JP2008237840A - Image analysis system and image analysis program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image analysis system capable of producing a highly accurate depth map by specifying the noise region included in an object to be photographed with roundedly changing unevenness and correcting the depth information. <P>SOLUTION: An image analysis computer 2 constituting the image analysis system 1 includes profile curve computing means 21, depth map creating means 22, center point determining means 23, contour part specifying means 25, noise area specifying means 28, noise area eliminating means 29, curve correcting means 30 with spline curve computing means 27, eliminated depth information computing means 32, and corrected map output means 33. The depth information 13 is corrected based on a depth curve 8a to create an accurate depth map, and the created map can be displayed. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image analysis system and an image analysis program, and more particularly to an image analysis system and an image analysis program that can provide doctors with information useful for diagnosis of ophthalmic diseases such as glaucoma. .

  Conventionally, the eyeball is irradiated with light from the outside, and the state of the fundus is observed through the eyeball. In addition, a fundus photograph obtained by imaging the state of the fundus using an optical device such as a camera is left as a medical record. By examining this fundus photograph in detail, a doctor can make a diagnosis of various diseases. Here, observation of the fundus is not generally imposed on a subject (patient) and can be obtained by a device with a relatively simple configuration, and thus is generally performed in a medical institution. It is one of the inspection methods.

  In addition, blood vessels taken by fundus photography are a part of the cerebral blood vessels in the brain located in the back of the eyeball, and are the only method that can directly observe the inside of the brain from outside the body. Therefore, very useful information can be provided to doctors and the like. In addition, this fundus photograph is used not only for diagnosis of ophthalmic diseases such as glaucoma but also for diagnosis of lifestyle-related diseases such as diabetes.

  Here, in the fundus photograph of a healthy subject, the entire fundus is yellowish reddish brown, and is a circle or oval of 1.2 mm to 1.7 mm at a position about 15 degrees outward (ear side direction) of the subject's line of sight. An optic nerve head is observed. This optic nerve head is anatomically called an optic disc, and the boundary between the fundus portion and the optic nerve head is more clearly observed on the ear side than on the nose side. Is. In the vicinity of the center of the optic nerve head, an area called a optic nerve crevice or a physiologic cup having a circular shape is observed.

  When diagnosing glaucoma using fundus photographs, as one of the criteria for diagnosis, the ratio of the diameters of the optic nerve head and the optic nerve crevice (depression) described above, that is, the substantially circular optic nerve. The value of the diameter ratio (hereinafter referred to as “C / D ratio”) of the optic disc depression (C: Cup) to the nipple (D: Disc) is obtained. Among the C / D ratios, the C / D ratio in the vertical direction (vertical C / D ratio) is particularly useful for determining the presence or absence of glaucomatous changes. As one criterion, the vertical C / D ratio is 0.7. Glaucoma is suspected in the above cases or when the difference in the vertical C / D ratio between the right and left fundus is 0.2 or more. Diagnosis of glaucoma is not made only by the value of the C / D ratio, but is made in combination by other examinations, doctor's findings, etc. The value of the C / D ratio is determined by the diagnosis. This value is not one of the most useful information for glaucoma.

  Here, the structure of the fundus will be described in more detail. The fundus, which is a part of the eyeball, is three-dimensionally configured with a three-dimensional curved surface (spherical surface). However, the above-described fundus photograph is actually configured by converting it into a two-dimensional (planar) form even if it has a three-dimensional structure. Therefore, when diagnosing various diseases using the fundus photograph, the two-dimensional fundus photograph is imaged in a three-dimensional manner in the head, and the optic disc depression is based on it. Judgment of hypertrophy, etc. was sometimes made. Therefore, there is a possibility that an ophthalmologist who has little diagnostic experience cannot make an accurate diagnosis. Therefore, attempts have been made to add three-dimensional information and increase the accuracy of the C / D ratio to enable accurate diagnosis.

  For example, a blood vessel region included in a photographed fundus image is extracted, a data missing portion from which the blood vessel region has been removed is interpolated based on the surrounding data, and the fundus depth information is obtained using this. Are known.

Japanese Patent Laid-Open No. 04-276232

  However, the image processing method for the fundus image shown above may cause the following problems. That is, in the above-described fundus image, a plurality of blood vessels (blood vessel portions) travels in a complicated manner, and recognition and removal of the blood vessel portions are relatively difficult. Furthermore, the fact that the blood vessel part is agglomerated may cause the accuracy of the three-dimensional information of the fundus obtained by the interpolation to be significantly reduced because the distance from the surrounding data used for the interpolation is too far. there were. In addition, the fundus image may include noise (such as imaging noise) that hinders accurate acquisition of depth information in addition to the blood vessel region that travels in the fundus. These image processing methods However, it was not possible to interpolate the area occupied by noise other than blood vessels.

  Therefore, in view of the above situation, the present invention specifies a noise region included in a subject to be photographed having curved irregularities, corrects depth information, and makes it possible to create a highly accurate depth map. In particular, an image analysis system and an image analysis program capable of identifying a blood vessel region and noise other than blood vessels included in the optic disc concavity region of the fundus image and creating an accurate depth map by correction Providing is an issue.

