JP5870419B2 - Lens image analyzer - Google Patents

Description

  The present invention relates to a lens image analysis apparatus for assisting a doctor in diagnosing the degree of progression of cataract.

  An anterior ocular segment observation apparatus that captures an optical cut surface image of a cornea or a crystalline lens of a subject's eye using a slit lamp is known. In particular, the anterior ocular segment observation device disclosed in Patent Document 1 below can focus on a wide range of a deep object, and can capture a stable and high-quality optical cut surface image.

  In the treatment of cataracts, doctors visually determine the degree of progression of cataracts while viewing images, and explain treatment methods such as whether or not to perform surgery on patients. One of the international standards widely used in preoperative examination of cataract surgery is called NUC classification (Grading Nuclear Cataract of WHO Cataract Grading Group. “A Simplified Cataract Grading System”). . The change of the aging lens nucleus is accompanied by turbidity and coloring, but the NUC classification is defined from the viewpoint of turbidity.

  Grades based on this NUC classification are classified into four grades as shown in FIG. The doctor observes the optical cut surface image of the eye to be examined and makes a grade determination while comparing it with a standard image.

JP 2009-56149 A

  However, skill is required to determine the NUC classification, and it cannot be easily determined. Moreover, a difference may arise in a determination result by an individual difference of a doctor.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a lens image analyzing apparatus capable of objectively and automatically evaluating a NUC classification in cataract diagnosis.

In order to solve the above problems, a crystalline lens image analyzing apparatus according to the present invention includes:
Image acquisition means for acquiring optical cut surface image data of the eye to be examined obtained by photographing the eye to be examined;
From the acquired optical cut plane image data, a specific region setting means for setting a specific region that is at least partly included in the embryonic nucleus region or set in the vicinity of the embryonic nucleus region, inside the lens region,
A center-to-center layer detecting means for detecting whether a center-to-center layer exists in the set specific region;
When the center layer exists, it is determined that it belongs to the first grade group with a low NUC classification grade for nuclear cataract evaluation. When the center layer does not exist, the second grade group with a high NUC classification grade Grade judging means for judging that it belongs to the above.

  The operation and effect of the crystalline lens image analyzing apparatus having this configuration will be described. First, optical cut surface image data of the eye to be examined is acquired, and a specific region is set in the cut surface image for determination of NUC classification grade. The specific region includes at least a part of the embryonic nucleus region or is set in the vicinity of the embryonic nucleus region. Here, the embryonic nucleus is a part constituting the lens nucleus, and the lens nucleus is composed of an embryonic nucleus that is the most core part, a fetal nucleus that surrounds it, and an adult nucleus that surrounds them. The present invention focuses on the embryonic nucleus. This is because as the cataract progresses, the lens nucleus becomes harder and the turbidity in the embryonic nucleus region and its vicinity changes markedly.

  When the specific area is set, it is detected whether or not an inter-center layer exists in the specific area. The center-to-center layer usually refers to a region (transparent region) that is located almost in the center of the embryonic nucleus in the front-rear direction and has little scattered reflected light. As cataract progresses, it becomes difficult to detect the center layer due to turbidity. Therefore, grades can be classified based on the presence or absence of the center-to-center layer. Specifically, if there is an inter-center layer, it is determined that it belongs to the first grade group with a lower grade, and if there is no inter-center layer, it is determined that it belongs to the second grade group with a higher grade. Thereby, it is possible to provide a crystalline lens image analysis apparatus capable of objectively and automatically performing NUC classification evaluation in cataract diagnosis.

In the present invention, first luminance value calculation means for calculating the luminance values of the front region and the rear region of the specific region,
First luminance value comparison means for comparing the calculated luminance value and sample data of known grade,
When it is determined by the grade determining unit that the grade belongs to the first grade group, the grade determining unit preferably determines a final grade based on a comparison result by the first luminance value comparing unit.

  In the case of the first grade group in which the center layer is present, the luminance values of the front area and the rear area of the specific area are calculated in order to determine the final grade. As the cataract progresses, both the front and rear regions become cloudy, but when not so advanced, the front region becomes nearly transparent. Therefore, the final grade can be determined by calculating the luminance values before and after the above and comparing with the sample data with a known grade.

In the present invention, a second luminance value calculation means for calculating a luminance value in the specific area;
A second luminance value comparing means for comparing the calculated luminance value with sample data of a known grade;
When it is determined by the grade determining means that the member belongs to the second grade group, it is preferable that the grade determining means determines a final grade based on a comparison result by the second luminance value comparing means.

  In the case of the second grade group in which the center layer does not exist, the luminance value of the specific region is further calculated in order to determine the final grade. This is because as the cataract progresses, the specific area becomes totally turbid. Therefore, the final grade can be determined by calculating the luminance value in the specific region and comparing it with sample data with a known grade.

  In the case of the second grade group, it is preferable that the specific region is set at a rear portion of the lens region. As nuclear cataract progresses, turbidity in the posterior region of the lens region in particular progresses. Therefore, in order to determine a more accurate grade, it is preferable to reset the specific area at the rear of the crystalline lens area.

