JP2005296434A - Method and apparatus for analyzing spectral image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To precisely specify the position of even a slightly lesioned part at the eye fundus from a spectral image of the eye fundus. <P>SOLUTION: The spectral data of the slit image of the eye fundus is obtained at the position of a line vertical to the longitudinal direction of a slit is obtained by each line, and the spectral data at each pixel (Pi, Qj) is obtained from the spectral data of a slit image at each line. The spectral data at each pixel is compared with that of a lesion (A) to specify the position of a pixel image having the spectral data nearly coincident or close to it, and the position is displayed on a monitor 40 together with the eye fundus image. Thus, even a slightly lesioned part which is often passed by can be detected to realize early and precise diagnosis. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a spectral image analysis method and apparatus, and more particularly to a spectral image analysis method and apparatus for acquiring and spectrally analyzing a spectral image of a subject such as a fundus of a subject's eye.

  2. Description of the Related Art Conventionally, as a method for diagnosing a subject, diagnosis is performed by color imaging of an affected area, detecting a lesioned portion, and displaying that portion. For example, the fundus of a subject is photographed, and only a specific color component is extracted from the fundus photographed image for image diagnosis. However, for that purpose, it was necessary to acquire a special interpretation called image interpretation.

  Thus, a spectral image of a subject is acquired and the spectral image is analyzed (Patent Document 1 and Patent Document 2). However, those described in these documents obtain the temperature distribution of the subject (Patent Document 1) or display the spectral pattern of the lesioned part (Patent Document 2), and accurately determine the position of the lesioned part from the spectral image. There is a disadvantage that it cannot be specified.

Also, from Patent Document 3, a fundus camera is known that can record a spectral image of a slit image in an eyeball for each of a plurality of slit positions. Since the image obtained from the fundus camera has spectral information for each pixel, it is possible to evaluate a specific change in wavelength. There is a drawback that technology is required.
JP 2003-166880 A JP-A-61-33639 JP 2002-224041 A

  An object of the present invention is to solve the above-mentioned conventional problems. By analyzing a spectral image of a subject, even if there is an anomalous portion having a slight change in the subject, the position can be accurately identified. Is to be able to do that.

The present invention (Claim 1)
A spectral image analysis method for acquiring spectral data of a subject and performing spectral analysis,
Capture the spectral image of the subject, obtain the spectral data of the pixel image for each pixel of the subject,
Compare the spectral data of each pixel image of the subject with the spectral data of the reference pixel image,
Based on the comparison result, the position of the pixel image having spectral data substantially coincident with or close to the spectral data of the reference pixel image is specified.

The present invention (Claim 8)
A spectral image analysis device for acquiring spectral data of a subject and performing spectral analysis,
Computing means for capturing a spectral image of the subject and obtaining spectral data of the pixel image for each pixel of the subject;
Comparison means for comparing the spectral data of each pixel image of the subject and the spectral data of the reference pixel image;
And a specifying means for specifying the position of a pixel image having spectral data that substantially matches or is close to the spectral data of the reference pixel image based on the comparison result.

  In the present invention, there is provided storage means for capturing spectral data of a slit image of a subject at a line position perpendicular to the longitudinal direction of the slit, and storing the spectral data of the captured slit image for each line. The spectral data of the slit image at each line stored in the storage means is read, each line is divided into a predetermined number of pixels, and the spectral data of the pixel image is obtained for each pixel.

  In the present invention, the spectral data of the pixel image is obtained for each pixel of the subject from the spectral image of the subject, and the spectral of the reference pixel image is obtained based on the comparison result between the spectral data of each pixel image and the spectral data of the reference pixel image. Since the position of the pixel image having spectral data that substantially matches or is close to the data is specified, the position of the abnormal part (lesioned part) of the subject can be specified in units of pixels, and accurate diagnosis is possible. Become.

  The spectral image analysis apparatus of the present invention specifies the position of a specific portion such as a lesion from the spectral image of the subject, and will be described by taking the fundus of the eye to be examined as an example of the subject. A fundus camera is used. However, the present invention is not limited to the fundus only, but can be applied to analysis of spectral images of other subjects.

  Hereinafter, the present invention will be described in detail based on embodiments shown in the drawings.

  In the fundus camera shown in FIG. 1, light emitted from a light source 1 such as a halogen lamp removes reflections from a condenser lens 2, a strobe 3, a ring slit 4 for forming ring-shaped illumination, a lens 5, and an objective lens 9. Is reflected by a perforated total reflection mirror 8 having a hole in the center, passes through an objective lens 9, and enters the fundus 10a from the pupil of the eye 10 to be examined. Illuminate.

