JP2010158279A - Ocular-fundus image analyzing system and ocular-fundus image analyzing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an ocular-fundus image analyzing system and ocular-fundus image analyzing program, capable of providing useful information for diagnosis by analyzing ocular-fundus image data and discriminating symptoms of glaucoma based on a pixel value profile. <P>SOLUTION: An analysis computer 2 in the analyzing system 1 includes: an ocular-fundus image data input receiving means 11 for receiving the input of ocular-fundus image data 3 of a subject; an analysis target extracting means 12 for extracting an analysis target region A including an optic papilla region D; a blood vessel erasing means 13 for specifying and erasing a blood-vessel region B; a converting means 16 for image-converting blood-vessel erased image data 5; a contour point forming means 17 for forming contour points OP between the optic papilla region D and a retina region R; a profile creating means 18 for creating the pixel value profile 7; and a glaucoma discriminating means 19 for comparing the pixel value profile 7 with the profile in normal time and discriminating symptoms of the glaucoma. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a fundus image analysis system and a fundus image analysis program. In particular, the present invention uses fundus image data obtained by photographing the fundus of a subject, and provides doctors with useful information on diagnosis of ophthalmic diseases such as glaucoma. The present invention relates to a fundus image analysis system and a fundus image analysis program that can be provided to support the diagnosis.

  Conventionally, the eyeball is irradiated with light from the outside, and the state of the fundus behind the eyeball is observed through the eyeball. From the fundus image taken by irradiating light, it is possible to obtain information such as the size of the optic nerve head area and the optic nerve head depression area (details will be described later), their ratio, and the crossover state of the blood vessels. It is useful information that allows doctors to diagnose ophthalmic diseases such as glaucoma and other diseases such as diabetes. This diagnostic method using fundus images is a non-invasive method that does not impose a physical burden on subjects (patients), only by irradiating light from outside and taking images. It is a diagnostic technique that is easy to apply to the elderly. Furthermore, a device (fundus camera) that acquires a fundus image is composed of a relatively simple device, and is one of inspection devices that are widely used in current medical institutions.

  Here, a specific example (diagnosis of glaucoma) based on the fundus image will be described. The fundus image captured by the fundus camera generally has a yellowish reddish brown color in the entire fundus, and a circular shape of 1.2 mm to 1.7 mm or approximately 15 degrees inward (in other words, the nasal direction) of the subject's line of sight A region of the optic disc (Disc) indicated by pixels having a higher brightness than the egg-shaped fundus region is observed, and on the inner side, a small circular shape observed in a brighter and whitish region than the peripheral portion of the optic disc An area of optic disc depression (Cup) is observed. Here, a region between the optic nerve head and the optic nerve head depression is called a rim.

  As glaucoma progresses, there is a symptom in which the optic disc is compressed and dented due to an increase in intraocular pressure. As a result, since the width of the rim tends to narrow, as one of the guidelines for suspected glaucoma symptoms, the diameter (or radius) of the region of the optic disc (D) and the optic disc recess (C) described above is used. Ratio values are used. That is, since each is configured to have a substantially circular shape, the value of the diameter or radius ratio (= Cup / Disc ratio: hereinafter referred to as “C / D ratio”) of the optic disc region and the optic disc recess region is A rough judgment can be made based on whether or not a certain standard is exceeded. In the case of glaucoma, the C / D ratio tends to increase because the optic disc recessed area tends to be enlarged. Generally, when the C / D ratio is 0.6 or more, the optic disc recessed area Hypertrophy is observed and the symptoms of glaucoma are suspected.

  On the other hand, when the C / D ratio is a value smaller than 0.6, it is determined that the possibility of glaucoma is low. Even when the difference is less than 0.6, glaucoma may be similarly suspected when the difference in C / D ratio between the left and right fundus images is 0.2 or more. In the diagnosis of glaucoma, the determination based on the C / D ratio is not absolute, and is comprehensively determined based on other test results and doctor's findings. Here, examples of other factors with high possibility of glaucoma symptoms include 1) the intraocular pressure value always exceeds 25 mmHg, 2) there are symptoms due to high intraocular pressure such as “the eye feels heavy” 3) The optic disc findings change with progress, 4) the retinal nerve fiber layer defect expands with progress, and the like. That is, the value of the C / D ratio is used as one of useful information for diagnosing glaucoma symptoms, and glaucoma is not directly determined by such values.

