JP5246798B2 - Biological tissue identification apparatus and method - Google Patents

Description

  The present invention relates to a biological tissue identification apparatus and method for identifying whether a biological tissue is a normal tissue or an abnormal tissue (such as a tumor) based on infrared reflection spectrum information.

Conventional techniques for distinguishing normal / abnormal tissue from infrared spectrum information are disclosed in, for example, the following documents.
In Patent Document 1, an optical system that splits radiated light or reflected light from a subject into a plurality of wavelength regions defined by a hyperspectrum, and light that has been split by the optical system is received for each wavelength region, and is subjected to photoelectric conversion. Calculates the intensity of light in the wavelength region based on the electrical signal and the imaging device having a plurality of pixels that generate the electrical signal, and performs hyperspectral analysis based on the relationship between the multiple wavelength regions and the light intensity in the pixel. There is disclosed a system and method comprising an analysis unit for analyzing a component of radiation or reflected light from a subject in order to identify an affected part of the subject.
That is, the technique described in the same document identifies an affected area based on the relationship between a plurality of wavelength regions and light intensity in a specific pixel (a specific portion of a living tissue).

In Patent Document 2, the effective marker for identifying, 1280 cm infrared spectrum of connective tissue ^- the slope of ^one band of the baseline, normal biopsy samples is a positive slope, cancer samples Based on the finding that it exhibits a relatively flat baseline, systems and methods are disclosed for identifying the presence of a lesion based on the presence of infrared markers in extracellular material in a biological sample.
That is, the technique described in this document identifies an affected area based on a feature indicated by a predetermined wave number (reciprocal of wavelength) in a specific part of a living tissue.

  Both technologies are characterized in that they identify normal / abnormal biological tissues based on the characteristics of the components in the wavelength direction of a specific biological tissue part. There is a limit in power.

JP2009-39280 Special table 2004-515759

  The present invention has been made in view of the above problems, and it is an object of the present invention to provide an apparatus and method that can identify normality / abnormality of a living tissue with high accuracy.

In order to solve the above problems, according to an aspect of the present invention, an infrared reflection spectrum acquisition unit that acquires the intensity of a three-dimensional infrared reflection spectrum consisting of a surface position and a wavelength direction from a biological tissue, and the infrared A biological tissue identification device having a computing means for identifying normality / abnormality of the biological tissue based on the infrared reflection intensity obtained by the reflection spectrum acquisition means, wherein the computing means is a surface of the biological tissue. There is provided a biological tissue identification device characterized in that normality / abnormality of the biological tissue is identified based on a variation in the infrared reflection intensity between adjacent voxels.
According to this configuration, normality / abnormality of a living tissue can be identified with high accuracy. This was obtained as a result of the inventor actually measuring and comparing infrared reflection spectrum information of normal tissue and abnormal tissue (tumor, cancer tissue, etc.), and the voxel values of normal tissue and tumor tissue were obtained. When the spatial variation is evaluated by the standard deviation of the voxel values in the minute space, as shown in FIG. 1, it is based on the knowledge that a large difference is observed in the vicinity of 1220-1380 nm.

  Further, the variation between the adjacent voxels is a voxel included in a voxel cube made up of an odd number of voxels whose one side is 3 or more with the evaluation target voxel at the center, and is not parallel to each other on the same surface of the voxel cube. Evaluation may be performed based on a voxel value of a voxel through which a plane including two sides passes.

  The evaluation may be performed based on a voxel value of a voxel having the same wavelength as that of the evaluation target voxel.

  In addition, evaluation may be performed by assigning a large weight to voxels near the evaluation target voxel and assigning a weight equal to or smaller than the large weight to voxels far from the near voxel.

  Further, the variation between the adjacent voxels is calculated by calculating a sum of squares of residuals of voxel values in a voxel surface composed of an odd number of voxels whose side is 3 or more and having the same wavelength, and includes an evaluation target voxel. Evaluation may be performed by summing over a wavelength region in which the wavelength on the voxel surface is a median value.

  Further, the variation between the adjacent voxels is a median value of the wavelength in the voxel surface including the evaluation target voxel with the voxel surface formed of an odd number of voxels having a side of 3 and having the same wavelength as the bottom surface. Evaluation may be performed by calculating a sum of squares of residuals of voxel values in a voxel cuboid having a height in the wavelength region as follows.

  The wavelength range of the infrared reflection spectrum may be 1220 to 1380 nm.

  Further, the infrared reflection spectrum acquisition means may acquire the intensity of the infrared reflection spectrum of the living tissue in the living body with an endoscope.

  The biological tissue identified as abnormal in the computing means may be a digestive system tumor.

