonal components. On the other hand, in the case of two-wavelength measurement, any change in absorbance can be attributed to a change in hemoglobin concentration, and thus such detection cannot be performed. For this reason, it is necessary that m ≧ 3, and m = 3 in this discussion.
Molar extinction coefficient matrix

Then, the orthogonal component h with respect to the plane P of the absorbance change a is obtained as follows.

Here, E ⁺ is a pseudo inverse matrix of E. When using three wavelengths, the above equation can be written as:

Here, u is the third column vector of U (vector not corresponding to the singular value) when the singular value decomposition of E is E = USV ^T. As can be seen from this equation, the components of the vector h have the same time series except for their magnitudes. Therefore, the BL plane orthogonal component h as a scalar quantity is defined by the following equation and used.

The following factors can be considered as factors causing the BL plane orthogonal component.
(1) Motion artifact Since the probe is basically on the hair, if the distance between the tip of the probe and the skin changes due to body movement, or if the angle of the probe with respect to the head surface changes, the contact with the probe changes accordingly. The attenuation of the measurement light changes, and this is observed as a change in absorbance. Moreover, when the hair moves, the state of absorption and reflection of measurement light by the hair changes, which may lead to a change in absorbance. Artifacts caused by such factors are called motion artifacts. Since the motion artifact is caused by body movement, it is considered to have a relatively low frequency component.
(2) Measurement noise White noise mainly caused by photoelectric change in the detection probe. Since this occurs uncorrelated with each wavelength, it cannot be explained by changes in hemoglobin concentration. Since the absorbance is calculated from the logarithm of the light intensity, the magnitude of the measurement noise included in the absorbance change of each channel is inversely proportional to the intensity of the measurement light detected by the detection probe. For this reason, if there is a probe with a large attenuation of light at the contact, regardless of whether it is an incident probe or a detection probe, the measurement noise of the channel related to the probe is greatly observed.

As described above, the change in the contact causes a change in the intensity of the measurement light, which becomes a motion artifact. Further, the magnitude of the measurement noise varies due to the steady attenuation of the measurement light at the contact. These aspects of motion artifacts and measurement noise are directly related to the way the contact between the probe and the head surface, so here we give some definitions for the attenuation of the measurement light at the contact.
Assume that the attenuation of the measurement light at the contact of the incident probe or the detection probe is r (t) (0 ≦ r (t) ≦ 1, and the smaller r (t) is, the greater the attenuation), and r (t) is as follows: Decomposes into two components.

はｒ₀のまわりでの減衰度のふらつきを示している。

の平均を０とする。
ｆＮＩＲＳの各チャンネルは一対の入射プローブと検出プローブで構成されるため、そのチャンネルで観測される計測ノイズの平均的な大きさは、入射プローブと検出プローブにおける定常的な減衰の大きさの積によって定まる。ｒ₀が小さいコンタクト(弱いコンタクトと呼ぶ)では平均的に計測光が強く減衰するため、関係するチャンネルには大きな計測ノイズが惹起されることになる。この意味から、計測ノイズに関する各コンタクトの評価をｒ₀(あるいはｌｏｇｒ₀)の大きさを用いて行うことができ、そのような評価値で測られるコンタクトの特性をコンタクトの弱さと呼ぶ。
同様に、入射プローブおよび検出プローブにおける減衰のふらつきの積によって、モーションアーティファクトの時間的変化が定まる。

の大きさ(分散)が大きいコンタクト(不安定なコンタクトと呼ぶ)を含むチャンネルでは、体動等によるノイズが強く観測されることになる。このため、計測ノイズの場合と同様に、モーションアーティファクトに関する各コンタクトの評価を

（あるいは

）の分散を用いて行うことができ、そのような評価値で測られるコンタクトの特性をコンタクトの不安定性と呼ぶ。 Where r ₀ represents the steady decay magnitude, and

Indicates the fluctuation of the degree of attenuation around r ₀ .

The average of is assumed to be 0.
Since each channel of fNIRS is composed of a pair of incident probe and detection probe, the average magnitude of measurement noise observed in that channel depends on the product of the steady attenuation magnitudes of the incident probe and detection probe. Determined. A contact with a small r ₀ (referred to as a weak contact) attenuates the measurement light strongly on average, which causes a large measurement noise in the related channel. In this sense, each contact regarding measurement noise can be evaluated using the magnitude of r ₀ (or logr ₀ ), and the contact characteristic measured with such an evaluation value is referred to as contact weakness.
Similarly, the product of the attenuation wander at the entrance and detection probes determines the temporal variation of the motion artifact.