  In order to solve the above-described problems, the image analysis system of the present invention corrects depth information calculated from image data of a concave or convex object that changes in a curve using an image analysis computer. An image analysis system, wherein the image analysis computer is configured to obtain one-dimensionally the depth of the object based on information data acquisition means for acquiring the depth information from the image data and the acquired depth information. A profile curve calculating means for calculating a depth profile curve expressed by a simple distribution, a center point determining means for determining a center point of the object based on at least one of the image data and the depth information; An outline specifying means for specifying the outline of the object, and a point on the curve of the depth profile curve corresponding to the position of the specified outline as a starting point, Noise area specifying means for specifying a noise area of the depth profile curve based on a gradient change in the depth direction of the depth profile curve from the starting point toward the center point, and the depth corresponding to the specified noise area A curve correction unit that corrects the depth profile curve based on a change in the surrounding depth profile curve, and a depth information correction unit that corrects the depth information based on the corrected depth profile curve. Mainly configured.

  Here, the image data constitutes a concave or convex object that changes in a curve, for example, a topographic map of a mountainous terrain taken from the sky as an aerial photograph, or a part of an eyeball that exhibits a substantially spherical shape The fundus image obtained by photographing the fundus is obtained as electronic data, and the depth information can display the curved state of the fundus as a numerical value. Three-dimensional position information is given to the depth information acquired from the image data. Based on this depth information, it is possible to construct a three-dimensional fundus image that displays the fundus in three dimensions.

  As a method for acquiring depth information from image data such as a fundus image, a known technique can be used. For example, an image can be taken using a photographing device such as a stereo fundus camera capable of stereoscopic shooting. Based on this, parallax is obtained from two fundus images displayed two-dimensionally obtained by shifting the photographing position for photographing the fundus, and obtaining image data including depth information based on the image data. The three-dimensional structure is constructed and acquired as a three-dimensional fundus image.

  The information data acquisition means is a means for appropriately reading the image data stored in advance in a storage means (hard disk drive or the like) built in the image analysis computer by operating a keyboard or the like, or storing it in a storage medium such as a CD-R. One that reads the image data that has been read, or one that acquires the image data via a communication network. That is, any data can be used as long as data stored in a format that can be analyzed by the image analysis computer can be read as appropriate. Furthermore, it is also possible to combine the above-described stereo fundus camera and the image analysis system of the present invention, and use the image data including the captured depth information as it is for the analysis processing.

  The profile curve calculation means calculates a depth profile curve capable of displaying the depth of the object in a one-dimensional distribution based on depth information included in the acquired image data. For example, when the object is the fundus, assume a virtual component (approximately equivalent to the diameter of the optic disc region) that passes through the approximate center of the fundus (corresponding to the optic disc region) and place this on the horizontal axis. The value corresponding to the depth of the fundus is sequentially plotted on the vertical axis. As a result, a curvilinear change with respect to the depth direction of the fundus can be indicated by the depth profile curve.

  The center point determining means determines the center point of the object based on image data and / or depth information (at least one of image data and depth information), and the image data and / or Even if it is determined directly from the depth information, or a depth profile curve is derived based on image data and / or depth information, and then determined using the depth profile curve. Also good. In other words, it may be derived indirectly from image data and / or depth information.

  The contour portion specifying means specifies a reference position (starting point in the depth profile curve) from the acquired image data using the difference in pixel value. For example, when the object is a fundus, the fundus region may be significantly different from the rest of the fundus region, and the fundus region displayed in a relatively dark color and the optic nerve head region displayed in a light color such as milky white And the like can be specified as a contour portion.

  The noise region specifying means starts from a point on the curve of the depth profile curve corresponding to the specified contour portion, and noise is determined based on the gradient change in the depth direction from the start point to the previously determined center point. The area is specified. That is, when the object is a fundus, the fundus is theoretically changed by a smooth curve toward the fundus center. Therefore, in these curve changes, when the rate of change changes significantly, it is a depression to assume that so-called noise that does not indicate the original fundus shape is included. Therefore, a place where the curve change rate changes greatly is identified, and the place (area) is specified as a noise area.

  Therefore, according to the image analysis system of the present invention, the depth profile curve of the object is obtained based on the depth information, and the center point determined from the position of the depth profile curve corresponding to the specified contour portion The gradient change in the depth direction toward is detected. Then, a range in which the change in the depth gradient is remarkably changed is specified as a noise region, a depth profile curve corresponding to the specified noise region is corrected based on a change in the surrounding depth profile curve, and the Based on the depth profile curve, the depth information is corrected, and the result is reflected in the depth information and displayed. Thereby, when the object is the fundus, the blood vessel and other noise regions included in the fundus are specified as the noise region, and the depth information is constructed in a state in which the portion is erased. As a result, highly accurate three-dimensional image information can be obtained. Thereby, in the case of an optic nerve head region constituted by a curved surface curved in a concave shape, it is possible to provide a diagnostic method particularly useful for diagnosing glaucoma. Here, the curve includes a linear curve, that is, a “straight line”.

  Further, the image analysis system according to the present invention has, in addition to the above configuration, “the image analysis computer further includes a noise region erasing unit that erases the identified noise region from the depth profile curve, and the curve correcting unit. Calculates a correction profile curve that substantially matches the rate of change of the depth profile curve remaining after the noise region is erased, and the depth information correction means is based on the calculated correction profile curve, It may be “correcting the depth information”.