  Here, when the C / D ratio is obtained, a ruler is generally applied to the developed fundus image, and the diameters of the optic disc area and the optic disc depression area are directly measured. In order to simplify this measurement work, a dedicated scale for glaucoma determination (for example, see Patent Document 1) may be used. However, the use of these rulers and dedicated scales tends to cause measurement errors due to differences in measurers, and it has been difficult to measure them as reliable and accurate data.

  In order to solve the above-mentioned problems, the applicants of the present application specify the optic disc region and the optic disc depression region by using a difference in pixel values from the surroundings using an image analysis technique, and for each specified region, A technique for accurately calculating the C / D ratio by applying an approximate circle when approximated by a perfect circle has been developed and disclosed (see Patent Document 2). This makes it possible to automate the calculation and determination of the C / D ratio. In particular, when diagnosing the fundus images of a large number of people such as a health checkup, useful information for diagnosis is provided to the doctor. The time required for diagnosis of glaucoma can be dramatically reduced.

  However, the above-described diagnostic method may cause the following problems. In other words, using an image analysis technique to automate the extraction of each region and the specification of an approximate circle certainly facilitates the calculation of the C / D ratio, and can simplify the work of a doctor or the like. did it. However, the optic disc and the optic disc concavity region are not necessarily perfect circles, and there are many cases in which an elliptical shape or a part of the optic disc protrudes into distortion. Therefore, since the approximate circle is specified on the assumption that each region has a perfect circle shape, the accuracy of the obtained C / D ratio may not be high. In particular, when glaucoma symptoms progress and the optic disc depression area deviates significantly from the perfect circle, it is difficult to identify a matching approximate circle and cannot provide useful information to the doctor. There was a case.

  In view of the above circumstances, the present invention analyzes the fundus image data obtained by photographing the fundus using image analysis technology, specifies the optic disc region and the optic disc depression region, and calculates the C / D ratio. A fundus image analysis system and a fundus image analysis program capable of determining the possibility of glaucoma symptoms based on a pixel value profile, providing useful information to a doctor, and supporting diagnosis The issue is to provide.

  In order to solve the above-described problems, the fundus image analysis system of the present invention provides a doctor by acquiring fundus image data obtained by photographing the fundus of a subject and analyzing the fundus image data using an analysis computer. A fundus image analysis system capable of providing information for assisting diagnosis of glaucoma, wherein the analysis computer receives fundus image data input of the subject obtained using a fundus camera A receiving unit; an analysis target extracting unit that identifies an analysis target region including an optic disc region using a difference in pixel value from the received fundus image data; and extracting the target region data; and the extracted target region data From this, the blood vessel region is identified using the difference in pixel value of the green component, and the pixel value of the identified blood vessel region is interpolated from the surrounding retinal region. Blood vessel erasure image creation means for creating blood vessel erasure image data, luminance value conversion processing means for performing processing for converting RGB values into luminance values for the created blood vessel erasure image data, and converted luminance value conversion image On the other hand, conversion processing means for increasing the luminance value difference between the optic nerve head region and the retinal region of the blood vessel erasure image data, and density gradation conversion means for performing processing for converting density gradation, Contour point forming means for detecting a boundary between the optic papilla region and the retinal region from the converted image data processed by the conversion processing unit and forming a contour point of the optic papilla region, and the formed contour point A profile that creates a pixel value profile in which a baseline for longitudinally traversing the optic disc area is plotted and the pixel values of the blue component on the baseline are plotted. And yl creating means is composed mainly comprises a "glaucoma identifying means for identifying the symptoms of glaucoma in accordance with the left and right skirt length of the pixel value profile created.

  Here, the fundus image data is image information obtained by photographing the state of the fundus of the subject using a fundus camera, and electronic data obtained by photographing the state of the fundus using a digital camera or the like is used as it is. Alternatively, a fundus image made of a silver salt photograph may be scanned and digitized. The fundus image data input accepting unit is a reading unit for analyzing the fundus image data photographed or acquired by the above-described method using an analysis computer, and is stored in a CD-R, HDD, USB memory or the like. What is stored in advance on a medium may be read using a reading device, or transmitted fundus image data may be received via a communication network and received.