In a channel including a contact having a large size (dispersion) (called an unstable contact), noise due to body movement or the like is strongly observed. For this reason, as in the case of measurement noise, each contact is evaluated for motion artifacts.

(Or

The contact characteristic measured by such an evaluation value is called contact instability.

  In multi-point measurement of brain activity by fNIRS, an incident probe and a detection probe are used across several channels. For this reason, it is necessary to determine to which probe the noise of the measurement data originates. Here, a method for evaluating the instability / weakness of each probe contact using the BL plane orthogonal component defined above as a clue is proposed.

(Probe arrangement matrix)
It is defined as a probe arrangement matrix how the incident probe and the detection probe configure each channel.
_Let M _s and M _{d be} the number of incident probes and detection probes, respectively, and N be the total number of channels. If the incident probe i (1 ≦ i ≦ M _s ) and the detection probe j (M _s + 1 ≦ j ≦ M _s + M _d ) constitute a channel k (1 ≦ k ≦ N), the probe arrangement matrix G is N × ( M _s + M _d ) matrix, and its element g is determined as follows.

(BL plane orthogonal component caused by probe contact failure)
Assume that the incident probe i (1 ≦ i ≦ M _s ) and the detection probe j (M _s + 1 ≦ j ≦ M _s + M _d ) constitute a channel k (1 ≦ k ≦ N). Since measurement noise is mainly photoelectric conversion noise at the detection probe, if photoelectric conversion noise n _j (t) at detection probe j is superimposed on light I _k (t) measured at channel k, Then, the absorbance change in channel k is

It is. If the magnitude of the detection probe noise is sufficiently smaller than the magnitude of the measurement light (n _j (t) << I _k (t)),

とする。入射プローブ ｉからの入力光の強度をＩ_i、チャンネルｋの生体における減衰をＲ_k(ｔ)とすると、 Can be approximated. It can be seen that the second term on the right side is measurement noise, and its magnitude is inversely proportional to the intensity of the measurement light.
Next, the attenuation of the measurement light at contacts i and j, respectively, according to equation (5)

And When the intensity of the input light from the incident probe i is I _i and the attenuation in the living body of the channel k is R _k (t),

Because

It becomes. Here, the first term is a change in absorbance associated with a change in hemoglobin in the tissue, the second term is the effect of the fluctuation of the contact on the absorbance change, and the third term is the measurement noise.
Considering the above equation for the three wavelengths used for measurement, the BL plane orthogonal component is obtained using equation (4). Since the first term of equation (13) is the change in absorbance associated with the change in hemoglobin, the orthogonal component is 0 and can be ignored. If the attenuation of the measurement light at the contact does not depend on the wavelength, the BL plane orthogonal component h _k (t) is given by the following equation.

  The first term on the right side is a term related to motion artifacts, and the second term is a term related to measurement noise. here

1 is a column vector in which all elements are 1.

(Evaluation of contact instability)
For the term h _{motion, k} (t) relating to the motion artifact, the following equation is obtained from the equation (14). Here, the average of both sides was set to zero.

として、これをすべてのｉ(１≦ｉ≦Ｍ_s＋Ｍ_d)についてまとめたベクトルをｄ(ｔ)とすると Let h _{motion, k} (t) be a vector summarizing all k, and let h _motion (t)

Assuming that d (t) is a vector in which all i (1 ≦ i ≦ M _s + M _d ) are combined

  Can be written. G is a probe arrangement matrix. from now on

Is required. The instability of the contact of each probe is evaluated by the variance of each d _i (t), and a vector summarizing this is written as α.

(Contact weakness assessment)
For the term h _{noise, k} (t) relating to measurement noise, the following equation is obtained from equation (14).