  Here, the noise area erasing means refers to the position of the depth profile curve corresponding to the end of the specified noise area, that is, the range indicated by the noise area on the imaginary line segment (noise profile curve). The curve in the range is identified and deleted from the depth profile curve. Thereby, the depth profile curve expressed one-dimensionally as one depth profile curve is divided into a plurality, at least two or more.

  Furthermore, the curve correction means in the present invention calculates a correction profile curve that corrects the location of the erased noise region profile curve based on the remaining depth profile curve that is divided into a plurality of lines. The calculation of the curve is, for example, to calculate a curve that connects the ends of the divided depth profile curves adjacent to each other. At this time, the correction profile curve is calculated such that the ends are connected with a smooth curve based on the change rate of the two depth profile curves to be connected. As a result, the depth profile curve divided by the correction profile curve is again one-dimensionally expressed by a smooth curve. Based on the result, the depth information of the imaging target is corrected. Thereby, the gradient change of the depth which goes to the center point from the starting point becomes smooth. Then, the depth information is corrected by the depth information means using the calculated correction profile curve. Correction of the correction profile curve by the curve correction means may be performed by correcting the correction profile curve of the singular region portion, or by correcting the correction profile curve of the singular region portion and interpolating between them. Any method may be used.

  Furthermore, the image analysis system according to the present invention has a depth map used for displaying the distribution of the depth of the object based on at least one of the acquired image data and the depth information. A depth map creating means for creating the depth map, a depth map correcting means for correcting the depth map based on the information obtained by the depth information correcting means, and the depth corrected by the depth map correcting means. And a depth map display means for displaying a depth map ”. Further, “the depth information correcting unit may correct the depth map instead of the depth information” may be used.

  Depth map creation means displays so that the depth distribution in the object (for example, the fundus) can be visually recognized based on the acquired image data or depth information corresponding thereto. (Expression). More specifically, for example, in the fundus image in which the optic disc region is photographed, the optic disc concavity region present at the deepest position is expressed in dark color (for example, black), and the periphery is gradually stepped. For example, a light color (for example, white) is used. As a result, the optic disc area of the dusk color (gray color) slightly darkened around the optic disc depression area expressed in black, and the fundus area expressed in light color (white) around it In other words, it is expressed step by step in so-called “grayscale”. Note that the depth map expression method is not limited to the one that expresses the distribution by the shade of a single color as in the gray scale described above, and uses a plurality of colors, for example, a cold color ( If the distribution of depth is clearly shown to the observer, such as those expressing deep parts in blue, green, purple, etc., and expressing shallow parts in warm colors (red, yellow, orange, etc.) I do not care.

  Therefore, according to the image analysis system of the present invention, the depth map is corrected based on the information obtained by the depth information correcting means, and the depth map obtained thereby is displayed. The depth map display means displays the depth distribution in one to three dimensions. Here, the derivation of the depth profile curve may be determined directly from the image data and the depth information, or the depth map and / or the depth profile curve is derived based on the image data and the depth information. However, it may be derived using a depth map and / or a depth profile curve. In other words, it may be derived indirectly from image data and / or depth information.

  Furthermore, in addition to the above configuration, the image analysis system of the present invention may be configured so that “the noise region specifying means has a change in gradient in the depth direction of the depth profile curve from the starting point toward the center point based on a preset change rate. And a portion corresponding to a noise profile curve on at least one of the depth profile curves where the gradient change changes in the opposite direction may be specified as the noise region.

  Therefore, according to the image analysis system of the present invention, when the noise region is specified when the gradient change in the depth direction becomes larger than the preset change rate or the change in the depth direction changes in the reverse direction. It is specified as the noise region. Such a location is assumed to be noise or the like generated in the processing when acquiring depth information.

  Furthermore, in addition to the above configuration, the image analysis system of the present invention may include “the curve correction unit includes a spline curve calculation unit that calculates a spline curve as the correction profile curve”.

  Therefore, according to the image analysis system of the present invention, the correction profile curve is calculated as a spline curve. Here, the spline curve changes in a curve connecting two points (corresponding to the end of the depth profile curve), and a method for obtaining the spline curve is already well known. Therefore, the spline curve formed as the correction profile curve can be approximately approximated to the curve change rate of the depth profile curve before and after the correction profile curve. As a result, for example, when the object is the fundus, correction can be performed with a correction profile curve that substantially matches the curve change rate of the fundus, and the accuracy of the depth information obtained is improved.

  Furthermore, the image analysis system of the present invention has the above-described configuration, and “the image data uses a fundus image acquired by using a fundus image capturing apparatus and imaging the optic disc depression region”. It doesn't matter. Further, “the noise region erasing means includes a plurality of blood vessel regions that travel in the optic disc depression region of the fundus, imaging noise that occurs during imaging of the fundus, and depth information processing for acquiring the depth information. The processing noise generated at the time may be specified as the noise region and erased ”.

  Therefore, according to the image analysis system of the present invention, the optic disc recessed area that is a part of the fundus is selected as the imaging target (corresponding to the object), the depth information for the area of the fundus image is corrected, and The accuracy can be increased. As a result, there is no error in the depth information due to a plurality of blood vessels running in the optic disc concavity region of the fundus, and by using the corrected depth information and depth map, the optic disc concavity of the fundus is used. It becomes possible to construct a three-dimensional image of a region with high accuracy. Thereby, the calculation accuracy of the C / D ratio that can be used for diagnosis of glaucoma and the like is improved, and it is possible to provide useful information for diagnosis by a doctor or the like.