  The reflected light from the fundus 10a is received again from the pupil of the eye to be examined through the objective lens 9, is passed through the hole of the perforated total reflection mirror 8, and is narrowed by the photographing aperture 14 disposed behind it, and the focus lens 15 Then, the light passes through the imaging lens 16 and enters the rotating mirror 17. The light reflected by the rotating mirror 17 passes through the mask 18 arranged at a position conjugate with the fundus, then is reflected by the mirror 19, passes through the lens 20, and enters the infrared CCD 21 for observation. Since the infrared CCD 21 is arranged at a position conjugate with the fundus in the observation optical system, the examiner can observe the fundus image formed by the infrared CCD via the monitor 40 and perform alignment and focusing operations. It can be carried out. In this observation mode, a filter 1 ′ that transmits infrared light and cuts visible light is inserted into the illumination optical path so that observation with the infrared CCD 21 can be performed with a non-mydriatic pupil.

  When acquiring a spectral image, the mirror 19 is detached from the optical path, and the fundus image illuminated by the light source 1 is once formed by the mask 18, passes through the lens 22, and extends in a direction perpendicular to the paper surface. Re-imaged at 23 position. Then, the fundus image that has become a slit image is incident on an image sensor 27 including a monochrome CCD through a spectroscopic optical system including the lens 24, the spectroscopic element 25, and the lens 26.

  The spectroscopic element 25 has a configuration similar to that of the prism / grating / prism (PGP) described in Patent Document 3, and a fundus image that is formed into a slit image by the slit 23 has a predetermined shape in a direction perpendicular to the longitudinal direction of the slit. Since the spectrum is split over the wavelength bandwidth, the image sensor 27 can capture a spectral image of the fundus slit image.

  Further, the rotating mirror 17 is rotated by a rotating mirror driving mechanism 28 composed of, for example, a stepping motor by a predetermined step about the shaft 17a, and therefore a fundus image line formed on the slit 23 according to the rotation. The position changes, and the spectral data of the slit image at each line position of the fundus is captured according to the position of the rotating mirror.

  Spectral data of the fundus slit image at each line position imaged by the image sensor 27 is input to the spectral image analyzer 30, and slit image spectroscopic data is synchronized with the fundus line position obtained from the rotating mirror drive mechanism 28. Each line is stored in a memory (storage means) 32. The CPU 31 of the spectral image analyzer 30 reads the spectral data of the fundus slit image stored in the slit image spectral data memory 32, divides each line into a predetermined number of pixels, and separates the spectral image of the pixel image for each pixel. Data is calculated (calculation means). The obtained spectral data for each pixel in each line is stored in a pixel image spectral data memory (storage means) 33 for each pixel.

  Further, the spectral data of the pixel image of the fundus lesion or the part having an abnormality is measured in advance, and the spectral data of the pixel image of various lesions is used as the spectral data of the reference pixel image for each lesion. It is registered in the library (memory) 34. The CPU 31 sequentially reads the spectral data of each pixel image stored in the pixel image spectral data memory 33, compares it with the spectral data of the lesion stored in the lesion library 34 (comparison means), and refers to the reference pixel based on the comparison result. The position of a pixel image having spectral data substantially coincident with or close to the spectral data of the image is specified (position specifying means).

  In addition, a monitor 40 is connected to the spectral image analysis device 30, and the position of the specified lesioned part is displayed together with the fundus image by assigning brightness or a predetermined color.

  In such a configuration, when the alignment or focusing operation is completed for the eye to be examined, imaging preparation is completed, and a shutter button (not shown) is operated. As a result, the mirror 19 leaves the optical path, and the fundus image illuminated by the light source 1 is formed on the slit 23. Initially, as shown in FIG. 2, the position of the rotary mirror 17 is positioned so that a slit image is taken at the uppermost line position P1 of the fundus. An image is formed. Since the spectroscopic element 25 spectrally decomposes the slit image into the wavelengths λ1 to λk, the image sensor 27 calculates the spectral intensity at the wavelengths λ1 to λk of the slit image of the line P1, as shown in the upper center portion of FIG. Imaging as a spectral image (spectral waveform). This spectral image is A / D converted and stored in the slit image spectral data memory 32 of the spectral image analyzing device 30 as spectral data on the line P 1 in synchronization with the signal from the rotary mirror drive mechanism 28.