とし、また生体での減衰についても各チャンネルおよび波長で等しくＲ_k，λ(ｔ)＝Ｒ₀と仮定する。入射プローブからの入力光の強度も入射プローブおよび波長によらず一定Ｉ₀とする。さらに、各検出プローブでのノイズｎ_j，λ(ｔ)の大きさも、検出プローブおよび波長によらず一定σ_n ²であり、波長間で独立であると仮定する。これらの仮定から、ｃ₂(ｔ)の分散は

となるため、ｈ_noise,k(ｔ)の分散を

とすると、次式が得られる。 To consider the steady noise level observed with each detection probe,

It is also assumed that attenuation in the living body is equal for each channel and wavelength as R _{k, λ} (t) = R ₀ . The intensity of input light from the incident probe is also constant I ₀ regardless of the incident probe and wavelength. Furthermore, it is assumed that the magnitude of noise n _{j, λ} (t) at each detection probe is constant σ _n ² regardless of the detection probe and wavelength, and is independent between wavelengths. From these assumptions, the variance of c ₂ (t) is

_Therefore, the variance of h _{noise, k} (t)

Then, the following equation is obtained.

はi、jに依存しないため、r_i,0、r_j,0のスケールを問題としない場合には、

とできる。

Is independent of i and j, so if the scale of r _{i, 0} and r _{j, 0} is not a problem,

And can.

  Here, taking the logarithm of both sides,

Let s _noise be a vector summarizing logσ _{noise, k} for all k, and let β _i = logri _{i, 0} be a vector summarizing all i (1 ≦ i ≦ M _s + M _d ) as β. Then

  Can be written. from now on

Is required. The smaller the β _{i, the} weaker the contact.

(Decomposition of h _k (t) into h _{motion, k} (t) and h _{noise, k} (t))
In order to evaluate the instability / weakness of the contact, it is necessary to decompose the BL plane orthogonal component h _k (t) into motion artifact h _{motion, k} (t) and measurement noise h _{noise, k} (t). For this reason, a low-pass filter and a high-pass filter are applied to h _k (t). Since h _{noise, k} (t) is white noise, the output of the low-pass filter is very small. However, since h _{motion, k} (t) is a component generated with body movement, the low-pass filter has a very large effect. Not receive. As a result, the output of the low-pass filter can be used as h _{motion, k} (t). On the other hand, h _{motion, k} (t) is hardly included in the output of the high-pass filter. h _{noise, k} (t) also includes a component that varies with the fluctuation of the contact, but this is hardly included in the output of the high-pass filter. For this reason, the output of the high-pass filter can be used as h _{noise, k} (t).
H _{motion, k} (t) and h _{noise, k} (t) determined by such processing may not be correct in absolute magnitude (scale), so contact instability / weakness estimated from it The scale of the evaluation value may not be correct. However, in order to identify relatively unstable or weak contacts from a large number of contacts, it is not necessary to evaluate the absolute size, and it is sufficient if the relative evaluation value between each contact is known. Therefore, there is no problem with such processing.

(Identification of unstable / weak probe contacts)
The instability and weakness of each probe contact can be evaluated by the magnitude of α _i and β _i . However, since it is difficult to inspect and correct a large number of probe contacts at the same time in an actual experiment site, a contact with particularly large instability / weakness is identified among the contacts, and inspection and correction are given priority. It is efficient to quickly realize good measurement. Therefore, a contact i in which α _i and β _i satisfy the following criteria is selected as a contact having a particularly large instability / weakness.

Here, μ _α and σ _α are α _i , and μ _β and σ _β are the average and standard deviation of β _i . t _α and t _β are parameters that determine how many contacts are selected.