In the present invention, comprising a transparency determination means for determining whether or not the luminance value in the specific region is a transparency of a predetermined level or less,
Preferably, the grade determining means determines that the luminance value in the specific region is the lowest grade among the NUC classification grades regardless of the presence or absence of an inter-center layer when the luminance value in the specific region is determined to be equal to or lower than the predetermined level.

  When the luminance value in the specific region is a transparency level equal to or lower than a predetermined level, it is possible to determine that the degree of turbidity is low and the cataract does not progress so much. Therefore, it can be determined that it is the lowest grade among NUC classification grades regardless of the presence or absence of the center-to-center layer. Thereby, the processing time until determination can be shortened.

In the present invention, it comprises a lens extraction means for extracting a lens region by image processing from the acquired optical cutting plane image data,
Preferably, the specific area setting means sets the specific area inside the lens area.

  According to this configuration, the crystalline lens region is extracted by image processing. In order to determine the grade of the NUC classification, a specific region is set inside the lens region. Therefore, since the specific region is set after first extracting the lens region, it is possible to ensure the setting of the specific region and increase the accuracy.

  In the present invention, it is preferable that the specific area setting means calculates a correlation function value while scanning a preset shape, and sets a position having the highest degree of correlation as the specific area.

  In setting the specific area, the shape is set in advance. This set shape is scanned. While scanning, a correlation function value is calculated, and a position where the calculated degree of correlation is the highest is set as a specific region. As a method for obtaining the correlation function value, a suitable method for setting the specific region may be determined in advance, and various calculation methods are conceivable. As a result, the specific area can be automatically set at an appropriate position.

<Crystal image analysis program>
The crystalline lens image analyzing apparatus according to the present invention can be realized by installing a crystalline lens image analyzing program having the following features in a computer.

That is, in order to solve the above-mentioned problem, the lens image analysis program according to the present invention is:
Obtaining optical section image data of the eye to be examined obtained by photographing the eye; and
From the acquired optical cutting plane image data, including at least partially the embryonic nucleus region, or setting a specific region set in the vicinity of the embryonic nucleus region inside the lens region;
Detecting whether a center-to-center layer exists in the set specific region;
When the center layer exists, it is determined that it belongs to the first grade group with a low NUC classification grade for nuclear cataract evaluation. When the center layer does not exist, the second grade group with a high NUC classification grade And determining that the computer belongs to the computer.

In the present invention, calculating the luminance values of the front region and the rear region of the specific region,
Comparing the calculated luminance value with sample data of known grade, and
When it is determined by the determining step that the image belongs to the first grade group, it is preferable that the grade determining step determines a final grade based on a comparison result by the comparing step.

In the present invention, calculating a luminance value in the specific area;
Comparing the calculated luminance value with sample data of known grade, and
When it is determined by the determining step that the second grade group belongs, it is preferable that the grade determining step determines a final grade based on a comparison result of the comparing step.

  In the present invention, it is preferable that the specific region is set at a rear portion of the lens region.

In the present invention, comprising the step of determining whether or not the luminance value in the specific area is a transparency below a predetermined level,
Preferably, in the step of determining the grade, when the luminance value in the specific region is determined to be equal to or lower than the predetermined level, it is determined that the grade is the lowest among the NUC classification grades regardless of the presence or absence of the center layer. .

  Furthermore, it is preferable that the lens region is extracted by image processing from the acquired optical cutting plane image data, and the specific region setting step sets the specific region inside the lens region.

  In the specific area setting step, it is preferable that the correlation function value is calculated while scanning a preset shape, and a position having the highest degree of correlation is set as the specific area.

Block diagram showing functions of an anterior ocular segment observation apparatus including a crystalline lens image analysis apparatus Block diagram explaining the functions of the computing unit The figure which shows the horizontal sectional view of the eyeball partially Flow chart showing the procedure for determining the grade Flow chart showing the procedure for determining the grade Figure showing the optical cut surface image on the monitor Diagram showing the setting of a specific area within the crystalline lens area The figure which shows the setting of the line segment orthogonal to the major axis of the ellipse representing the specific area Diagram showing brightness distribution data Flow chart showing the procedure for determining a specific area The figure which shows the procedure for confirming the existence of the center layer The figure which shows the detailed conditions for judging the presence or absence of a center layer Diagram showing the front and rear regions Diagram showing sample data of grade 0 and 1 The figure which shows the state which set the specific area to the rear part of the crystalline lens area Diagram showing sample data of grades 2 and 3 The figure explaining another embodiment which sets up a specific field The figure explaining another embodiment which sets a specific area | region Diagram showing the contents of NUC classification

  A preferred embodiment of a crystalline lens image analyzing apparatus according to the present invention will be described with reference to the drawings. FIG. 1 is a block diagram illustrating functions of an anterior ocular segment observation apparatus including a crystalline lens image analysis apparatus.