  On the other hand, the image analysis program of the present invention is based on “the information data acquisition means for acquiring the depth information calculated from the image data of the concave or convex object that changes in a curve, based on the acquired depth information. A profile curve calculating means for calculating a depth profile curve expressing the depth of the object in a one-dimensional distribution, based on at least one of the image data and the depth information. Center point determination means for determining a center point, contour part specifying means for specifying the contour part of the object, starting from a point on the curve of the depth profile curve corresponding to the position of the specified contour part, A noise region specifying means for specifying a noise region of the depth profile curve based on a gradient change in the depth direction of the depth profile curve from the starting point toward the center point; The depth profile curve corresponding to the corrected noise region is corrected based on a change in the surrounding depth profile curve, and the depth information is calculated based on the corrected depth profile curve. As the depth information correcting means for correcting, an image analysis computer is allowed to function "," noise area erasing means for erasing the identified noise area from the depth profile curve, and the noise area remaining after being erased " The curve correction unit that calculates a correction profile curve that substantially matches the rate of change of the depth profile curve, and the depth information correction unit that corrects the depth information based on the calculated correction profile curve. "A further function of the image analysis computer" or "the acquired image data or Based on at least one of the depth information, a depth map creating means for creating a depth map used for displaying the distribution of the depth of the object, based on information obtained by the depth information correcting means The image analysis computer further functions as a depth map correction unit that corrects the depth map and a depth map display unit that displays the depth map corrected by the depth map correction unit. It does not matter. Further, the image analysis program according to the present invention is “the depth information correcting unit corrects a depth map instead of depth information” or “the image data uses a fundus image photographing device, The fundus image acquired using the nipple depression region as an imaging target may be used. "

  Therefore, by executing the image analysis program of the present invention, it is possible to cause the image analysis computer to exhibit the above-described excellent operational effects.

  As an effect of the present invention, a one-dimensional depth profile curve is calculated from image data including depth information, and the gradient change in the depth direction from the origin of the fundus etc. to the center point on the depth profile curve is smooth. By assuming that the noise region is one, it is possible to identify a noise region, create a highly accurate depth map corrected by deleting the noise region, and display it. As a result, when performing diagnosis of the fundus etc. using these image data, it is possible to provide highly reliable and useful information.

  Hereinafter, an image analysis system 1 and an image analysis program according to an embodiment of the present invention will be described with reference to FIGS. Here, FIG. 1 is a block diagram showing a functional configuration of the image analysis computer 2 in the image analysis system 1 of the present embodiment, and FIG. 2 images the fundus 5 including the optic disc area 3 and the optic disc depression area 4. FIG. 3A is a schematic diagram showing a cross section of the extraction region 7, FIG. 3B is a curve diagram showing an example of the depth profile curve 8a of the extraction region 7, and FIG. FIG. 4C is a map diagram showing an example of the depth map 9a of the extraction region 7. FIG. 4A is a schematic diagram showing a cross section of the extraction region 7. FIG. 4B shows an example of the identified noise profile curve 11. FIG. 5 is a map diagram illustrating an example of a depth map 9b from which a noise region 10 has been removed, and FIG. 5 illustrates an example of a depth profile curve 8b from which the noise profile curve 11 has been removed. Curve diagram, b) a curve diagram showing an example of a depth profile curve 8c connected by the correction profile curve 12, and (c) a map diagram showing an example of a corrected depth map 9c. FIG. 6 and FIG. FIG. 8 is a flowchart showing a flow of processing of the image analysis computer 2 related to the correction of the map 9a. FIG. 8A is a curve diagram showing another example of the correction of the depth profile curve 8a, and FIG. 8B is a corrected depth file. It is a curve figure which shows the curve 8d. Here, in the following specification, the depth profile curves 8a, 8b, 8c, 8d are “depth curves 8a, 8b, 8c, 8d”, and the noise profile curve 11 is “noise curve 11”. The correction profile curve 12 is referred to as a “correction curve 12”.

  As shown in FIGS. 1 to 7, the image analysis system 1 according to the present embodiment captures a fundus 5 including a photographed optic disc region 3 and a optic disc recessed region 4 (corresponding to an object in the present invention). ). The depth information 13 included in the acquired fundus image data 6 and the depth maps 9a and 9b and the depth curve 8a created based on the depth information 13 correspond to the boundary between the fundus region 14 and the optic disc region 3. It is assumed that a change from the contour portion 15 to the deepest portion (corresponding to a center point C described later) of the optic disc recessed region 4 is indicated by a smooth concave curve, and based on this The depth information 13 and the depth map 9a are corrected. As a result, a cupline indicating the outline of the optic disc recessed region 4 is accurately determined, and an accurate value of the ratio (C / D ratio) of the optic disc recessed region 4 / optic disc region 3 is provided to an ophthalmologist or the like. However, the image analysis computer 2 can mainly support diagnosis of ophthalmic diseases such as glaucoma.