  Furthermore, the analysis target extracting means specifies and extracts an analysis target region including a region of the optic disc that is to be analyzed and a optic disc concavity existing inside the optic disc. Here, since the optic disc area and the optic disc depression area are indicated by pixels having higher brightness than the surrounding area (retinal area), a known image analysis technique (for example, binarization processing method) is used. And it becomes possible to specify easily the analysis object area | region containing the said area | region using the difference in a pixel value.

  On the other hand, the blood vessel erasure image creating means removes a blood vessel region included in an analysis target region extracted as target region data. Here, since a plurality of blood vessels are running in the optic nerve head region or the like, the blood vessel hides the vicinity of the boundary between the optic nerve head region and the retinal region in forming a contour point or creating a pixel value profile, which will be described later, May affect profile values. Therefore, the blood vessel region displayed in a darker color than the optic disc region and the optic disc recessed region is identified using a difference in pixel value, in particular, using the difference in the pixel value of the green component, At the same time as removal, blood vessel erasure image data is created by interpolating the removed erasure region from the surrounding retinal region. As a result, the region where the blood vessel originally exists is displayed with pixels that approximate the retinal region.

  Furthermore, the conversion processing means includes two means: a luminance value conversion processing means for converting to a luminance value and a density gradation conversion means for performing density gradation conversion. Thereby, the brightness value difference between the optic disc region and the optic disc depression region and the other region (retinal region) can be increased, and the distinction between them can be made clearer.

  Further, the contour point forming means detects the boundary of the entire circumference of the optic nerve head region using the difference in pixel value from the converted image data. That is, since the optic papilla region has a substantially circular shape, the contour of the optic papilla region can be shown by combining the contour points specified at each position. Then, a pixel value profile is created by plotting pixel values on a baseline that cuts through the optic disc area. At this time, the pixel values on the baseline are plotted against the blue component pixel values. The baseline passes through the optic disc area and the optic disc depression area, and is designated as a contour point with both ends specified. Therefore, the pixel value profile generally exhibits a mountain-shaped convex shape with the optic disc recessed region having the highest brightness as the top. At this time, if enlargement of the optic disc depression area is progressing (when the possibility of glaucoma is high), the optic disc area formed on the left and right sides (corresponding to the hem) of the optic disc depression area It is confirmed that the range of the corresponding pixel value profile is narrowed, in other words, the left and right hem lengths of the mountain shape are shortened. On the other hand, when the optic disc concavity region is not enlarged (when the possibility of glaucoma is low), a flat hem continues, and only the optic disc concavity region sharply protrudes upward. This makes it possible for the glaucoma identifying means to identify glaucoma symptoms according to the left and right hem lengths.

  Therefore, according to the fundus image analysis system of the present invention, glaucoma symptoms can be identified and determined according to the left and right foot lengths of the created pixel value profile, unlike the conventional calculation of the C / D ratio. . Thereby, it is not necessary to determine the boundary between the optic disc area and the optic disc depression area, and a highly accurate determination can be made.

  Further, the fundus image analysis system of the present invention has, in addition to the above-described configuration, “a ratio of the analysis target region to be specified by the analysis target extraction unit is specified in advance for all regions of the fundus image data, A p-tile extraction unit that uses a p-tile method for determining a threshold value of the number of pixels in the analysis target area according to the ratio may be included.

  Therefore, according to the fundus image analysis system of the present invention, the analysis target region is specified and extracted by applying the p-tile method used for specifying the region in the image analysis technique. Here, it is known in advance that the ratio of the optic disc area to the retinal area of the fundus is specified in a general range, and by specifying the value, the analysis target area can be easily specified.

  Further, the fundus image analysis system of the present invention has the above-described configuration, in which “the contour point forming unit performs a Canny filter process on the converted image data, and a boundary between the optic disc region and the retinal region”. Emphasizing edge, emphasizing the converted image data with enhanced edge, expanding / isolating point removing means for removing isolated points, and converted image data from which the isolated points have been removed Centroid point specifying means for specifying a centroid point of the optic nerve head region from the specified centroid point toward the edge 360 degrees, and the edge located at the shortest distance from the centroid point as a contour point candidate The candidate designating means for designating and combining the plurality of designated contour point candidates by executing an expansion process, and removing the micro area by the area determination process, Be one further comprising, "said contour point determining means for determining the contour points of the nervous papilla region may be.