Further, according to an aspect of the present invention, an apparatus capable of discriminating normality / abnormality of a biological tissue is a three-dimensional infrared reflection having a surface position and a wavelength direction obtained from the biological tissue by an infrared reflection spectrum acquisition unit. A biological tissue identification method for identifying normality / abnormality of the biological tissue based on a spectrum intensity, the step of calculating a variation between adjacent voxels of the infrared reflection intensity on the surface of the biological tissue, and the adjacent And a step of identifying normality / abnormality of the biological tissue based on variation between voxels.

  As described above, according to the present invention, normality / abnormality of a living tissue can be identified with high accuracy. Details of the effect will be described later.

The graph which showed the voxel value standard deviation in the micro space of a normal tissue and a tumor tissue. The figure which showed the structure of the biological tissue identification device which is one Embodiment of this invention. << FIG. 2A >> Whole block diagram. << FIG. 2B >> The figure which showed the structure in the case of using a linear variable filter. << FIG. 2C >> The figure which showed the structure in the case of using a grism. The processed image which calculated the differential value sum of squares in three-dimensional space with the apparatus and method of one Embodiment of this invention, and showed it with the shading. The processed image which calculated the integral of the wavelength direction of the two-dimensional dispersion | distribution by the apparatus and method of one Embodiment of this invention, and showed it with the shading. The processed image which calculated the standard deviation in a micro voxel with the apparatus and method of one Embodiment of this invention, and showed it with the shading. The figure which showed the comparison with the identification result by the embodiment of this invention, and a pathological test result.

  Below, the apparatus etc. which concern on each embodiment of this invention are demonstrated. First, an apparatus capable of sweeping the wavelength and discriminating normality / abnormality of a living tissue from a change in reflection luminance with time will be described with reference to FIG. 2A.

  This device changes the center wavelength of narrow-band illumination light with a bandwidth of 5 nm in the range of 1225 nm to 1375 nm, and in a three-dimensional data space by adding wavelength axes to x and y coordinates from continuous images acquired at intervals of 5 nm. The non-uniformity of the voxel value is evaluated on the basis of the intensity of the infrared reflection spectrum of the tissue and imaged.

  In FIG. 2A, the femto wave infrared light source 1 is generally called an SC light source. When a femtosecond infrared monochromatic laser is incident on a highly nonlinear optical fiber, the monochromatic laser beam is converted into broadband white light by a non-linear optical effect called self-phase modulation. Here, wavelength conversion is performed by a nonlinear optical fiber using a femtosecond laser having a wavelength of 1550 nm as a seed light source, and white light having a wavelength of 1100 to 2500 nm is obtained. The light generated by the femtowave infrared light source 1 is developed on the surface of the movable slit 3 by the spectrometer 2. The movable slit 3 is driven by the electrostatic actuator 4 in a triangular wave shape, and light having a half-value width of 5 nm and a center wavelength of 1225 to 1375 nm is extracted from the continuous spectrum 1220 to 1380 nm. Guided to the outer fiber 6. The image of the observation object 7 illuminated with this light is converted into an electrical signal by an infrared CCD camera 9 through an infrared fiber scope 8 composed of an infrared multi-core fiber and an objective lens, and passes through an image input board 10. Guided to the arithmetic unit 11.

  The arithmetic device 11 averages the results calculated based on the arithmetic method described later for the past 45 frames, and displays them on the image display device 12. In addition, a clock signal is sent to the control circuit 13 of the electrostatic actuator 4 so that a wavelength change of 5 nm can be obtained for each screen input in synchronization with the image capture of the image input board 10. That is, a symmetric triangular wave with a period of 0.75 seconds is output from the control circuit 13, and the amplitude of the variable attenuator R (14) is such that the output light center wavelength from the movable slit 3 is 1225-1375 nm. It is adjusted with.

Further, instead of the spectroscope 2, the movable slit 3, and the electrostatic actuator 4 in the apparatus, a linear variable filter (LVF) 23 and an LVF driving motor 24 as shown in FIG. 2B may be used. In general, the linearity is better and alignment is easier than a spectroscope. In FIG. 2B, two collimating lenses 22 are used for both the fiber (light source side) 20 and the fiber (for illumination) 21, but only one may be provided on the fiber (for illumination) 21 side. .
Similarly, a grism 26 as shown in FIG. 2C may be used. By including the grating 25, the grism 26 can shorten the distance to the collimating lens 22.