  Since the rotating mirror 17 is rotated by a predetermined amount, the lines P2. . . . . Spectral data of the fundus slit image at Pn is sequentially captured, and the spectral data of the slit image at each line is stored in the memory 32 for each position of the rotating mirror, that is, for each line. This line P2,. . . . . The spectral data of the slit image at Pn is shown in the center of FIG. Each line P1. . . . . Since the width of Pn is selected so that the image width at the slit 23 of the line is equal to the slit width of the slit 23, the rotary mirror 27 is rotated stepwise by an amount corresponding to the slit width, The entire area is scanned without a gap with a line width corresponding to the slit width.

  In this way, when the spectral image of the fundus slit image at each line position is captured, the CPU 31 of the spectral image processing apparatus reads the spectral data of the slit image at the line P1 from the slit image spectral data memory 32, A line is connected to a predetermined number of pixels Q1, Q2,. . . . . The pixel data is divided into Qm (for example, m = 768), and the spectral data of the pixel image is obtained for each pixel. The spectral data of the pixel image on the line P1 obtained in this way is the pixel (P1, Q1),. . . . . A spectral waveform of (P1, Qm) is shown in the upper right of FIG.

  The CPU 31 then continues with the lines P2,. . . . . The spectral data of the slit image at Pn are sequentially read out, and each line is similarly read from the pixels Q1, Q2,. . . . . Dividing into Qm, the spectral data of the pixel image is obtained for each pixel. The spectral data of the pixel image in each line obtained in this way is shown on the right side of FIG. 2, and these spectral data are the pixels (Pi, Qj) (i = 1,... N). ; Q = 1,... M) and stored in the pixel image spectral data memory 33.

  Here, the number m of pixels Q is the same as the number of effective pixels in the line direction of the image sensor 27, and the number (k) of spectral wavelengths λ is the number of effective pixels in the direction orthogonal to the line direction of the image sensor. If the same, spectral data with the highest resolution can be obtained, but if the amount of data becomes enormous, it can be compressed appropriately. For example, the line is divided into pixels having a size corresponding to a plurality of (for example, two) pixels in the line direction of the image sensor, and a plurality of (for example, two) pixels in a direction orthogonal to the line direction of the image sensor. By setting the average value of the spectral intensity of each wavelength in the pixel as the spectral intensity of the average value of the wavelength, the data amount can be compressed.

  Subsequently, the CPU 31 of the spectral image processing apparatus sequentially reads out the spectral data of each pixel image in each line stored in the pixel image spectral data memory 33 and obtains the pixel image of the lesion part registered in the lesion library 34. Compare with spectroscopic data.

  This state is shown in FIG. 3. For example, when it is desired to know the position where the lesion A is on the fundus, the spectral data of the lesion A registered in the lesion library 34 is read as reference pixel image spectral data. Then, it is sequentially compared with the spectral data of each pixel image (Pi, Qj) read from the pixel image spectral data memory 33.

  For example, an average value of spectral intensity differences (or squares thereof) at the respective wavelengths λ1 to λk is obtained, and when the value is equal to or smaller than a predetermined value, the spectral data of the pixel image at the predetermined position (Pi, Qj) is obtained. Then, it is determined that the spectral data substantially coincides with or close to the spectral data of the pixel image of the lesion A. When obtaining an average value, a weighted average value obtained by assigning a different weighting factor to each wavelength may be obtained. In a wavelength region peculiar to a lesion, if the weighting factor is increased, a good determination result is obtained. be able to. Since the spectral data appears as a pattern, the spectral pattern of the pixel image of the lesion A is used as a template, and the spectral pattern of each pixel image on each line, the template of the lesion A, and the like using a known template matching algorithm. Both spectral data may be compared by performing the pattern matching and obtaining the correlation value.

  Spectral data of the pixel image at the position (Pi, Qj) substantially equal to or close to the spectral data of the pixel image of the lesion A when the average value of the differences is not more than a predetermined value and / or the correlation value is not less than the predetermined value. If it is determined as data, it is determined that there is a lesion at the pixel position (Pi, Qj), and the position is specified and displayed at a corresponding position on the monitor 40 that displays the fundus image. As shown by the thick solid line on the monitor 40 in FIG. 3, the display is performed by changing the brightness in accordance with the magnitude of the correlation or by coloring with a specific color corresponding to the correlation value. The lesion is emphasized and imaged. Therefore, the examiner can detect a lesion having a slight change that tends to be overlooked, and early detection and accurate diagnosis are possible.

  In addition, the comparison between the two spectral data is performed by comparing data of characteristic points having spectral characteristics that are characteristic among the spectral data, for example, data of inflection points of spectral waveforms or data on points specified in advance in the spectral waveforms. Can be done.