<Configuration of anterior ocular segment observation device>
As shown in FIG. 1, an anterior ocular segment observation device 100 of the present embodiment includes an optical device 1 and an analysis device 2 that functions as a lens image analysis device. In FIG. 1, only the configuration relating to image analysis is shown for the optical device 1. The analyzer 2 is provided with a function for automatically determining the grade of NUC classification.

  The optical device 1 includes a longitudinal cross-sectional image sensor 10 that captures an image of a longitudinal optical cut surface of an eye to be examined (hereinafter referred to as a longitudinal optical cut surface image) and outputs the image, and a lateral optical cut surface of the eye to be examined. And a cross-sectional image pickup device 11 that picks up an image (hereinafter referred to as a horizontal optical cut surface image) and outputs the image. In addition, the optical device 1 includes various optical elements and the like. As a specific configuration of the optical device 1, the configuration of the optical device disclosed in Japanese Patent Application Laid-Open No. 2009-56149 by the present applicant can be used as it is.

  The analysis device 2 includes a computing unit 20, a monitor 21, and an input device 22. The crystalline lens analysis device 2 is composed of, for example, a personal computer and its peripheral devices, the computing unit 20 is composed of a personal computer as a main body, the monitor 21 is composed of an image display device, and the input device 22 is composed of a mouse, a keyboard, and the like. It consists of operation tools for data entry.

  The computing unit 20 includes, for example, a CPU and a main memory, and a predetermined program including a lens analysis program described later is stored in the main memory. Then, the CPU reads out and executes this program, so that necessary processing and control are performed. Specifically, the computing unit 20 receives a vertical optical cut surface image and a horizontal optical cut surface image from the vertical cross-sectional image pickup device 10 and the cross-sectional image pickup device 11, respectively, and based on these, the cornea and the crystalline lens of the eye to be examined. Measure the radius of curvature of the front and back surfaces. Further, as described later, the lens image is analyzed to determine the grade of NUC classification. Further, based on information input from the input device 22 and image data input from the optical device 1, necessary display is performed on the monitor 21 and the overall operation of the optical device 1 is controlled.

<Functional configuration of computing unit>
Next, the function of the computing unit 20 will be described with reference to the block diagram of FIG. The analysis device 2 also has an analysis function other than the lens image analysis. FIG. 2 shows only functions related to the lens image analysis.

  The image acquisition unit 200 acquires optical cut surface image data of the eye to be inspected that is taken by the optical device 1. The optical cut surface image here is an image photographed by the longitudinal cross-section image sensor 10. The crystalline lens extracting means 201 extracts a crystalline region from the obtained optical cut surface image data using an image processing technique. A specific example of extracting the lens region will be described later, but the technology itself for extracting the lens region is known. Although the cross-sectional image sensor 11 may be used instead of the vertical cross-section image sensor 10, the vertical cross-section is less susceptible to wrinkles and analysis is easier.

  The specific area setting unit 202 sets a specific area inside the extracted lens area. The specific region has an area smaller than that of the lens region and includes at least a part of the embryonic nucleus region or is set in the vicinity of the embryonic nucleus region.

  Here, the embryonic nucleus will be described. FIG. 3 is a diagram partially showing a vertical sectional view of the eyeball. The cornea 30, the pupil 31, the lens 32, and the vitreous body 33 are positioned in order from the front of the eyeball. The crystalline lens 32 has a substantially spheroid shape (convex lens shape), and the embryonic nucleus 32a is located at substantially the center thereof. A region at least partially including the embryonic nucleus 32a or the vicinity of the embryonic nucleus 32a is set as a specific region. By looking at the turbidity within this specific region, the degree of progression of nuclear cataract (NUC classification grade) can be determined.

  As a standard for determining the degree of progression of nuclear cataract, there is a grade classification according to the NUC classification of WHO. FIG. 18 is a diagram showing the contents of each grade. This classification focuses on the gradual expansion of turbidity as nuclear cataract progresses. There are four grades, 0, 1, 2, and 3. The larger the value, the more advanced the cataract.

  In order to determine the grade, three types of standard images are prepared, and the determination by the doctor is performed by comparing the optical cut surface image of the subject with NUC Standard 1, NUC Standard 2, and NUC Standard 3. That is, if it is less than NUC standard 1, it is judged as grade 0, if it is NUC standard 1 or more and less than 2, it is judged as grade 1, if it is NUC standard 2 or more and less than 3, it is judged as grade 2, and if it is NUC standard 3 or more, it is judged as grade 3. If it cannot be determined, it is determined as grade 9.

  Incidentally, NUC Standard 1 is sufficient for consideration as a case and is a clinically significant nuclear cataract. The front and back nuclei are significantly milky white than usual, but the intercenter layer is at a level that can be recognized as a whole. NUC standard 2 is a state in which nuclear cataract has progressed considerably. The nucleus region is uniformly opaque, and the central layer cannot be clearly recognized. The rear region of the nucleus region is opaque and the red reflections gradually darken. NUC Standard 3 is a serious enough condition to consider surgery. The nucleus region exhibits deep opacity with turbidity that uniformly spreads to its edge, and the features of the nucleus are at a level that can be partially recognized. The red reflection becomes dull.