  Therefore, according to the fundus image analysis system of the present invention, in order to form a contour point, an edge is emphasized by processing using a well-known canny filter (Canny filter), and a contour point is determined by expansion processing and isolated point removal. The contour point can be finally determined by further performing the expansion process and the area determination process again based on the candidates from the center of gravity point to remove the points other than the contour point.

  On the other hand, the fundus image analysis program of the present invention is based on “fundus image data input receiving means for receiving fundus image data input obtained by photographing the fundus of a subject using a fundus camera, from the received fundus image data, An analysis target extraction unit that identifies an analysis target region including an optic nerve head region using a difference in pixel value, extracts target region data, and uses a difference in pixel value of a green component from the extracted target region data A blood vessel erasure image creating means for identifying a blood vessel region and interpolating the pixel value of the identified blood vessel region from the pixel value of the surrounding retinal region to create blood vessel erasure image data, and for the created blood vessel erasure image data A luminance value conversion processing means for performing processing for converting RGB values into luminance values, and density gradation conversion means for performing processing for converting density gradation on the converted luminance value converted image. Conversion processing means for increasing the luminance value difference between the optic nerve head area and the retinal area of the blood vessel erasure image data, and conversion image data processed by the conversion processing means, from the optic nerve head area and the retinal area. A boundary point forming means for detecting a boundary between the optic nerve head regions and forming a contour point of the optic nerve head region, based on the formed contour points, setting a baseline for longitudinally cutting the optic nerve head region; Profile creation means for creating a pixel value profile in which pixel values of the blue component are plotted, and glaucoma identification means for identifying glaucoma symptoms according to the left and right hem lengths of the created pixel value profile, It mainly consists of what makes the fundus image analysis computer capable of providing information to support the diagnosis, A p-tile extraction unit that uses a p-tile method in which the ratio of the analysis target area to be specified is specified in advance for all the areas of the fundus image data, and the threshold value of the number of pixels in the analysis target area is determined according to the ratio As said analysis object extraction means having "or" an edge enhancement means for executing a canny filter process on the converted image data and enhancing a boundary between the optic nerve head area and the retinal area, the edge enhanced Expansion / isolation point removing means for performing expansion processing on the converted image data and removing isolated points, and centroid point specifying means for specifying the centroid point of the optic disc region from the converted image data from which the isolated points have been removed The 360-degree search from the specified centroid point toward the edge is performed, and the edge located at the shortest distance from the centroid point is used as a contour point candidate. The candidate designating means for designating and a plurality of designated contour point candidates are expanded and combined, and the contour point of the optic nerve head region is determined by removing a minute region by area determination processing. As the contour point forming means having the contour point determining means, the fundus image analysis computer may be caused to function.

  Therefore, according to the fundus image analysis program of the present invention, by executing the above program, it becomes possible for the analysis computer to exert an excellent action by the fundus image analysis system.

  As an effect of the present invention, by creating a pixel value profile from acquired fundus image data, the right and left of the pixel value profile can be determined without obtaining the values of the diameter (or radius) of the optic disc region and the optic disc recess region. The glaucoma symptom can be identified by the hem length. Thereby, compared with the conventional method which calculated C / D ratio with the image analysis technique, the identification accuracy of the glaucoma symptom can be improved.

It is a block diagram which shows the functional structure of the analysis computer in the fundus image analysis system of this embodiment. It is a flowchart which shows the flow of the analysis process by an analysis computer. It is explanatory drawing which shows (a) fundus image data, and (b) blood vessel erasure image data. It is explanatory drawing which shows (a) luminance value conversion processing and (b) density gradation conversion processing. (A) Canny filter processing, (b) expansion processing, (c) isolated point removal processing, (d) contour point candidate search processing, (e) expansion processing, (f) area determination processing, (g) contour point determination It is explanatory drawing which shows an example of an image, respectively. It is explanatory drawing which shows an example of the pixel value profile of (a) glaucoma, (b) normal fundus.

  Hereinafter, a fundus image analysis system 1 (hereinafter simply referred to as “analysis system 1”) according to an embodiment of the present invention will be described with reference to FIGS. 1 to 6. Here, FIG. 1 is a block diagram showing a functional configuration of the analysis computer 2 in the analysis system 1 of the present embodiment, and FIG. 2 is a flowchart showing a flow of analysis processing by the analysis computer 2, and FIGS. FIG. 6 is an explanatory diagram showing an example of an image obtained by each process, and FIG. 6 is an explanatory diagram showing an example of a pixel value profile of (a) glaucoma and (b) normal fundus.