An example of a high-density probe arrangement for fNIRS multipoint measurement used by the present inventors is shown in FIG. A black circle is an incident probe, a white circle is a detection probe, and a line represents each channel. There are 16 incident probes and 16 detection probes, and 44 channels are formed from these probes. When this is mounted on the head, the center portion of the probe arrangement is matched with the Cz position of the EEG 10-20 system (see FIG. 2 (b)). As the NIRS apparatus, OMM-3000 manufactured by Shimadzu Corporation was used. The measurement light has three wavelengths of 780 nm, 805 nm, and 830 nm. The sampling interval of measurement is 0.13 seconds, and the output from the measuring instrument is not filtered.
The test subjects (A, B, C) were measured using this probe arrangement. For each subject, the task of alternately opening and closing the entire palm was performed, and the change in absorbance at that time was measured. The experiment was conducted with the approval of the National Institute of Advanced Industrial Science and Technology ergonomics experiment committee.
From the obtained change in absorbance, the BL plane orthogonal component was determined using equation (4). A portion of what appears to be typical is shown on the right side of FIG. These are all measured with ch-20 (channel marked with “*” in FIG. 2A), and FIGS. 3A, 3B, and 3C correspond to subjects A, B, and C. To do. The part colored in gray indicates the period during which the task was performed. In (b), a slow drift over the entire measurement time is observed, and in (c) a large and abrupt fluctuation is observed throughout. On the other hand, although measurement noise which seems to be white noise is on each channel, the magnitude varies depending on the subject, and it is understood that subjects A and B are relatively large and subject C is extremely small. The corresponding original absorbance change (780 nm) is shown on the left side of the figure, but the absorbance change that seems to be derived from the hemoglobin concentration change seen there is hardly observed in the right-hand orthogonal component.
In order to separate the motion artifacts and measurement noise seen in FIG. 3, the orthogonal component was processed with a filter. FIG. 4 shows the results of processing with a low-pass filter having a cutoff frequency of 0.1 Hz and a high-pass filter having a 1.0 Hz. The one processed with the low-pass filter (left side in FIG. 4) shows the motion artifact, and the one processed with the high-pass filter (right side in FIG. 4) shows the measurement noise.
From the above results, the instability / weakness of each probe contact is evaluated as follows. First, the BL plane quadrature component is obtained from the change in absorbance of all the channels obtained by fNIRS imaging using equation (4), and a low-pass filter (0.1 Hz) and a high-pass filter (1.0 Hz) are applied thereto. . By using the output of the low-pass filter as h _{contact, k} (t) in the equation (19), d _i (t) is obtained, and the variance α _i of each element is used as an evaluation of the instability of each probe contact. Further, β _i is obtained by using the logarithm of the standard deviation of each channel output of the high-pass filter as s _{noise, k} , and each element thereof is used as an evaluation of the weakness of each probe contact.

FIG. 5A shows the BL plane orthogonal component obtained from the absorbance data of each channel obtained for the fNIRS multipoint measurement of the subject B, and the magnitude (variance) displayed as a map. Also, after processing this component using a low-pass filter (0.1 Hz) and a high-pass filter (1.0 Hz), the magnitude (variance) of each channel is also displayed as a map as shown in FIG. ) And (c). It can be seen that the noise shown in FIG. 5A is successfully separated into two types of components shown in FIGS. 5B and 5C by such filter processing.
Next, d _i (t) is obtained by using the output of the low-pass filter as h _{motion, k} (t) in equation (19), and the variance α _i of each element is obtained. It is obtained as follows. α _{1 to} α ₁₆ indicate instability of the incident probe contact, and α _{17 to} α ₃₂ indicate instability of the detection probe contact. The higher the value, the higher the instability. As can be seen, in this data, large alpha ₁₄ protrudes, it can be seen that the 14th of the incident probe is unstable.
Further, when β _i is obtained by using the logarithm of the standard deviation of the output of the high-pass filter as s _{noise, k} , it is obtained as shown in FIG. β _{1 to} β ₁₆ indicate the weakness of the incident probe contact, and β _{17 to} β ₃₂ indicate the weakness of the detection probe contact. In this case, the smaller the value, the weaker the contact. As can be seen from the figure, there are no poorly evaluated contacts for contact weakness, but some contacts (β ₁₀ , β ₁₈ , β ₂₁ ) are given bad evaluation.
According to the above definition, contacts with particularly large instabilities / weaknesses were selected and written with white marks on the maps (b) and (c) of FIG. The round mark is the incident probe contact, and the square mark is the detection probe contact. Here, t _α = t _β = 1.5. As a result, contacts with instability / weakness can be selected effectively.
Finally, the results of performing the same processing for two examples of fNIRS multipoint measurement data of other subjects are shown in FIGS. 7 and 8 (t _α = t _β = 1.5). 7 and 8 show the results of subjects A and C, respectively. In either case, good results have been obtained. In the case of subject C, the stability is generally poor, but it can be seen that there are very few weak contacts.