  Here, the image analysis computer 2 can use a general-purpose personal computer, and can acquire fundus image data 6 including depth information 13 for photographing the fundus 5 and displaying it in three dimensions. The stereo fundus camera 16, the liquid crystal display 17 as a display means for displaying the results of various processes and analysis results for the acquired fundus image data 6, and various commands and data are input to the image analysis computer 2. These are connected to operation input devices 18 such as a keyboard and a mouse (not shown). Various data and information acquired or created by the stereo fundus camera 16 and the like are stored in the data storage means 19 built in the image analysis computer 2. The data stored in an external storage device (for example, HDD or the like: not shown) connected to the image analysis computer 2 is read at any time, or is read into a data server (not shown) connected to the network. You may accumulate | store and acquire this by reading via LAN or the internet.

  Next, the functional configuration of the image analysis computer 2 will be described. As mainly shown in FIG. 1, a data storage means 19 that is photographed by a stereo fundus camera 16 and stores fundus image data 6 including depth information 13 is stored. Information data acquisition means 20 for reading the stored fundus image data 6 from the data storage means 19 and acquiring depth information 13 relating to the fundus image data 6 and the depth of the fundus 5 to be imaged of the fundus image data 6; Based on the acquired depth information 13, profile curve calculation means 21 that calculates a depth curve 8 a expressing the depth of the fundus 5 in a one-dimensional distribution, and acquired fundus image data 6 and depth information 13 Based on the depth map creating means 22 for creating the depth map 9a used for displaying the depth distribution of the optic disc region 3, the calculated depth curve 8a, Center point determination means 23 for determining the center point C of the fundus 5 from the depth map 9a, and contour part specifying means 25 for specifying the contour part 15 of the optic nerve head region 3 using the difference in pixel values; The points on the curve of the depth curve 8a in the specified contour portion 15 are set as starting points S1 and S2, and the gradient change in the depth direction of the depth curve 8a toward the center point C determined from the starting points S1 and S2 Is determined as a noise curve 11 and a noise region 10 corresponding to the noise curve 11 is specified. Means 28, a noise area erasing means 29 for erasing the identified noise curve 11 from the depth curve 8a, and further erasing the identified noise area 10 from the depth map 9a; Remaining A curve correction means 30 having a spline curve calculation means 27 for calculating a correction curve 12 represented as a spline curve substantially matching the rate of change of the depth curve 8b divided into at least two or more. The depth curve 8b divided based on the corrected curve 12 is connected and corrected, and the depth information 13 of the erased region 26 of the depth map 8b from which the noise region 10 is removed is corrected. Based on the depth information 13 in the depth map 8b around the erasing region 26 and based on the calculated removal depth information 31, the removal depth information calculation means 32 calculates the removal depth information 31. A correction map output means 33 for correcting the depth map 8b, generating a depth map 8c in which the erase region 26 is corrected, and outputting the depth map 8c is mainly configured. Here, the noise region 10 included in the fundus image data 6 is displayed in a three-dimensional manner from a two-dimensional fundus image mainly using a blood vessel region that travels along the fundus 5 or a stereo fundus camera 16, for example. Therefore, noise or the like in image processing that occurs when acquiring the depth information 13 or the like for this purpose is assumed. Here, the fundus image data 6 of the present embodiment corresponds to the image data of the present invention.

  As another configuration of the image analysis computer 2, the acquired fundus image data 6 and the calculated depth curve 8a and the like, such as the depth map 9a, are displayed on the display means of the liquid crystal display 17 so that they can be visually recognized. A signal sent from the operation input device 18 that is operated to input data or the like to the display control means 34 that controls the signal and the image analysis computer 2 or to input an instruction related to the analysis processing is received. Operation control means 35 is provided. In addition, the extraction area data 36 relating to the extraction area 7 extracted including the optic disc recessed area 4 and the noise area data 37 relating to the identified noise area 10 are stored in the data storage means 19. Further, the depth curve data 38 related to the depth curve 8 a calculated by the profile curve calculation means 21, the noise curve data 39 related to the noise curve 11 specified by the noise region specifying means 28, and the curve correction means 30 are calculated. The correction curve data 40 related to the corrected curve 12 is stored and stored together in the profile data 41 in the data storage means 19. Further, map data 42 relating to the created and corrected depth map 9 a and the like is also stored in the data storage means 19. Furthermore, the center point data 24a relating to the determined center point C, the contour portion data 24b relating to the identified contour portion 15, the estimated removal depth information 31, and the corrected correction map data 45 are also stored in the data storage means 19. Remembered. The data and the like are appropriately stored in the respective steps of the flowcharts shown in FIGS. 6 and 7 shown below. For the sake of simplification of description, unless otherwise specified, the storage / storing processing related to these is described. Description is omitted. Further, the progress in each process shown in FIGS. 3 to 5 can be appropriately displayed through the display control means 34 or the like.

  Next, the flow of processing of the image analysis computer 2 in the image analysis system 1 of the present embodiment will be described based on the flowcharts of FIGS. 6 and 7 in particular. Here, it is assumed that the image analysis computer 2 acquires the fundus image data 6 including the depth information 13 that can display the fundus 5 in a three-dimensional manner by a known technique and stores it in the data storage unit 19 in advance. Here, Steps S1 to S18 in FIGS. 6 and 7 correspond to the image analysis program of the present invention.