  As shown in FIGS. 1 to 6, the analysis system 1 of the present embodiment analyzes fundus image data 3 obtained by photographing the fundus of a subject by applying an image analysis technique, and identifies the symptoms of glaucoma by the analysis computer 2. It is possible to provide useful information for a doctor to make a diagnosis. As shown in FIG. 1, the analysis computer 2 has, as its functional configuration, fundus image data that accepts input of fundus image data 3 of a subject photographed using a known fundus camera (not shown). The input receiving means 11 and the analysis target extraction for identifying the analysis target area A for image analysis including the optic disc area D and the optic disc recessed area C inside from the received fundus image data 3 and extracting it as the target area data 4 The blood vessel region B is identified from the means 12 and the extracted region target data 4 by using the difference in the pixel value of the green component, and the pixel value of the identified blood vessel region B is interpolated from the surrounding retinal region to obtain the blood vessel erased image Blood vessel erasure image creation means 13 for creating data 5, luminance value conversion processing means 14 for performing processing for converting the RGB values into luminance values for the created blood vessel erasure image data 5, and darkness A boundary between the optic disc region D and the outer retinal region R is obtained from the conversion processing unit 16 having the density gradation conversion element processing unit 15 for gradation conversion and the converted image data 6 processed by the conversion processing unit 16. Contour point forming means 17 that detects and forms the contour point OP of the optic nerve head region D, and passes through the optic nerve head region D and the optic nerve head recessed region C longitudinally based on the formed contour point OP. A profile creation means 18 for creating a pixel value profile 7 in which a blue line pixel value on the baseline L is plotted along the baseline L, and a pixel value profile 7 The right and left hem length W is examined, and glaucoma identifying means 19 for identifying glaucoma symptoms is mainly provided based on comparison with the normal fundus.

  Here, the analysis target extraction unit 12 further includes a p tile extraction unit 20 using a p tile method (details will be described later). At this time, since the optic disc area D is indicated by pixels brighter than the surrounding retinal area R with respect to the analysis target area A, identification by the p tile method is facilitated. In addition, in order to specify the analysis target area A, it may be used in combination with the p-tile method and further use the characteristics of the circularity, area, and color information of the optic nerve head area D.

  The contour point forming unit 17 specifies the contour point OP of the optic nerve head region D from the converted image data 6 image-processed by the conversion processing unit 16 and forms a contour OL surrounding the entire circumference. Specifically, the converted image data 6 is processed by a Canny filter, and an edge enhancement unit 21 that enhances the boundary (edge) between the optic disc region D and the retina region R, and enhancement of the edge E A dilation / isolation point removal unit 22 that performs an expansion process on the converted image data 6 ′ and further removes isolated points; and a centroid point specification unit 23 that specifies a centroid point C from the optic nerve head region D; A candidate designating means 24 for searching 360 degrees from the specified centroid point C toward the emphasized edge E and designating each edge E located at the shortest distance from the centroid point C as a contour point candidate OP ′; The contour points of the optic nerve head region are obtained by performing expansion processing on a plurality of contour point candidates OP ′ and combining them, and further removing the minute regions by area determination processing. And a contour point determining means 25 for determining the P.

  Here, as the analysis computer 2, a commercially available personal computer and its peripheral devices can be used. As other general configurations, the read fundus image data 3 and each image data generated by each process are included. Storage means 26 for storing or temporarily storing 6 etc., display control means 27 for signal-controlling and sending display signals to display means (not shown) such as a liquid crystal display for displaying processing results, In addition, various commands and data are input to the analysis computer 2 and connected to operation means (not shown) such as a keyboard for operating the analysis computer 2, and an operation input signal input via the operation means Has a functional configuration such as an operation control means 28 for accepting.

  Next, an example of glaucoma symptom identification processing by image analysis of the fundus image data 3 by the analysis system 1 of the present embodiment, the image example in each processing of FIGS. 3 to 5, and the pixel of FIG. A description will be given based on the value profile 7.