Since each channel is composed of a pair of incident probes and detection probes, when noise in a certain channel is large, the cause is in one or both of the incident probe and detection probe constituting the channel. When only the channel is viewed, it is not known where the cause is, but it is possible to identify the cause probe by comprehensively viewing the channels related to those probes. For example, ch18 in FIG. 5A (channel composed of the incident probe S6 and the detection probe D2) is a channel that shows a large noise, but it is orthogonal that the noise is mainly derived from measurement noise. It can be found by decomposing the components with a filter and calculating the respective variances (see FIGS. 5B and 5C). Further, it can be seen from FIG. 5C that the noise is mainly caused by the detection probe D2. Certainly, the channel related to the detection probe D2 (channel around D2) shows relatively large noise, but the channel related to the incident probe S6 shows no significant noise other than ch18. As described above, it is possible to pinpoint a problematic probe by examining the noise of the entire channel analytically and comprehensively.
If unstable or weak probe contact is detected, remove the hair between the probe and the scalp as much as possible, and recheck how to bundle the fibers and how to fix the probe holder to the head. To stabilize the probe. After that, by applying this method again, it is confirmed whether or not this correction was successful. By repeating such a cycle several times, the entire contact can be converged to a stable one.
In the conventional detection of defective contacts, attention is paid to the magnitude of noise for each channel. For this reason, it is possible to detect which channel has a problem, but it is not possible to identify which problem is the incident probe or detection probe that constitutes the channel, and it is possible to sufficiently narrow down bad contacts. Could not do. Also, if there is a problem with a certain channel, both probes that make up that channel must be adjusted, but in some cases, the probe on the side with no problem is tampered with, and conversely the contact deteriorates. It could have happened. The method of the present invention can pinpoint which probe contact has a problem rather than a channel, and there is no such problem.
Furthermore, the method of the present invention can indicate whether the problem arising from the defective contact is a motion artifact or measurement noise. Since the causes of motion artifacts and measurement noise are different, using the proposed method, it is possible to take different approaches for probe adjustment. In general, motion artifacts have a greater problem with data than measurement noise, but it is possible to select only adjustment of probes related to motion artifacts without regard to probes related to measurement noise. As a result, it is possible to avoid a situation in which a probe that has not caused a motion artifact is mistakenly tampered with and is unstable.
It is easiest to display the position of the detected unstable or weak probe contact on the monitor of the measuring device. In this case, however, the position of the detected probe must be confirmed by the monitor, while the correction of the probe needs to be performed near the subject, and in order to stabilize the entire contact, there is nothing between the monitor and the probe. You have to come and go again and again. On the other hand, if a device such as an LED capable of displaying the instability and weakness of the contact can be attached to each probe, the position of the problematic probe can be intuitively displayed to the experimenter. By doing this, the position of the problematic probe can be known at a glance, so that the stabilization process for the entire contact can be rapidly advanced.

Claims (4)

Absorbance change obtained by arranging many incident probes and detection probes on the head, irradiating the brain with near-infrared light of 3 wavelengths or more from the incident probes, and measuring the luminance of the light emitted from the detection probes From the measurement of the state of brain function by determining the change in oxy and deoxyhemoglobin concentration from the above, in the biological light measurement device that detects the contact failure of the probe by extracting the BL plane orthogonal component as a noise component from the absorbance change,
A low-pass filter and a high-pass filter are provided, and the output obtained by low-pass filtering the extracted noise component is decomposed into two as a result of only the motion artifact, and the output of the extracted noise component as a high-pass filter is decomposed into two. A biological light measuring device characterized by the above.   Since the magnitude of motion artifacts and measurement noise included in the change in absorbance are determined from the steady magnitude of the attenuation of the measurement light and the temporal fluctuation of the attenuation at the contact of the incident probe and the detection probe, the low-pass filter processing is performed. The inverse problem is solved from the output to calculate the temporal fluctuation of the attenuation of the measurement light at the contact of each probe, and the inverse problem is solved from the output of the high-pass filter to solve the attenuation of the measurement light at the contact of each probe. The living body light measurement apparatus according to claim 1, wherein a steady size is calculated and used as an evaluation index for motion artifacts and measurement noise of the contact.   The biological light measurement apparatus according to claim 2, further comprising a monitor that displays an evaluation index for the calculated motion artifact and an evaluation index for measurement noise.   The living body light measurement apparatus according to claim 2, wherein a display device that displays the evaluation index for the calculated motion artifact and the evaluation index for measurement noise for each probe is installed in each probe.

Images