  The present invention is not a visual determination based on the standard image, but does not use the standard image, and the final grade is calculated by the calculator 20 by the calculation process. Since the NUC classification grade has four stages, grades 0 and 1 having a low NUC classification grade are defined as a first grade group, and grades 2 and 3 having a high NUC classification grade are defined as a second grade group.

  As the grade value increases, the turbidity gradually increases. In the present invention, a specific area is set, the turbidity in the area is examined, and the grade is automatically determined.

  The center-to-center layer detection unit 203 has a function of detecting whether or not the center-to-center layer exists in the specific area set by the specific area setting unit 202.

  The grade determination means 204 has a function of finally determining the NUC classification grade. Further, based on the detection result of the center-layer detection means 203, when the center-layer is present, the grade is determined to belong to the first grade group (grade 0, 1), and when the center-layer is not present The grade is determined to belong to the second grade group (grades 2 and 3).

  The first luminance value calculating means 205 calculates the luminance values of the front area and the rear area of the specific area, respectively, when the grade determining means 204 determines that it belongs to the first grade group.

  The first luminance value comparison unit 206 compares the luminance value calculated by the first luminance value calculation unit 205 with sample data whose grade is known, and the grade determination unit 204 determines the final grade. That is, it is finally determined whether the grade is 0 or 1.

  The second luminance value calculating unit 207 calculates the luminance value of the specific area when the grade determining unit 204 determines that the second luminance group belongs to the second grade group.

  The second luminance value comparison unit 208 compares the luminance value calculated by the second luminance value calculation unit 207 with sample data whose grade is known, and the grade determination unit 204 determines the final grade. That is, it is finally judged whether it is grade 2 or 3.

  The transparency determination means 209 determines whether or not the luminance value in the specific area is a transparency that is a predetermined level or less. When it is determined that the luminance value is equal to or lower than the predetermined level, the grade determination unit 204 finally determines that it is the lowest grade, that is, grade 0.

  The data storage means 210 stores a large number of data with known NUC classification grades. For each grade of grades 0 to 4, past data is stored in various formats. It is preferable to store a large number of data, not just one for each grade. Thereby, the determination accuracy can be increased. The data storage unit 210 can use a mass storage device such as a hard disk or an external storage device.

<Grade judgment procedure>
Next, the procedure for determining the grade will be described with reference to the flowcharts of FIGS. 4A and 4B. FIG. 5 shows a state where an optical cut surface image photographed by the optical device 1 is acquired (S1) and displayed on the monitor 21. First, the cornea 30 is extracted by image processing (S2). As can be seen in FIG. 5, the cornea 30 is positioned in front of the eye, and images before and after the cornea 30 are dark. Focusing on such characteristics, the front line 30a and the back line 30b of the cornea 30 are extracted by image processing. Thereby, the cornea 30 can be extracted.

  Next, the lens 32 is extracted (S3). It is known that the lens 32 is located at the rear of the cornea 30. Further, since the crystalline lens 32 has a greenish tone and the periphery of the crystalline lens 32 is dark, first, the front boundary 32b of the crystalline lens 32 is extracted by image processing. Next, the rear boundary 32c is similarly extracted.

  Since the entire shape of the crystalline lens 32 can be approximated to an ellipse, when the front boundary 32b and the rear boundary 32c are extracted as described above, the entire crystalline lens can be approximated to an elliptical shape, which is referred to as a crystalline region 40. . FIG. 6 shows a crystalline lens region 40 approximated by an ellipse. Thereby, the position of the crystalline lens 32 is determined. When the crystalline region 40 is determined, it is next necessary to set the specific region 41. First, the shape of the specific region 41 is set (S5). The shape is an ellipse that is similar to the crystalline lens region 40, and is determined by reduction at a predetermined ratio. For example, the length of the long axis is reduced to 40 to 60%. Preferably, it is set to be around 50%.

  Next, it is necessary to set the position of the specific area 41 in the crystalline lens area 40. For this purpose, the elliptical scan process is executed (S6). As shown in FIG. 6, this is a process of calculating a correlation function value while scanning an elliptical shape from the foremost part (indicated by 41a) to the last part (indicated by 41b) in the crystalline lens region. Since the specific area 41 is estimated to be at the center position in the left-right direction of the crystalline lens area 40, it is sufficient that the scanning direction is from the center frontmost part (41a) to the center rearmost part (41b). This process will be described in detail below with reference to the flowchart of FIG.

<Extraction of specific area>
First, as shown in FIG. 7, a line segment 411 orthogonal to the elliptical long axis 410 representing the specific region 41 (in the left-right direction) is set. The line segment 411 is divided into an upper part 411a and a lower part 411b of the long axis 410. The brightness distribution in the line segment 411 is shown, for example, as shown in FIG. However, this distribution is a G (green) component in the color image data. The R (red) component tends to be biased to the rear of the lens region 40 and the B (blue) component tends to be biased to the front, whereas the G component has no bias in the front-rear direction. Therefore, image processing for setting the specific area 41 is performed based on the G component.