  First, the analysis computer 2 acquires the fundus image data 3 related to the fundus of the subject, which is captured by the fundus camera and converted into electronic data, from a storage medium such as an HDD in which the fundus image data 3 is stored in advance. (Step S1: FIG. 3A).

  Then, the analysis target area A including the optic disc area D is extracted by the analysis target extraction means 12 (step S2: FIG. 3A). Here, the analysis target extraction unit 12 is extracted by the p tile extraction unit 20 using the p tile method. The p-tile method is one of well-known methods in image analysis processing. Specifically, the area S of the entire image and the area S ′ of the analysis target area A are set first, and the analysis target area A for the entire image is set. The area ratio p = S ′ / S is calculated. At this time, when the pixels of the analysis target area A are brighter than the pixels of the entire image, the number of pixels is counted from the higher density value, while when the pixels of the analysis target area A are darker than the pixels of the entire image. Is counted from the one with the smallest density value, the number of pixels at that time is N ′, and the number of pixels of the entire image is represented by N, the density value satisfying the value of N ′ / N = p is determined as a threshold value. Thereby, the process of extracting the analysis target area A according to the threshold value is performed. Here, the pixels of the optic nerve head region D that occupies most of the analysis target region A are brighter than the pixels of the surrounding retinal region R. Therefore, identification by the p-tile method is easily performed and extracted as target region data 4 Is done. At this time, the characteristics of circularity, area, and color information of the optic papilla region D may be further used when the analysis target region A is specified.

  Then, a plurality of blood vessel regions B included in the extracted target region data 4 related to the analysis target region A and erased by the blood vessel erased image creating means 13 are erased by image analysis and erased by the surrounding retinal region R. A process for interpolating the region is performed (step S3). Here, since the blood vessel region B is generally displayed in a darker color than the retinal region R, the optic nerve head region D, or the like, the blood vessel region B can be easily identified. At this time, the blood vessel region B is specified using a difference in pixel value of the green component included in the target region data 4 indicated by the color image. Then, the specified blood vessel region B is erased from the target region data 4, and the erased region is interpolated using the pixel values of the surrounding retinal region R, so that the blood vessel erased image data 5 without the blood vessel region B is present. Is formed (FIG. 3B).

  And the process which enlarges the luminance value difference of the optic disc area | region D and another area | region (retinal area | region R) with respect to the produced | generated blood vessel erasure | elimination image data 5 by the change process means 16 is performed. More specifically, first, a process of converting the RGB values of the red component, the green component, and the blue component contained in the blood vessel erasure image data 5 indicated by the color image into luminance values is performed (step S4: FIG. 4A). . Thereby, the pixel values of the respective color components are emphasized, and particularly, the respective pixel values indicating the optic nerve head region D are emphasized so as to show a constant value. After that, the density gradation conversion processing means 15 performs density gradation conversion processing on the blood vessel erasure image data 5 'after the conversion processing to the luminance value (step S5: FIG. 4B). By the conversion processing means 16 having these two processes, the relative luminance value difference between the optic disc region D and the retinal region R surrounding it increases (see pixel value profile).

Thereafter, a process of forming the contour point OP of the optic disc region D from the converted image data 6 obtained by the above process is performed. Specifically, a Canny filter process is performed on the converted image data 6 to perform an edge emphasis process for emphasizing the edge E (bending point) of the boundary between the optic disc region D and the retinal region R (step) S6: FIG. 5 (a)). The Canny filter process and the Canny method that defines such a process are well-known techniques in the image analysis process, and thus detailed description thereof is omitted here. Then, an expansion process is performed on the image after the canny filter process (FIG. 5A) (step S7: FIG. 5B). Here, the expansion process is a process for converting a noise value (particularly an isolated point) in a graphic component by converting the value of a pixel in the background component that is in contact with the boundary of the graphic component into a pixel value of the graphic component and expanding it by one pixel. N) is a process capable of displaying prominently. And the image after the expansion process (FIG. 5B)
A process for removing isolated points from the image is performed (step S8: FIG. 5C).

  Then, the centroid point C in the optic nerve head region D is specified from the image after the isolated point removal process (FIG. 5C) (step S9), the search is performed in the direction of 360 degrees from the centroid point C, and the searched edge E is designated as the contour point candidate OP ′ (step S10: FIG. 5D), and further expansion processing and area determination processing are executed for the plurality of contour point candidates OP ′ (step S11: FIG. 5E). And FIG. 5 (f)), the contour point OP is finally determined (step S12: FIG. 5 (g)).