  In addition to the lesion A, other various lesions B,. . . . . For X, spectral data peculiar to the lesion are measured in advance and registered in the lesion library 34, so that other lesions B,. . . . . If there is X, the corresponding position can be specified and the corresponding display can be performed. Further, the spectral data of the reference pixel image may be spectral data selected from the spectral data of the pixel image stored in the memory 33 instead of the spectral data prepared and registered in advance.

  Note that when the spectral image of the fundus is captured by the image sensor 27, the fundus is irradiated by the visible light source 1, but the fundus is illuminated with monochromatic light, infrared light, or a light flux having a predetermined bandwidth wavelength. You may make it do. In that case, since the spectroscopy is performed only in the predetermined wavelength region, the amount of calculation can be reduced.

It is the block diagram which showed the structure of the imaging device which acquires a decomposition image analysis apparatus and a spectral image. It is explanatory drawing which showed the spectral data of a slit image and a pixel image. It is explanatory drawing which showed the process of identifying the lesioned part of a fundus and displaying the lesioned part.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Eye to be examined 25 Spectroscopic element 27 Imaging element 30 Spectral image analyzer 31 CPU
32 Slit image spectroscopic data memory 33 Pixel image spectroscopic data memory 34 Lesion library 40 Monitor

Claims (14)

A spectral image analysis method for acquiring spectral data of a subject and performing spectral analysis,
Capture the spectral image of the subject, obtain the spectral data of the pixel image for each pixel of the subject,
Compare the spectral data of each pixel image of the subject with the spectral data of the reference pixel image,
A spectral image analysis method characterized by identifying a position of a pixel image having spectral data substantially coincident with or close to spectral data of a reference pixel image based on a comparison result.   The spectral data of the pixel image is obtained by acquiring spectral data of a slit image of a subject at a line position perpendicular to the longitudinal direction of the slit for each line, dividing each line into a predetermined number of pixels, The spectral image analysis method according to claim 1, wherein the spectral image analysis method is obtained for each pixel from spectral data of slit images in a line.   The spectral image analysis method according to claim 1 or 2, wherein the comparison is performed by comparing data of feature points having spectral characteristics that are characteristic among the spectral data.   4. The spectral image analysis method according to claim 1, wherein the comparison is performed by obtaining a correlation between the spectral data of the pixel image and the spectral data of the reference pixel image. 5.   5. The spectral image analysis method according to claim 3, wherein the feature point is an inflection point of a spectral waveform or a point on a spectral waveform designated in advance.   6. The spectral data of the reference pixel image is spectral data prepared in advance or spectral data selected from the spectral data of the obtained pixel image. 2. The spectral image analysis method according to item 1.   The spectral image analysis according to any one of claims 1 to 6, wherein the specified position is displayed on a monitor together with an image of a subject by assigning brightness or a specific color according to a comparison result. Method. A spectral image analysis device for acquiring spectral data of a subject and performing spectral analysis,
Computing means for capturing a spectral image of the subject and obtaining spectral data of the pixel image for each pixel of the subject;
Comparison means for comparing the spectral data of each pixel image of the subject and the spectral data of the reference pixel image;
A specifying means for specifying a position of a pixel image having spectral data substantially matching or close to the spectral data of the reference pixel image based on the comparison result;
A spectral image analysis apparatus comprising:   Storage means for capturing spectral data of the slit image of the subject at a line position perpendicular to the longitudinal direction of the slit and storing the spectral data of the captured slit image for each line is provided, and the computing means is the storage means 9. The spectral data of the slit image at each line stored in the line is read, each line is divided into a predetermined number of pixels, and the spectral data of the pixel image is obtained for each pixel. Spectral image analyzer.   10. The spectral image analysis apparatus according to claim 8, wherein the comparison unit compares data of characteristic points having spectral characteristics that are characteristic of the spectral data. 11.   11. The spectral image analysis apparatus according to claim 8, wherein the comparison unit performs a comparison by calculating a correlation value between the spectral data of the pixel image and the spectral data of the reference pixel image. .   The spectral image analysis apparatus according to claim 10 or 11, wherein the characteristic point is an inflection point of a spectral waveform or a point on a spectral waveform designated in advance.   The spectral data of the reference pixel image is spectral data prepared in advance or spectral data selected from the spectral data of the obtained pixel image. The spectral image analysis apparatus according to item 1.   14. The monitor according to claim 8, further comprising: a monitor that displays the specified position together with an image of the subject by assigning brightness or a specific color according to a magnitude of the correlation value. Spectral image analyzer.

Images