  Next, an average brightness value is obtained for the upper portion 411a, and an average brightness value is also obtained for the lower portion 411b (S20). Next, when the brightness distribution value is subtracted by the obtained average value, the DC component is removed as shown in FIG. This is performed for each of the upper part 411a and the lower part 411b. Further, a standard deviation σ is obtained for the brightness distribution data shown in FIG. 8A (S21). The distribution data shown in FIG. 8B is divided by the standard deviation σ. Thereby, the range of fluctuation is normalized to a value with a standard deviation of 1 (S22). This is also performed for each of the upper part 411a and the lower part 411b.

  As described above, normalized brightness distribution data is obtained for each of the upper part 411a and the lower part 411b. Then, a correlation function value is obtained between the function representing the distribution data of the upper portion 411a and the function representing the distribution data of the lower portion 411b (S23). This is a value representing how similar the upper and lower distribution data is.

  The above correlation function values are scanned along the elliptical long axis 410, for example, from left to right (S24), the correlation function values are calculated for each line segment 411, and all of them are added to obtain an added correlation value. (S25). This added correlation value is taken as the degree of correlation of the specific area 41 at that position.

  As described above, the elliptical shape is scanned from the center frontmost part (41a) to the center rearmost part (41b) (S26), and the degree of correlation at each position is calculated. The degree of correlation at each position is indicated by a graph [A] on the left side of FIG. The position having the highest degree of correlation is set as the specific area 41 (S7, S27). As described above, the specific area 41 can be automatically set.

  Returning to the flowchart of FIG. 4, when the specific area 41 is set, next, the luminance value in the specific area is calculated (S8). In the calculation of the luminance value, the luminance value of each pixel constituting the specific area 41 is obtained and the average value is calculated. If the average value of the calculated luminance values is equal to or lower than the predetermined level, the grade determination unit 204 determines that the final grade is grade 0 (S40). In a normal state in which the nuclear cataract is not progressing, the transparency of the specific region 41 (embryonic nucleus region) is high, so that it can be determined that it is the lowest grade 0 without performing the detection process of the center layer described later. is there. About a predetermined level, it can set suitably.

  If the level is not lower than the predetermined level, the center-center layer detection process is performed (S11). According to the NUC classification, there is an inter-center layer in grades 0 and 1 (first grade group) and not in grades 2 and 3 (second grade group).

<Center layer>
The presence / absence detection of the center layer is performed by the following procedure. FIG. 10 shows a state in which the specific region 41 is sliced along a line segment 412 parallel to the long axis 410. The center-to-center layer usually refers to a region (transparent region) that is located approximately in the center of the embryonic nucleus 32a in the front-rear direction and has little scattered reflected light.

  There are as many sliced line segments 412 as the number of pixels constituting the short axis 413 of the specific area 41. For each line segment 412, the average value is calculated only for the G (green) component. A graph of the average value is shown on the right side of FIG. If the center-to-center layer exists, the portion becomes dark, so that the graph [B] has a dent on the left side as shown in FIG. The presence or absence of the center layer can be determined by the presence or absence of this dent. The calculation may be performed using R (red) or B (blue), which are other colors, or a monochrome luminance value instead of the G component.

  FIG. 11 shows only the graph [B] in FIG. First, a local minimum point of a curve and a concave area sandwiching the local minimum point are extracted from the graph. There are many local minimum points in the graph. For example, in FIG. 11, a plurality of local minimum points can be seen as indicated by P1, P2, P3, and the like. Then, the recessed areas LP1, LP2, LP3 sandwiching these minimum points P1, P2, P3 can be detected. When these plural minimum points are detected, the one with the longest concave area is determined as a subsequent calculation target. In this case, since the length of the concave area LP1 is the longest, the concave area LP1 including the minimum point P1 is determined as a target. Next, criteria for determining whether or not the recessed area LP1 corresponds to the center-to-center layer is set as follows.

That is,
(A) The length of the concave area is 1/5 or more of the entire length of the curve.
(B) The inclination to the adjacent maximum point is greater than or equal to the threshold value.
(C) A concave area is applied to the central portion 1/3.
(D) The depth of the concave area is greater than or equal to a threshold value.

  Here, the entire curve length is represented by Lt. That is, the length of the short axis of the ellipse that constitutes the specific region 41 corresponds to the entire length of the curve. If LP1 is Lt × 1/5 or more, the condition (A) is satisfied. This is because the inter-center layer is considered to have a predetermined length or more.

  Since the local maximum points are represented by Q1, Q2, and Q3 as shown in FIG. 11, the adjacent local maximum points are Q1 and Q3. The inclination of the line segment connecting the minimum point P1 and the maximum point Q1 or Q3 is calculated, and if this inclination is equal to or greater than the threshold value, the condition (B) is satisfied. As shown in FIG. 11B, the maximum point Q10 is calculated by calculating a curve with a large period greater than a predetermined value, ignoring the irregularities with a small period as described above. Alternatively, the tilt may be calculated using Q11. In this figure, a line segment connecting the minimum point and the maximum point is indicated by a broken line. If the slope is smaller than the threshold value, the difference in luminance value is not so large.