  Thereafter, a baseline BL that vertically traverses the optic disc region D in which the contour point OP is specified is set (FIG. 5G), and a pixel value profile in which the pixel values of the blue component of the fundus image data 3 on the baseline BL are plotted. 7 (step S13: FIG. 6A), where the vertical axis indicates the pixel value of the blue component and the horizontal axis indicates the coordinates of the base line BL. The left and right axial ends of the horizontal axis are contour points OP corresponding to the boundary between the optic disc region D and the retina region R specified by the above processing.

  Then, glaucoma is identified based on the created pixel value profile 7, and identification data 30 including the identification result is stored (step S14). Here, the glaucoma is identified by the left and right skirt lengths W of the pixel value profile 7. That is, when there is a suspicion of glaucoma, the width of the rim is narrowed, so that a pixel value profile having almost no skirt length W is formed on the image as shown in FIG.

  On the other hand, in the normal optic disc region D, a rim is formed around the optic disc recessed region C. That is, the skirt length W of the pixel value profile 7 'increases as shown in FIG.

  As described above, glaucoma symptoms are identified by forming the pixel value profile 7 from the fundus image data 3 acquired from the subject through various image processing steps by the analysis system 1 of the present embodiment. It becomes possible. The final diagnosis of glaucoma is comprehensively judged by a specialized ophthalmologist in consideration of other examination conditions, etc., and this system provides doctors with useful information for making the diagnosis. Can be provided. As a result, the analysis computer 2 can make a primary determination when diagnosing a large number of fundus image data 3 acquired by a health examination or the like.

  In particular, as compared with the conventional image analysis technique for the fundus image data 3, it is possible to omit the accurate specification of the optic disc recessed region C. That is, the calculation of the C / D ratio can be omitted. This eliminates the possibility of errors due to the measurement of the optic disc area D and the optic disc depression area C, which has conventionally been necessary when calculating the C / D ratio, and identifies glaucoma with high accuracy. Is possible.

  The present invention has been described with reference to preferred embodiments. However, the present invention is not limited to these embodiments, and various modifications can be made without departing from the spirit of the present invention as described below. And design changes are possible.

  That is, in the analysis system 1 of this embodiment, the target region data 4 is extracted from the fundus image data 3, and then the blood vessel region B is erased and interpolated to generate the blood vessel erased image data 5. However, the present invention is not limited to this, and the order of such processing may be reversed. If the fundus image data 3 targets a relatively small area, the extraction of the target region data 4 is omitted, the blood vessel region B is specified for the entire fundus image data 3, and the blood vessel erased image data 5 is stored. It may be generated. 3 to 4 exemplify an image displaying the entire fundus image data 3 after extraction of the target area data 4 in order to simplify the description.

1 Analysis system (fundus image analysis system)
2 analysis computer 3 fundus image data 4 target area data 5, 5 ′ blood vessel erasure image data 6 converted image data 7 pixel value profile 11 fundus image data input receiving means 12 analysis object extraction means 13 blood vessel erasure image creating means 14 luminance value conversion processing Means 15 Density gradation conversion processing means 16 Conversion processing means 17 Contour point forming means 18 Profile creation means 19 Glaucoma identification means 20 p tile extraction means 21 Edge emphasis means 22 Expansion / isolation point removal means 23 Centroid point specification means 24 Candidate designation means 24 25 Contour point determination means A Analysis target region B Blood vessel region BL Baseline C Optic disc depression region D Optic disc region E Edge OP Contour point OP 'Contour point candidate R Retina region

JP 2004-49259 A JP 2006-263127 A

Claims (6)