  First, the image analysis computer 2 reads the fundus image data 6 to be analyzed from the data storage unit 19 described above and acquires it (step S1, see FIG. 2). Then, from the read fundus image data 6, the optic nerve area 3 corresponding to the optic nerve head and the optic nerve cavity corresponding to the optic nerve crevice recess inside the optic nerve head are utilized by using the difference in pixel value of each pixel constituting the fundus image. The nipple depression region 4 is extracted (step S2). Here, the extraction process of the extraction area 7 including the optic disc area 3 and the optic disc depression area 4 is performed by using a well-known image processing technique. It is assumed that the information data acquisition means 20 includes the extraction function.

  Then, the depth of the fundus 5 is expressed in a one-dimensional distribution based on the depth information 13 of the fundus image data 6 for the extracted region 7 including the extracted optic disc region 3 and the optic disc recessed region 4. A depth curve 8a is calculated (step S3: see FIG. 3B)).

  Here, the calculation of the depth curve 8a will be described in detail. A virtual line segment L (see FIG. 2) substantially corresponding to the diameter of the optic disc region 3 is arranged on the horizontal axis, and each point on the virtual line segment L is arranged. The depth value is plotted on the vertical axis. A depth curve 8a that changes in a curved manner is calculated by connecting the plurality of plotted points. Here, FIG. 3A is a schematic cross-sectional view of the extraction region 7 viewed from the side.

  Furthermore, based on the acquired fundus image data 6 and the depth information 13 included in the acquired fundus image data 6, a depth map 9a used to display the depth distribution in the fundus 5 is created (step S4: see FIG. 3C). . Here, the depth map 9a visually displays the distribution of the depth by color shading or a plurality of different colors. In the present embodiment, as shown in FIG. 3C, the deep portion of the fundus 5 is expressed by a dark color, while the shallow portion is expressed by a light color, and the middle is dark. The example is expressed so as to change gradually from light to light color.

  Then, based on the created depth map 9a and depth information 13, the center point C of the extracted extraction region 7 is determined (step S5). Here, the center point C determines one point that is most darkly expressed in the created depth map 9a. The optic papilla region 3 is composed of a curved surface that curves in a concave shape, and the above-described optic papilla depressed region 4 exists in the deepest part of the fundus 5. Therefore, the determined center point C exists in the optic disc recessed region 4. The center point C is not determined based solely on the depth information 13, but is determined by comprehensively considering information relating to the point having the smallest depth value in the created depth curve 8a. It may be.

  Further, the contour portion 15 of the optic nerve head region 3 is specified (step S6). Then, the points on the depth curve 8a included in the specified contour portion 15 are set as the starting points S1 and S2 (see FIG. 4B).

  Then, the gradient change in the depth direction of the depth curve 8a toward the center point C determined from the starting points S1 and S2 (see the arrow in FIG. 4B) is detected (step S7). Here, since the optic disc region 3 is formed to be concavely curved, the gradient change in the extraction region 7 also has a smooth curved shape from the starting points S1 and S2 toward the center point C, usually obliquely downward. Should change. However, there is a possibility that there is a portion where the depth curve 8a does not change smoothly as described above due to a blood vessel region 43 such as a blood vessel traveling on the fundus 5 or a portion 44 caused by noise generated when the depth information 13 is acquired. Yes (FIG. 3C). Therefore, the gradient change is detected as described above.

  Here, the gradient change of the depth curve 8a from the starting points S1 and S2 toward the center point C is larger than the reference change rate set in advance, or the gradient change is in the reverse direction (the gradient change point in FIG. 3B). (Refer to E1 and E2) (YES in step S8), noise is detected in the range on the depth curve 8a of the change start point T1 at which the corresponding gradient change starts and the change end point T2 at which the gradient change returns to the original change rate again. It is specified as the curve 11 (step S9: see FIG. 4B)). Then, the area on the depth map 9a corresponding to the specified noise curve 11 is set as the noise area 10, the noise curve 11 is deleted from the depth curve 8a (see FIG. 5A), and from the depth map 9a. The noise region 10 is erased (step S10: see FIG. 4C). On the other hand, when a sudden gradient change or the like is not confirmed (NO in step S8), the processes related to step S9 and step S10 are cancelled.

  Thereafter, it is determined whether or not the detection of the gradient change started from the starting points S1 and S2 is completed up to the center point C (step S11), and the detection until reaching the center point C is not completed (step S11). NO), the process returns to step S7, and the detection of the gradient change in the depth curve 8a is continued.

  On the other hand, when the detection of the gradient change up to the center point C is completed (YES in step S11), the noise curve 11 is deleted and each end of the depth curve 8b divided into a plurality (see FIG. 5A). A correction curve 12 is calculated which connects P1, P1 ′ and P2, P2 ′ and substantially matches the curve change of the remaining depth curve 8b (step S12). Here, the calculated correction curve 12 is a known spline that passes through one end P1 (or P2) of one depth curve 8b and the other end P1 ′ (or P2 ′) of the other depth curve 8b. Calculated by obtaining a curve. Then, the divided depth curves 8b are connected to each other by the calculated correction curve 12 (see step S13: FIG. 5B). As a result, the depth curve 8c indicated by a single curve is obtained again by the smooth correction curve 12 that changes in accordance with the rate of change of the depth curve 8b. As a result, the depth curve 8a is corrected, a curve (noise curve 11) whose gradient change is remarkably different toward the center point C is removed, and the depth shown for the optic nerve head region 3 that changes in a concave curve. Curve 8c is obtained.