  Next, the region of the central portion 1/3 is indicated by Lc in FIG. The total length Lt is divided into three equal parts, and the middle part thereof is represented as Lc. If the concave area indicated by LP1 covers the area of Lc (if it overlaps), the condition (C) is satisfied. It can be set as appropriate as to how long it takes to satisfy the condition (C). For example, it may be determined that the condition is satisfied if it takes even a little, or it may be determined that the condition is not satisfied if there is no overlap of a certain length or more.

  The depth of the recessed area is indicated by h in FIG. If the depth h is greater than or equal to the threshold value, the condition (D) is satisfied. About the magnitude | size of a threshold value, it can determine suitably. This is because the inter-center layer is considered to have higher transparency than the other areas, and therefore, if the concave area is the inter-center layer, the depth h is considered to be greater than or equal to a predetermined value.

  When all the conditions (A) to (D) set as described above are satisfied, it can be determined that the center-to-center layer exists. However, the present invention is not limited to this, and further conditions may be added. Alternatively, it may be determined that an inter-center layer exists when at least one of conditions (A) to (D) is satisfied.

  If it is determined in step S11 that the center layer is present, the grade is determined to belong to the first grade group, and if it is determined that the center layer is not present, the grade belongs to the second grade group. It is determined (S12, S30). When it is determined that it belongs to the first grade group, it is further necessary to finally determine whether the grade is 0 or 1.

<Grade 0, 1 judgment>
The difference between grades 0 and 1 is represented by the turbidity ratio of the front region A1 and the rear region A2 of the specific region 41 (embryonic nucleus region) as shown in FIG. The specific area 41 can be divided into two equal parts by the major axis of the ellipse constituting the specific area 41 and divided into the front area A1 and the rear area A2. In grade 0, the rear area A2 is cloudy, but the front area A1 is transparent and the brightness is dark. In grade 1, the turbidity of the front area A1 is close to that of the rear area A2. Therefore, the brightness values of the front area A1 and the rear area A2 of the specific area 41 are calculated (S13). The luminance value can be obtained as an average value of the luminance values of the pixels constituting each region.

  The calculated luminance value (average value) is compared with known sample data (S14). FIG. 13 shows sample data stored in the data storage unit 210, and a data group analyzed in the past is two-dimensionally displayed (luminance value map). Grade 0 is plotted with ●, and Grade 1 is plotted with ○. The data is plotted in two-dimensional coordinates with the x-axis representing the rear luminance value and the y-axis representing the front luminance value. This sample data is created based on black and white luminance value data, but the luminance value of any one of RGB color information may be used.

  In FIG. 13, the actually calculated luminance value is plotted with X. Whether this brightness value is grade 0 or 1 is determined by determining the distance between the known grade 0 group data and X, the distance between the known grade 1 group data and X, and the closest grade as the final grade. Determine (S15). As a calculation method of the distance, it can be calculated based on the Mahalanobis method.

  The method for determining whether the grade is 0 or 1 is not limited to the above, and various modifications can be considered. The method shown in FIG. 13 is a calculation method based on the ratio of the luminance values of the front area A1 and the rear area A2, but may be determined based on the difference in luminance values instead of the ratio. For example, the difference between the luminance value (average value) of the front area A1 and the luminance value (average value) of the rear area A2 is calculated. can do. In this case, known sample data is not required.

  As yet another method, a method of using pixel color information is also conceivable. For example, the RGB luminance value information of the front area A1 (becomes three-dimensional information) and the RGB luminance value information of the rear area A2 are plotted on a three-dimensional space. In this case, the average value is also used as the luminance value. Then, taking the vector of the plot point of the front region and the plot point of the rear region and paying attention to the fact that the length and direction of this vector are different between grade 0 and grade 1, the final grade can be determined. Good.

<Grade 2 and 3 judgment>
In step S30, when it is determined that it belongs to the second grade group, it is necessary to finally determine whether it is grade 2 or 3.

  The difference between grades 2 and 3 can be determined by the spread of turbidity in the specific region 41. In grade 3, turbidity also extends to the rear region of the lens region 40. Therefore, grade 2 or 3 is determined based on the turbidity in the rear region of the lens region. Therefore, as shown in FIG. 14, the specific area 41 is set at the rear of the crystalline lens area (S31). Specifically, it is set at a position indicated by 41b in FIG. 6, and the size and shape thereof may be the same.

  Next, the luminance value (average value) in the set specific area 41 is calculated (S32). The luminance value is obtained as an average value of the luminance values of each pixel in the specific area. The calculated luminance value is compared with the sample data stored in the data storage unit 210 (S33). This sample data is shown in FIG.