A fundus image analysis system capable of providing information for supporting a diagnosis of glaucoma by a doctor by acquiring fundus image data obtained by photographing the fundus of a subject and analyzing the fundus image data using an analysis computer There,
The analysis computer
Fundus image data input accepting means for accepting input of the fundus image data of the subject acquired using a fundus camera;
From the received fundus image data, an analysis target extraction unit that identifies an analysis target region including the optic nerve head region using a difference in pixel value, and extracts target region data;
A blood vessel region is identified from the extracted target region data using the difference in pixel value of the green component, and the blood vessel erasure image data is created by interpolating the pixel value of the identified blood vessel region from the surrounding retinal region. A blood vessel erasure image creating means;
Luminance value conversion processing means for performing processing for converting RGB values into luminance values for the created blood vessel erasure image data, and a density scale for performing processing for converting density gradation on the converted luminance value converted image Conversion processing means for increasing a luminance value difference between the optic disc area and the retinal area of the blood vessel erasure image data,
Contour point forming means for detecting a boundary between the optic disc region and the retinal region from the converted image data processed by the conversion processing unit, and forming a contour point of the optic disc region;
Based on the formed contour points, a profile creating means for setting a baseline for longitudinally cutting the optic disc region and creating a pixel value profile in which pixel values of blue components on the baseline are plotted;
A fundus image analysis system comprising glaucoma identification means for identifying glaucoma symptoms according to the left and right hem lengths of the created pixel value profile. The analysis object extraction means includes
A p-tile extraction unit that uses a p-tile method in which the ratio of the analysis target area to be specified is specified in advance for all the areas of the fundus image data, and the threshold value of the number of pixels in the analysis target area is determined according to the ratio The fundus image analysis system according to claim 1, further comprising: The contour point forming means includes:
Edge enhancement means for performing canny filtering on the converted image data, and emphasizing a boundary between the optic disc region and the retinal region;
An expansion / isolated point removing means for performing an expansion process on the converted image data in which edges are emphasized to remove isolated points;
A barycentric point specifying means for specifying a barycentric point of the optic disc region from the converted image data from which the isolated points have been removed;
Search for 360 degrees from the specified centroid point toward the edge, and candidate specifying means for specifying the edge located at the shortest distance from the centroid point as a contour point candidate;
A contour point determining means for determining the contour point of the optic nerve head region by performing expansion processing on a plurality of designated contour point candidates and combining them, and removing a minute region by area determination processing; The fundus image analysis system according to claim 1, further comprising a fundus image analysis system.   A fundus image data input receiving unit that receives input of fundus image data acquired by photographing the fundus of a subject using a fundus camera, and using the difference in pixel value from the received fundus image data, the optic disc area is determined. An analysis target extraction unit for specifying an analysis target region including the target region data, specifying a blood vessel region from the extracted target region data using a difference in pixel value of a green component, and specifying the specified blood vessel region Blood vessel erasure image creating means for creating blood vessel erasure image data by interpolating the pixel values of the retinal region from the pixel values of the surrounding retinal regions, and processing for converting RGB values into luminance values for the created blood vessel erasure image data Luminance value conversion processing means, and density gradation conversion means for performing a process of converting density gradation on the converted luminance value conversion image, and the optic nerve milk of the blood vessel erasure image data Conversion processing means for increasing a luminance value difference between the region and the retinal region, detecting a boundary between the optic disc region and the retinal region from the converted image data processed by the conversion processing unit, and the optic disc region A contour point forming means for forming a contour point of the image, and a pixel value profile obtained by setting a base line for longitudinally cutting the optic disc region based on the formed contour point and plotting a pixel value of a blue component on the base line As a glaucoma identification means for identifying glaucoma symptoms according to the left and right foot lengths of the created pixel value profile, information for supporting a diagnosis of glaucoma by a doctor can be provided. A fundus image analysis program for causing a fundus image analysis computer to function.   A p-tile extraction unit that uses a p-tile method in which the ratio of the analysis target area to be specified is specified in advance for all the areas of the fundus image data, and the threshold value of the number of pixels in the analysis target area is determined according to the ratio The fundus image analysis program according to claim 4, further causing the fundus image analysis computer to function as the analysis target extraction unit having the function. Edge conversion means for emphasizing a boundary between the optic disc area and the retinal area, and expansion processing for the converted image data with the edge enhanced. , Expansion / isolation point removing means for removing isolated points, centroid point specifying means for specifying a centroid point of the optic disc region from the converted image data from which the isolated points have been removed, and from the specified centroid point to the edge 360 degree search, candidate designation means for designating the edge located at the shortest distance from the barycentric point as a contour point candidate, and performing a dilation process on the plurality of designated contour point candidates, As the contour point forming unit having a contour point determining unit that determines the contour point of the optic disc region by removing a minute region by area determination processing, Serial fundus image analysis program according to claim 4 or claim 5, characterized in that to the fundus image analysis computer further functions.