  Further, the depth information 13 is corrected using the corrected depth curve 8c (step S14). Thereafter, the virtual line segment L is set for all directions with the center point C as the reference position, and it is determined whether or not the correction of each depth curve 8a has been completed (step S15). Here, when the correction of the depth curve 8a for all directions has been completed (YES in step S15), that is, the imaginary line segment L is predetermined in the range of 0 ° to 180 ° with the center point C as the reference position. When the profile data 41 for each angle such as the depth curve data 38 that has been changed every predetermined angle (for example, 1 °) and corrected for each angle is acquired, the depth map 9b is corrected, Reconstruction is performed (step S16). On the other hand, if the correction of the depth curve 8a for all directions has not been completed (NO in step S15), the virtual line segment L is changed with reference to the center point C by a predetermined angle (step S17), and step The process returns to S3. Thereby, the correction of the depth curve 8a at the angle is performed by the same operation as described above. As a result, the erased region 26 is corrected based on the corrected depth curve 8c and the depth information 13 around the erased region 26, and a depth map 9c excluding the blood vessel region 43 and the like is obtained (FIG. 5 ( c)). Then, the depth map 9c is displayed (step S18).

  Accordingly, the depth curve 8a is corrected based on the gradient change of the depth curve 8a from the starting points S1 and S2 toward the center point C, and based on the corrected depth curve 8c obtained for all directions. The depth map 9c can be reconstructed.

  Therefore, useful information for diagnosing diseases such as glaucoma can be provided to doctors and the like. In particular, since the blood vessel region 43 and the like existing so as to overlap the optic disc region 3 and the optic disc recess region 4 can be erased, the cup line relating to the boundary of the optic disc recess region 4 can be accurately determined. Thus, the calculation accuracy of the C / D ratio is remarkably improved, and as a result, the accuracy of diagnosis by a doctor or the like is also improved. Furthermore, by using the corrected depth information 13, the accuracy between the actual fundus 5 and the obtained stereoscopic image is increased even when the fundus image data 6 is displayed in three dimensions.

  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 image analysis system 1 of the present embodiment, the virtual line segment L is set by a predetermined angle (for example, 1 degree) with the approximate center position (corresponding to the center point C) of the optic disc recessed area 4 as a reference position. Although the example in which the angle is changed and the depth curve 8a at each angle is corrected is illustrated, the present invention is not limited to this. For example, assuming the virtual line segment that crosses the extraction region 7 in the horizontal direction, the virtual line Minutes may be sequentially scanned in the horizontal direction from the upper end to the lower end of the extraction region 7 to correct the depth curve 8a at each scanning position. Further, the correction by all directions and the correction by scanning in the horizontal direction may be combined. Thereby, it becomes possible to correct the depth information 13 with high accuracy for the entire region of the fundus image data 6 obtained by photographing the fundus to be imaged, and the accuracy of C / D ratio calculation can be further improved.

  Further, as an example of the curve correction means 30 for correcting the depth curve 7a, a conventionally known spline curve has been shown. However, the present invention is not limited to this, and other techniques for correcting the depth curve 8a are available. It doesn't matter. For example, in the calculated depth curve 8a, when there is a portion where the gradient change changes significantly or the gradient slope changes in the opposite direction (for example, the direction in which the concave part protrudes), As shown in FIG. 8 (a), a line segment is extended in a substantially horizontal direction from the starting point X1 (or X2) of the gradient change on the depth curve 8a, and further, the end point Y1 of the gradient change intersecting the depth curve 8a ( Or specify X2). Then, the depth curve 8a may be corrected by line segments Q1 and Q2 (that is, straight lines) composed of linear curves connected by the start point X1 (X2) and the end point Y1 (Y2). That is, as shown in FIG. 8B, the corrected depth curve 8d may partially change in a step shape. Thereby, compared with what calculates a spline curve as mentioned above, the process which concerns on curve correction can be simplified. Further, the noise curve 11 and the noise region 10 are specified by the gradient change, and immediately after that, the depth curve 8a and the depth map 9a are deleted, but the detection by the gradient change is centered from the starting points S1 and S2. The process may be performed continuously up to point C, and thereafter, the noise curve 11 and the like may be erased collectively.

  Moreover, although the thing which mainly assumed the fundus 5 was shown as a target for correcting the depth information 13, it is not limited to this. For example, a mountain range or the like is made three-dimensional based on an aerial photograph taken from an airplane. It can also be used in accordance with various situations such as creating a 3D map to be displayed.

It is a block diagram which shows the functional structure of the image analysis computer in the image analysis system of this embodiment. It is a fundus photograph showing an example of fundus image data obtained by photographing the fundus including the optic disc area and the optic disc depression area. (A) A schematic diagram showing a section of an extraction region, (b) a curve diagram showing an example of a depth profile curve of the extraction region, and (c) a map diagram showing an example of a depth map of the extraction region. (A) A schematic diagram showing a section of an extraction region, (b) a curve diagram showing an example of an identified noise profile curve, and (c) a map diagram showing an example of a depth map from which the noise region has been removed. (A) a curve diagram showing an example of a depth profile curve from which a noise profile curve has been removed, (b) a curve diagram showing an example of a depth profile curve connected by a correction profile curve, and (c) a corrected depth It is a map figure which shows an example of a thickness map. It is a flowchart which shows the flow of a process of the image analysis computer which concerns on correction | amendment of a depth map. It is a flowchart which shows the flow of a process of the image analysis computer which concerns on correction | amendment of a depth map. (A) Curve diagram showing another example of correction of depth profile curve, (b) Curve diagram showing corrected depth file curve.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image analysis system 2 Image analysis computer 3 Optic nerve head area | region 4 Optic nerve disk depression area | region 5 Fundus (imaging object, target object)
6 Fundus image data (image data)
7 Extraction region 8a, 8b, 8c, 8d Depth curve (depth profile curve)
9a, 9b, 9c Depth map 10 Noise region 11 Noise curve (noise profile curve)
12 Correction curve (correction profile curve)
DESCRIPTION OF SYMBOLS 13 Depth information 15 Contour part 19 Data storage means 20 Information data acquisition means 21 Profile curve calculation means 22 Depth map creation means 23 Center point determination means 25 Contour part specification means 26 Erase area 27 Spline curve calculation means 28 Noise area specification means 28 29 Noise area elimination means 30 Curve correction means 31 Removal depth information 32 Removal depth information calculation means 33 Correction map output means 43 Blood vessel area C Center point E1, E2 Gradient change point L Virtual line segment S1, S2 Starting point

Claims (9)

An image analysis system that uses an image analysis computer to correct depth information calculated from image data of a concave or convex object that changes in a curve,
The image analysis computer
Information data acquisition means for acquiring the depth information from the image data;
Profile curve calculation means for calculating a depth profile curve expressing the depth of the object in a one-dimensional distribution based on the acquired depth information;
Center point determining means for determining a center point of the object based on at least one of the image data and the depth information;
Contour portion specifying means for specifying the contour portion of the object;
A point on the curve of the depth profile curve corresponding to the identified position of the contour portion is set as a starting point, and the depth is determined based on a gradient change in the depth direction of the depth profile curve from the starting point toward the center point. Noise area specifying means for specifying the noise area of the height profile curve;
Curve correcting means for correcting the depth profile curve corresponding to the identified noise region based on a change in the surrounding depth profile curve;
An image analysis system comprising: depth information correction means for correcting the depth information based on the corrected depth profile curve. The image analysis computer
Noise region erasing means for erasing the identified noise region from the depth profile curve,
The curve correcting means includes
Calculating a correction profile curve that substantially matches the rate of change of the depth profile curve remaining after the noise region is erased;
The depth information correcting means includes
The image analysis system according to claim 1, wherein the depth information is corrected based on the calculated correction profile curve. A depth map creating means for creating a depth map used for displaying the distribution of the depth of the object based on at least one of the acquired image data and the depth information;
Depth map correction means for correcting the depth map based on information obtained by the depth information correction means;
The image analysis system according to claim 1, further comprising a depth map display unit that displays the depth map corrected by the depth map correction unit. The noise region specifying means includes
At least one of the depth profiles in which the gradient change in the depth direction of the depth profile curve from the starting point toward the center point is larger than a preset change rate, and the gradient change changes in the opposite direction. The image analysis system according to claim 1, wherein a portion corresponding to a noise profile curve on a curve is specified as the noise region. The curve correcting means includes
5. The image analysis system according to claim 1, further comprising spline curve calculation means for calculating a spline curve as the correction profile curve. The image data is
The image analysis system according to any one of claims 1 to 5, wherein a fundus image acquired by using a fundus image capturing apparatus and capturing an image of the optic disc depression region is used.   Information data acquisition means for acquiring depth information calculated from image data of a concave or convex object that changes in a curve, and based on the acquired depth information, the depth of the object is one-dimensional A profile curve calculating means for calculating a depth profile curve expressed by a general distribution, a center point determining means for determining a center point of the object based on at least one of the image data and the depth information, Contour part specifying means for specifying the contour part of the object, the depth profile starting from a point on the curve of the depth profile curve corresponding to the position of the specified contour part, and going from the start point to the center point A noise region specifying means for specifying a noise region of the depth profile curve based on a gradient change in a depth direction of the curve, the depth corresponding to the specified noise region; Image analysis as a curve correction unit that corrects a profile curve based on changes in the surrounding depth profile curve, and a depth information correction unit that corrects the depth information based on the corrected depth profile curve An image analysis program characterized by causing a computer to function.   Noise area erasing means for erasing the identified noise area from the depth profile curve, the curve for calculating a correction profile curve that substantially matches the rate of change of the depth profile curve remaining after the noise area is erased 8. The image according to claim 7, wherein the image analysis computer further functions as a correction unit and the depth information correction unit that corrects the depth information based on the calculated correction profile curve. Analysis program.   Based on at least one of the acquired image data and depth information, a depth map creating means for creating a depth map used for displaying the distribution of the depth of the object, and the depth information correcting means The image analysis computer is used as a depth map correction unit that corrects the depth map based on the obtained information, and a depth map display unit that displays the depth map corrected by the depth map correction unit. The image analysis program according to claim 7 or 8, further comprising a function.

Images