  Sample data is displayed as a histogram, the x-axis is the luminance value, and the y-axis represents the frequency. Here, a grade 2 histogram and a grade 3 histogram are shown. Here, the calculated luminance value is plotted with X. The distance between the plot point and the grade 2 group data and the distance between the plot point and the grade 3 group point are calculated, and the closer grade is set as the final grade (S34). In addition, about the calculation method of distance, it can calculate by the above-mentioned Mahalanobis method.

  In the above description, the calculation is performed based on the black and white luminance values. However, similarly to the grades 0 and 1, the calculation may be performed based on the color RGB luminance values.

  The determination method of grade 2 or 3 is not limited to the above, and various modifications can be considered. The difference between grades 2 and 3 is a difference in spread in the turbidity in the embryonic nucleus, and can be determined based on this. For example, an appropriate region (corresponding to the specific region described above) larger than the embryonic nucleus region is set within the range of the lens region (see FIG. 16). When the center of the lens region is SO and the distance between each pixel (indicated by SA) in the region and the center SO is D, the moment of the turbidity distribution (SA brightness value × D) is obtained. For example, the moment is calculated for all the pixels in the region, and the average value or sum of the moments is calculated. It is considered that the turbidity progresses to the periphery as the value increases. Therefore, it is possible to determine whether it is grade 2 or grade 3 by appropriately setting a threshold value.

  As described above, grades 0 to 3 that are NUC classification grades can be automatically determined. In the present invention, the determination is made by software without using the NUC standard image. Accordingly, there is no influence of the doctor's experience or individual differences, and a uniform NUC classification can be performed.

<Another embodiment for detecting the presence or absence of the center-to-center layer>
Various methods other than the present embodiment are conceivable as a method for detecting the presence or absence of the center-to-center layer. For example, the approximate position of the center layer in the crystalline lens region (for example, a layer having a predetermined width is set at the center in the front-rear direction) is determined. The average value of the luminance values at this position is compared with the average value of the luminance values in the other areas. If it is less, it can be determined that there is no center-to-center layer.

<Another embodiment in which no lens extraction means is provided>
In the present embodiment, the specific area 41 is set after the crystalline lens area is extracted by providing the crystalline lens extracting means 201. However, the specific area 41 may be set without extracting the crystalline area 40. Good. For example, the following configuration is conceivable.

  Image data (pattern) of a specific area is prepared in advance, and the inside of the obtained optical cut surface image is scanned. The position with the highest degree of approximation can be set as the specific area. For example, the embryonic nucleus has a unique image pattern, such as having an inter-center layer. By scanning this pattern, an area including the embryonic nucleus can be extracted.

  Further, a method is conceivable in which a cross-sectional pattern of an optical cut surface image is stored in advance and a specific region is extracted by comparing the cross-sectional pattern with cross-sectional data of an optical cut surface image actually taken. As shown in FIG. 17, the cornea 30, the anterior chamber 34, and the crystalline lens 32 are located from the front of the eyeball, and the luminance distribution can also be estimated. Therefore, a typical luminance distribution pattern is stored in advance as a cross-sectional pattern (FIG. 17A), and this is compared with the cross-sectional data (shown by (b)) of the actually obtained optical cut surface image. At the time of comparison, considering the difference in eye size, comparison is made with (a) multiplied by a ratio (in the direction of arrow D), and a specific region is extracted from the cross-sectional pattern at the ratio when most matches. . Since the position of the specific region is determined in the cross-sectional pattern (indicated by E), the specific region can be set from this position and the ratio.

  In the present embodiment, a predetermined elliptical shape is scanned only inside the crystalline region 40, but when the crystalline region 40 is not extracted or set, a wide range from the front to the back of the optical cut surface image is set. And the region with the highest correlation function value may be set as the specific region as in the present embodiment.

<Another embodiment in the case of providing a lens extracting means>
Various other ways of setting the specific area 41 are conceivable. In the present embodiment, the degree of correlation is obtained, and the position with the highest degree of correlation is set as the specific region 41. Since the specific area 41 to be set is estimated to be substantially at the center of the crystalline lens area 40, the specific area 41 may be set to the central area of the crystalline lens area 40 as long as the rear surface is reliably reflected from the front surface of the crystalline lens.

  In the present embodiment, the specific area is set to be oval, but is not limited to this shape. In addition to an ellipse, various shapes such as a circle, an oval, a rectangle, and a square can be set. Moreover, when setting an ellipse, it does not need to be a similar shape. Depending on the shape and size of the specific region, the embryonic nucleus may be completely or partially included in the set specific region. Alternatively, there may be a case where the embryonic nucleus is partially included as in the specific area to be reset, or is outside the range of the embryonic nucleus but is set in the vicinity.

  In the present embodiment, the specific area 41 is set at the center in the left-right direction of the crystalline lens area 40, but the specific area 41 is not limited to this, and may be set at a position biased to either the left or right.

  In the present embodiment, when the specific area 41 is reset (see S31 in FIG. 4B), the specific area 41 is set at the rearmost position in the crystalline lens area 40, but the present invention is not limited to this. For example, you may make it set back only the predetermined distance of the specific area | region set initially. The size of the specific area 41 to be reset may be the same as or different from the size and shape of the specific area 41 set first.