According to this apparatus, since the wavelength band on the illumination side can be set sufficiently narrow, the irradiation energy applied to the observation object 7 can be suppressed sufficiently low, and the wavelength band on the light receiving side is limited. Compared with the hyperspectral endoscope, the thermal influence on the observation object can be greatly reduced. This is effective for preventing damage to the tissue, particularly in endoscopy, although high intensity illumination is usually required for hyperspectral imaging.
In addition, since the wavelength difference is handled as image information, the image buffer and the calculation mechanism can be simplified as compared with the normally used hyperspectral image processing apparatus, and observation with excellent real-time performance is realized. In this example, the image acquisition is 1 frame 1/60 sec. However, the real-time characteristics can be improved by using a high-speed imaging camera or changing the averaging of the difference amount to an exponential weighted average. It can also be increased.
In addition, by scanning only the wavelength range that is optimal for tumor detection, even in an example using a standard image acquisition board, one image can be formed in 1.3 seconds, further improving real-time performance. Yes, it has superiority at the surgical site.

  The infrared spectrum acquisition means of the present invention includes a femto wave infrared light source 1, a spectrometer 2, a movable slit 3, an electrostatic actuator 4, a condensing lens 5, an infrared fiber for illumination 6, an infrared fiber scope 8, Corresponding to the infrared CCD camera 9, the image input board 10, the control circuit 13, and the variable attenuator R14, the computing means corresponds to the computing device 11, the infrared light source is the femto wave infrared light source 1, the spectrometer 2, the movable slit. 3, the electrostatic actuator 4, the condenser lens 5, and the illumination infrared fiber 6.

  Next, calculation means for identifying normality / abnormality of the living tissue based on the infrared reflection intensity obtained by the infrared reflection spectrum acquisition means in the present invention will be described. This calculation means brings about discrimination with higher accuracy by discriminating in consideration of both the component in the wavelength direction and the component in the surface direction of the living tissue in the portion of the living tissue.

<First Embodiment>
Normal tissue and tumor tissue are discriminated by evaluating the degree of non-uniformity of the voxel values in the three-dimensional data space obtained by adding the wavelength axis to the x and y coordinates of the tissue surface. This non-uniformity is evaluated by defining a three-dimensional differential filter in the three-dimensional hyperspectral data space and adding the absolute value or square of this output over the wavelength band of interest (1220-1380 nm). Such an operation is approximately realized by taking a voxel value difference, for example, as in the following equation.
Here, h (λ ₀ ) represents a two-dimensional image plane obtained by cutting the hyperspectral data H (x ₀ , y ₀ , λ ₀ ) at the wavelength λ ₀ . The operator D _xλ represents taking the difference in the wavelength direction while taking the difference in the x direction for the three-dimensional hyperspectral data H. Similar calculations can be defined for the y direction. Actually, it is desirable to evaluate the sum of the absolute value of the change in the x direction and the y direction and the absolute value of the change in the λ direction.
FIG. 3 is a diagram showing light and shade based on the sum of squared differential values in a three-dimensional space. According to this, the tumor tissue is shown in white in the figure, and the normal tissue is shown in black, which almost coincides with the pathological examination result as described later.

Second Embodiment
In the present embodiment, by integrating the two-dimensional dispersion in the wavelength direction, the standard deviations in the integrated minute voxels are summed over the wavelength direction, and the degree of non-uniformity is evaluated. FIG. 4 shows the hyperspectral image of the inner wall of the isolated stomach specimen, the brightness of the standard deviation in a small rectangular parallelepiped of 7 × 7 pixels in the x and y directions around the evaluation target point and the wavelength direction of 1226 to 1270 nm as It is an image expressed by values, and the brightness value of the tumor part is expressed higher than the normal range. According to this, the white portion on the figure is the tumor tissue, and the black portion is the normal tissue, which is almost consistent with the pathological examination result.
The black dots in the figure are masked artifacts due to the specular reflection of the illumination light and are irrelevant to the presence or absence of a tumor.
Here, h (i, j, k) is the reflectance (or luminance value) in the three-dimensional space represented by the spatial coordinates i, j of the hyperspectral data and the index k in the wavelength direction, k ₁ , k ₂ Is an index _indicating the lower and upper limits of the wavelength to consider, h _av (k) is the average value of voxel values in the micro hyperspectral data space with the same wavelength to consider, and C is expressed as white with the maximum luminance value A normalization constant that adjusts the gray scale.

<Third Embodiment>
In the present embodiment, the magnitude of non-uniformity is evaluated based on the reflectance dispersion value in a minute space in the three-dimensional space of hyperspectral data composed of wavelength directions and spatial coordinates.
FIG. 5 shows an image of the hyperspectral image of the extracted stomach inner wall according to each spatial coordinate as shown in the following equation for the magnitude of the standard deviation of the voxel value in the micro space at the spatial coordinate 7 × 7 × wavelength (1226-1270 nm). It has become. According to this, the tumor tissue is shown in white in the figure, and the normal tissue is shown in black, which almost coincides with the pathological examination result as described later.
Here, h (i, j, k) is the reflectance (or luminance value) in the three-dimensional space represented by the spatial coordinates i, j of the hyperspectral data and the index k in the wavelength direction, k ₁ , k ₂ Is the index _indicating the lower and upper limits of the wavelength to be considered, h _av is the average value of the voxel values in the micro hyperspectral data space to be considered, and C is a standardization that adjusts the gray scale so that the maximum luminance value is expressed as white It is a constant.
The black dots in the figure are masked artifacts due to the specular reflection of the illumination light and are irrelevant to the presence or absence of a tumor.