<Another Embodiment for Obtaining Correlation Function Value>
In the present embodiment, the image processing until the specific area 41 is set is performed based on the G component (specific color data). Instead, the correlation value distribution is obtained from both the G component and the R component (which also correspond to specific color data), and the position of the specific region obtained based on the G component is inappropriate. In this case, the specific region may be reset at a position having the highest correlation value based on the R component. When the turbidity increases, the R component increases in the color tone. By setting as described above, the position accuracy of the specific region can be increased.

  Further, the correlation value may be obtained by the same procedure as in the present embodiment based on the monochrome image data (luminance data) obtained by combining the RGB components, not based on the G component.

  Alternatively, for each of the R, G, and B components, the correlation value distribution is obtained in the same procedure as in this embodiment. Then, among positions having a correlation value equal to or greater than the threshold value, positions that are not biased before and after the lens area (positions where the embryonic nucleus is presumed) are extracted as candidate positions of the specific area, From this, an appropriate position can be selected.

  In the present embodiment, even if the rear boundary 32c of the crystalline lens 32 is an unclear image, the crystalline region 40 can be set as long as the front boundary 32b can be extracted. This is because an ellipse representing the crystalline lens region 40 can be estimated from the curvature of the curve representing the front boundary 32b.

  In this embodiment, when calculating the luminance value, the calculation is performed based on the average value of the luminance values of all the pixels in the region, but the present invention is not limited to this. For example, calculation may be performed by thinning out pixels at a predetermined ratio instead of all pixels.

DESCRIPTION OF SYMBOLS 1 Optical apparatus 2 Analysis apparatus 20 Calculator 200 Image acquisition means 201 Lens extraction means 202 Specific area setting means 203 Center layer detection means 204 Grade determination means 205 First brightness value calculation means 206 First brightness value comparison means 207 Second brightness Value calculation means 208 Second luminance value comparison means 209 Transparency determination means 210 Data storage means 21 Monitor 22 Input device 30 Cornea 32 Lens 32a Embryonic nucleus 40 Lens area 41 Specific area A1 Front area A2 Rear area

Claims (6)

Image acquisition means for acquiring optical cut surface image data of the eye to be examined obtained by photographing the eye to be examined;
From the acquired optical cut plane image data, a specific region setting means for setting a specific region that is at least partly included in the embryonic nucleus region or set in the vicinity of the embryonic nucleus region, inside the lens region,
A center-to-center layer detecting means for detecting whether a center-to-center layer exists in the set specific region;
When the center layer exists, it is determined that it belongs to the first grade group with a low NUC classification grade for nuclear cataract evaluation. When the center layer does not exist, the second grade group with a high NUC classification grade Grade judging means for judging that it belongs to ,
First luminance value calculation means for calculating the luminance values of the front area and the rear area of the specific area,
First luminance value comparison means for comparing the calculated luminance value and sample data of known grade,
When it is determined by the grade determining means that the grade belongs to the first grade group, the grade determining means determines a final grade based on a comparison result by the first luminance value comparing means. Lens image analysis device. Image acquisition means for acquiring optical cut surface image data of the eye to be examined obtained by photographing the eye to be examined;
From the acquired optical cut plane image data, a specific region setting means for setting a specific region that is at least partly included in the embryonic nucleus region or set in the vicinity of the embryonic nucleus region, inside the lens region,
A center-to-center layer detecting means for detecting whether a center-to-center layer exists in the set specific region;
When the center layer exists, it is determined that it belongs to the first grade group with a low NUC classification grade for nuclear cataract evaluation. When the center layer does not exist, the second grade group with a high NUC classification grade Grade judging means for judging that it belongs to ,
A second luminance value calculating means for calculating a luminance value in the specific area;
A second luminance value comparing means for comparing the calculated luminance value with sample data of a known grade;
When it is determined by the grade determining means that the grade belongs to the second grade group, the grade determining means determines a final grade based on a comparison result by the second luminance value comparing means. Lens image analysis device. The lens image analyzing apparatus according to claim 2, wherein the specific region is set at a rear portion of the lens region. Transparency determination means for determining whether or not the luminance value in the specific area is a transparency of a predetermined level or less,
The grade determining means determines that the luminance value in the specific area is the lowest grade among the NUC classification grades regardless of the presence or absence of an inter-center layer when the luminance value in the specific area is determined to be equal to or lower than the predetermined level. The crystalline lens image analysis apparatus according to any one of claims 1 to 3 . A lens extracting means for extracting a lens region by image processing from the acquired optical cutting plane image data;
5. The lens image analysis apparatus according to claim 1 , wherein the specific area setting unit sets the specific area inside the lens area. 6. The specific area setting means, while scanning a predetermined shape, to calculate the correlation function value, any one of the preceding claims, characterized in that setting the highest correlation position as the specific region The crystalline lens image analysis apparatus described in the item.

Images