  FIG. 6 shows an evaluation of tumor tissue detection accuracy according to the present invention. A to D are tissue parts collected by total gastrectomy. In each of A to D, the part determined to be a malignant tumor tissue based on the pathological examination result of the tissue cross section in each tissue part is shown in the upper row, and the binarized one based on the standard deviation calculation result (third embodiment) is shown in the middle row. The binarized version based on the three-dimensional Sobel operation (first embodiment) is shown in the lower part. When each part is taken together, it can be seen that each embodiment shows a good agreement with the pathological examination result.

DESCRIPTION OF SYMBOLS 1 Femto wave infrared light source 2 Spectrometer 3 Movable slit 4 Electrostatic actuator 5 Condenser lens 6 Infrared fiber for illumination 7 Observation object 8 Infrared fiber scope 9 Infrared CCD camera 10 Image input board 11 Arithmetic unit 12 Image Display device 13 Control circuit 14 Variable attenuator R
20 Fiber (light source side)
21 Fiber (for lighting)
22 Collimating lens 23 Linear variable filter (LVF)
24 LVF drive motor 25 Grating 26 Grism

Claims (10)

An infrared reflection spectrum acquisition means for acquiring the intensity of a three-dimensional infrared reflection spectrum consisting of a surface position and a wavelength direction from a biological tissue;
Based on the infrared reflection intensity obtained by the infrared reflection spectrum acquisition means, a biological tissue identification device having a computing means for identifying normality / abnormality of the biological tissue,
The biological tissue identification device characterized in that the computing means identifies normality / abnormality of the biological tissue based on a variation between adjacent voxels of the infrared reflection intensity on the surface of the biological tissue. The variation between the adjacent voxels is
A voxel included in a voxel cube composed of an odd number of three or more voxels whose one side is located at the center of the evaluation target voxel, and passes through a plane including two parallel sides that are not on the same surface of the voxel cube Based on the voxel value of
The biological tissue identification device according to claim 1, wherein the biological tissue identification device is evaluated. In the biological tissue identification device according to claim 2, based on a voxel value of a voxel having the same wavelength as the wavelength in the evaluation target voxel,
The biological tissue identification device according to claim 2, wherein a variation between the adjacent voxels is evaluated. A voxel close to the voxel to be evaluated in the previous period is given a large weight, and voxels far from the close voxel are given the same or a smaller weight than the large weight,
The biological tissue identification device according to claim 2, wherein a variation between the adjacent voxels is evaluated. The variation between the adjacent voxels is
Calculate the sum of squares of the residuals of the voxel values in the voxel plane consisting of an odd number of voxels with 3 or more sides and voxels having the same wavelength,
By summing over the wavelength region where the wavelength in the voxel surface including the evaluation target voxel is a median value,
The biological tissue identification device according to claim 1, wherein the biological tissue identification device is evaluated. The variation between the adjacent voxels is
In a voxel rectangular parallelepiped having an odd number of voxels of 3 or more and a voxel surface consisting of voxels having the same wavelength as a bottom surface, and a height in which the wavelength region in the voxel surface including the evaluation target voxel has a median value. By calculating the sum of squares of the residual voxel values,
The biological tissue identification device according to claim 1, wherein the biological tissue identification device is evaluated. The wavelength range of the infrared reflection spectrum is 1220-1380 nm.
The biological tissue identification device according to any one of claims 1 to 6. The infrared reflection spectrum acquisition means acquires the intensity of the infrared reflection spectrum of the living tissue in the living body with an endoscope.
The biological tissue identification device according to any one of claims 1 to 6. In the calculation means, the biological tissue identified as abnormal is a digestive system tumor,
The biological tissue identification device according to any one of claims 1 to 6. An apparatus capable of discriminating normality / abnormality of a biological tissue is based on the intensity of a three-dimensional infrared reflection spectrum obtained from the biological tissue by the infrared reflection spectrum acquisition means and comprising a surface position and a wavelength direction. A biological tissue identification method for identifying normal / abnormal,
Calculating a variation between adjacent voxels of the infrared reflection intensity of the surface of the biological tissue;
Identifying normality / abnormality of the biological tissue based on variation between the adjacent voxels;
A biological tissue identification method comprising: