WO2004075742A1

WO2004075742A1 - Hollow organ blood vessel extracting method, hollow organ blood vessel extraction processing program and image processing device

Info

Publication number: WO2004075742A1
Application number: PCT/JP2004/002369
Authority: WO
Inventors: Hideo Yokota; Satoko Takemoto; Taketoshi Mishima; Akitake Makinouchi; Ryutaro Himeno
Original assignee: Riken
Priority date: 2003-02-28
Filing date: 2004-02-27
Publication date: 2004-09-10
Also published as: JP4017157B2; JP2004261377A

Abstract

A frame-based characteristic region for extracting a blood vessel region is given to a hollow organ stereomodel prepared in advance, a region belonging to the above blood vessel region is judged according to a pixel-based judging criterion based on a pixel value in the characteristic region of a fist frame, and a judging criterion in a second frame continuing from the first frame is set in the above blood vessel region based on a pixel value in a region including the above judged region.

Description

Hollow organ blood vessel extraction method, hollow organ blood vessel extraction processing program and image processing apparatus

Technical field

The present invention relates to, for example, a hollow organ blood vessel extraction method, a hollow organ blood vessel extraction processing program, and an image processing apparatus.

Akita

Background art

Conventionally, in the case of hollow organs, for example, the eyeball, it is common practice to extract blood vessels of the fundus oculi having a three-dimensional structure in a two-dimensional space projected onto a plane.

Because the imaging range of the fundus camera is small, it is possible to obtain a panoramic photograph of the fundus by connecting the images taken from multiple directions from the center, up, down, left, and right. A fundus panoramic image obtained by developing the three-dimensional structure of the eyeball in a two-dimensional space is used as a criterion for physicians to determine a disease.

In addition, it has been practiced to automatically extract a fundus blood vessel from a fundus panoramic image serving as a criterion for the diagnosis and use it for diagnosis such as disease detection. Several products have been sold or research has been conducted on automatic extraction for joining the fundus panoramic images and extracting blood vessels in a two-dimensional space. For example, "Automatic analysis of blood vessel shape distribution in fundus photographs", Abstracts of the 21st Annual Meeting of the Japanese Society of Medical Imaging Technology (pp. 515-516), Nari Inui, "Fundus photography using color images" Detection of Lesions from ”(Information Processing Society of Japan, May 1993, λo 1.34, No. 5, pp. 873-8883).

However, since the panoramic photograph of the fundus obtained by joining is a two-dimensional projection of the three-dimensional information inside the eyeball, there is no continuity between adjacent images, and joining is difficult. there were. In addition, since the above fundus panoramic photograph has a shape in which the fundus is opened, images cannot be discriminated unless a doctor is proficient in the structure of the eyeball. It was difficult for patients to understand the images.

Furthermore, since the captured images in the three-dimensional space are projected two-dimensionally, there is no continuity between the captured images, and each captured image is distorted or centered. There was distortion around. Therefore, it was not possible to quantitatively measure a measurement object such as a blood vessel in the fundus and an ablated area.

For these reasons, it was difficult to extract blood vessels from the fundus panoramic image by image processing, and quantification could not be expected.

An object of the present invention is to provide a hollow organ blood vessel extraction method and a hollow organ blood vessel extraction processing program capable of quantitatively measuring a measurement target with respect to a hollow organ such as an eyeball and automating a mouth-fast blood vessel extraction. And an image processing apparatus. Disclosure of the invention

An object of the present invention is to solve at least the above-mentioned problems.

The hollow organ blood vessel extraction method of the present invention is a method for extracting a blood vessel region from a continuous frame of a cross-sectional image of a living body of a hollow organ, which is used to extract a blood vessel region from a three-dimensional model of a hollow organ prepared in advance. The region belonging to the blood vessel region is determined based on the pixel value of the measured region of the first frame according to the determination criteria for each pixel, and the determination is performed for the blood vessel region. A determination criterion in a second frame that is continuous with the first frame is set based on a pixel value of an area including the selected area. .

Further, the hollow organ blood vessel extraction method of the present invention is a method of extracting a blood vessel region from a continuous frame of a cross-sectional image of a living body of a hollow organ, wherein the blood vessel region is extracted from a prepared hollow organ stereo model. A framed region for each frame to be extracted is provided, and a region belonging to the blood vessel region and a region not belonging to the blood vessel region are determined based on the pixel value of the framed region of the first frame according to a determination criterion for each pixel, A criterion in the vascular region in the second frame that is continuous with the first frame is set based on the pixel value of the region including the region belonging to the vascular region, and the categorization is not performed outside the vascular region. A criterion for setting a determination outside the blood vessel region in a second frame continuous with the first frame is set based on a pixel value of a region including the region.

Further, a hollow organ blood vessel extraction program according to the present invention is a program for causing a computer to execute a process for extracting a blood vessel from a continuous frame of a biological organ cross-sectional image of a hollow organ. The first step of inputting the quality region of each frame for extracting the blood vessel region from the three-dimensional model of the organ, and based on the pixel value of the quality region of the first frame according to the criterion for each pixel A second step of determining a region belonging to the blood vessel region, and a determination criterion in a second frame continuous with the first frame based on a pixel value of a region including the determined region in the blood vessel region. The third step of setting and executing are performed.

Further, a hollow organ blood vessel extraction program according to the present invention is a program for causing a computer to execute a process for extracting a blood vessel from a continuous frame of a biological organ cross-sectional image of a hollow organ. The first step of inputting the quality region of each frame for extracting the blood vessel region from the three-dimensional model of the organ, and based on the pixel value of the quality region of the first frame according to the criterion for each pixel A second step of determining a region belonging to the blood vessel region and a region not belonging to the blood vessel region, and a second step which is continuous to the first frame based on a pixel value of the region including the belonging region in the blood vessel region. A determination criterion in the vascular region in two frames is set, and the determination is continuously performed in the first frame based on a pixel value of a region including the non-attributed region outside the vascular region. That a third step of setting a criterion extravascular region in the second frame, is characterized in that for the execution.

The image processing device according to the present invention is an image processing device, comprising: a storage unit that stores a three-dimensional model of a hollow organ; and a quality region for each frame for extracting a blood vessel region from the three-dimensional model. An input unit for inputting, a determination unit that determines a region belonging to the blood vessel region based on a pixel value of the quality region of the first frame according to a determination criterion for each pixel, and a determination unit that determines the region determined as the blood vessel region. Based on the pixel values of the included area And a setting unit for setting a criterion in a second frame that is continuous with the first frame.

The image processing device according to the present invention is an image processing device, comprising: a storage unit that stores a three-dimensional model of a hollow organ; and a quality region for each frame for extracting a blood vessel region from the three-dimensional model. An input unit for inputting, a determining unit for determining a region belonging to the blood vessel region and a region not belonging to the blood vessel region based on a pixel value of the quality region of the first frame according to a determination criterion for each pixel; The determination criteria in the blood vessel region in the second frame that is continuous with the first frame are set based on the pixel values of the region including the region belonging to the region, and the region not belonging to the blood vessel region is included outside the blood vessel region. And a setting unit for setting a criterion outside the blood vessel region in a second frame that is continuous with the first frame based on a pixel value of the region.

The foregoing and other objects, features, and advantages of the present invention will be apparent from the following detailed description of the invention. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a diagram showing a fundus photograph, FIG. 2 is a sectional view illustrating an optical system of an eyeball, FIG. 3 is a diagram illustrating coordinates of a fundus, and FIG. FIG. 5 is a diagram for explaining a case where the fundus is projected from a planar image, FIG. 5 is a diagram for explaining a template change in a projected image, and FIG. 6 is a diagram for a case where a fundus is photographed from different angles. FIG. 7 is a cross-sectional view illustrating an optical system, FIG. 7 is a cross-sectional view illustrating an optical system when the fundus is photographed from different angles, and FIG. 8 is a diagram illustrating a case in which a crystalline lens is considered. FIG. 9 is a diagram for explaining the construction of a fundus shape stereo model. FIG. 10 is a diagram showing a display example of a fundus shape stereo model. FIG. 11 is a diagram showing a fundus shape stereo model. FIG. 12 is a diagram illustrating the HSV color space, and FIG. 13 is a diagram illustrating the HSV color space. FIG. 14 is a diagram illustrating polar coordinates applied to the present invention. FIG. 14 is a diagram illustrating a region extracting method applied to the present invention. FIG. 15 is a diagram illustrating an area extracting method applied to the present invention. FIG. 16 shows a region extraction method applied to the present invention. 9

Five

FIG. 17 is a diagram illustrating a region extraction method applied to the present invention. FIG. 18 is a diagram illustrating a region of a blood vessel region in a three-dimensional space applied to the present invention. FIG. 19 is a diagram illustrating a glowing method. FIG. 19 is a drawing illustrating a configuration of an image processing apparatus according to Embodiment 1 of the present invention. FIG. FIG. 21 is a block diagram showing functions, FIG. 21 is a diagram showing an example of a display transition according to the first embodiment, and FIG. 22 is a flowchart for explaining an operation according to the first embodiment. FIG. 23 is a diagram for explaining the management of flag information according to the first embodiment. FIG. 24 is a diagram for explaining the local region of the second embodiment. FIG. FIG. 9 is a diagram illustrating extended region growing according to a second embodiment. BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The present invention is not limited by the embodiment.

Embodiment 1

First, the principle of the present invention will be described. An example of the hollow organ is the fundus. The fundus is spherical, as shown in the photograph shown in Fig. 1. Therefore, a method is needed that can reproduce fundus images placed in a three-dimensional space as three-dimensional information. As an example, the fundus can be improved by adopting the “3D display and coordinate measuring method and apparatus of the fundus” of Japanese Patent Application No. 2002-18925 already proposed by the present inventors. It can be reproduced as a three-dimensional shape. Thus, a stereoscopic model of the fundus based on the coordinate data is constructed. In the first embodiment, this method is adopted as an example. In particular, the invention also covers methods that have timed out in the case of full color.

To extract the fundus blood vessels from this three-dimensional model, it is necessary to perform extraction in a three-dimensional space. Therefore, there is an image extraction method from a three-dimensional space as a method suitable for this extraction. When the space in which the fundus is located is placed in the poxel space, the extraction technique is described in Japanese Patent Application No. 2003-255720, which is already proposed by the present inventors. Output method, tissue of interest region extraction program, and image processing device ”can be used. With this method, it is possible to extract the fundus blood vessels. In the first embodiment, this method is adopted as an example.

However, if only the above two methods are adopted, and if the direction of region expansion at the time of extraction is not determined when adapting to fundus blood vessels, an enormous number of repetitions of cutting directions will be required. Therefore, in the present invention, it is intended to realize the processing efficiency as well as the extraction of the blood vessel of the hollow organ.

For a given fundus stereo model, it is necessary to define the section and its area of the extended region growing method as follows. That is,

(1) Apart from the three-dimensional coordinate axes of the three-dimensional model of the fundus, the plane that touches the site at the point with the optic disc of the fundus as the reference is the two-dimensional reference plane that extends that plane. (2) The moving direction of the cross section in the stereoscopic model of the fundus moves toward the center of the eyeball. Here, the cross section refers to a layer composed of an arbitrary number of frames when the three-dimensional model is composed of a plurality of frames.

(3) The local area for obtaining the criterion for extracting blood vessels is a three-dimensional space of, for example, 7 × 7 × 4 (pixels). This local region is a part of a plurality of frames forming a three-dimensional model.

(4) The thickness direction of 4 in the local region 7 X 7 X 4 is set as the axis of the cutting direction, and this 4 is the thickness in the moving direction from the target pixel to be determined as the blood vessel region. In this case, 4 corresponds to the number of frames “4” because the cross section is 4 pixels. In the blood vessel extraction performed for each cross section (four frames) in the fundus three-dimensional model, the extended region glowing is performed by moving the local region in a plane.

(4) After the extraction of the three-dimensional model of the fundus in one axis direction is completed (from the optic papilla to the end of the model), the same extraction operation is repeated in the opposite direction according to the one axis direction to increase the accuracy.

First, a three-dimensional model of the fundus oculi prepared before extraction of the blood vessels of the fundus will be briefly described. FIG. 2 is a cross-sectional view illustrating the optical system of the eyeball, FIG. 3 is a view illustrating the coordinates of the fundus, Fig. 4 illustrates the case of projecting from the planar image to the fundus, Fig. 5 illustrates the change of template in the projected image, and Figs. 6 and 7 show the fundus from different angles. FIG. 8 is a cross-sectional view illustrating an optical system in the case where the above-described operation is performed, and FIG. 8 is a diagram illustrating a case where a crystalline lens is considered.

The three-dimensional fundus model is, for example,,

① Eyeball shape measurement

② Eyeball setting

③ fundus photography

④Setting eyeball parameters

⑤Image and position recording

⑥Paste image

⑦ Image alignment

Is obtained by The three-dimensional model obtained in this way can be displayed or saved as a file.

To briefly explain the above (1) to (4), first, in the eyeball shape measurement, the size L of the eyeball 1 and the shape R of the fundus 2 are measured by a measuring device (not shown) according to the imaging position 〇 (see FIG. 2). . The size L of the fundus 1 includes at least the depth from the corneal surface to the fundus 2. The shape R of the fundus 2 includes at least the curvature of the fundus. As the measuring device, for example, an ultrasonic sensor can be applied.

In the eyeball setting, a fundus template is set based on the measured size L and shape R. In this setting, as shown in Fig. 3, the coordinate map of the fundus 2 including the retina

Created three-dimensionally. In the fundus photography, a plurality of fundus 2 including the overlapping part are photographed by shifting the photographing position of the fundus 2 by a predetermined amount, as shown in the relationship of FIG. 6 and FIG. In this shooting, a plurality of images are shot from slightly shifted positions such as shooting positions Om and On.

In the setting of the eyeball parameter, the eyeball parameter indicating the positional relationship between the fundus 2 and the photographed image from the position of the overlapping portion on the photographed image based on the shift amount of the plurality of photographed images Is required. In the recording of the image 'position, a captured image obtained in fundus photography and a positional relationship between the eyeball 1 corresponding to the image and an imaging device (not shown) are stored.

In image pasting, as shown in FIG. 4, a plurality of photographed images are pasted on an eyeball template based on eyeball parameters. At this time, as shown in FIG. 5, it is preferable to change the template in the projected image to the eyeball template 3 set in the eyeball setting.

In the image registration, the position of the fundus image in the plane of the fundus 2 based on the corresponding point 7 in the overlapping portion described above (see FIG. 7), that is, the optic disc, blood vessels, nerves, etc. of the fundus 2 is determined. Done. In the three-dimensional image display, a three-dimensional fundus image on the eyeball template 3 is displayed on a display device (not shown).

In addition, in the above ① to ⑦, as shown in FIG. 8, when considering the crystalline lens, it is necessary to consider the shape of the lens.

Here, the stereoscopic model of the fundus will be described. FIG. 9 is a diagram for explaining the construction of a three-dimensional fundus shape model, FIG. 10 is a diagram showing a display example of a three-dimensional fundus shape model, and FIG. 11 is a display example of a three-dimensional fundus model. FIG.

As shown in FIG. 9, the three-dimensional fundus shape model forms a vertical circular arc and a horizontal circular arc with a radius R on the XY Z axis, with the axial length on the X axis. The stereoscopic model of the fundus shape is displayed, for example, as shown in FIG. 10, and the stereoscopic model of the fundus is displayed only with an actual fundus image, for example, as shown in FIG. Therefore, this three-dimensional fundus model has the minimum necessary coordinate information for extracting blood vessels, that is, information on the measurement area.

Next, extraction of blood vessels from the fundus will be described. First, the principle of automatic region extraction from full-color continuous cross-sectional images will be described. FIG. 12 is a diagram illustrating the HSV color space, FIG. 13 is a diagram illustrating polar coordinates applied to the present invention, FIGS. 14, 15, 16, and 17. Fig. 18 is a diagram for explaining a region extracting method applied to the present invention, and Fig. 18 is a diagram for explaining a region growing method for a blood vessel region in a three-dimensional space applied to the present invention. A full color image has three color vector values for each pixel in the RGB color system. In each two-dimensional image, in order to extract a blood vessel region, it is necessary to determine a boundary of a blood vessel fibrous tissue. To do so, we must recognize the slight color differences that occur at the boundaries between vascular tissues. In other words, attention is paid to minute color changes between neighboring pixels.

When extracting a blood vessel region, a method of applying a threshold value to the pixel value can be used.However, there is little obvious color change near the tissue boundary in a living body, and a complex color distribution Therefore, it is difficult to extract a blood vessel region with a unique threshold.

Therefore, in the present invention, the following two points are taken into the extraction method.

(1) HSV (H: Hue, S: Color, V: Lightness) Calculation of density difference between pixels using pixel values expressed by the color specification method

(2) Rather than applying the same threshold value to the entire image, the inside / outside discrimination of the vascular region focusing on the local region (extended region growing method)

There are two reasons for using the HSV colorimetric method.

(1) Points that make it easier to distinguish the boundaries of vascular tissue than the RGB colorimetric method

(2) In the vascular tissue, the H value and S value do not show a large color change, and the values are unified.

Now, the ability to perform segmentation using pixel values represented by the HSV colorimetric method, in particular, segmentation is performed using H and S values. However, since the H, S, and V vector values represented by the HSV colorimetric method are not equivalent to each other and create a special space as shown in Fig. 12, the density difference between pixels is calculated. The correct value at normal Euclidean distance! / ,.

Therefore, when calculating the pixel-to-pixel density difference between two points, as a calculation method that does not impair the features of the HSV space (see FIG. 13), the distance on polar coordinates by H and S was calculated. As an example, the pixel values of pixels P 1 and P 2

_{fp, = (,, v Pl} ), f P 2 = (h P 2, S P ,, V P 2 j Then, the density difference << 5 between the pixels P l and P 2 is calculated by the following equations (1), (2),.

_{= Cos (2TT.h Pi) - s} n cos (2jrfip 2 · ■ · (1) y 2 = (s Pi 3Ϊη (2 Η Ρι) - s Pz sin (27ihp 2 f ■■■ (2) δ (Ρ I

(3) The present invention is to apply the above-described method of obtaining the density difference between pixels to a three-dimensional local region to determine whether or not a force is attributed to a blood vessel region. Therefore, a brief description will be given with reference to FIGS. 14 to 18.

First, it is necessary to set the cutting coordinates for the three-dimensional fundus model. This is to set the slice direction (cutting direction) for the three-dimensional fundus model. This slice direction corresponds to the axial direction of the eye so as to pass through the center of the eyeball.

FIG. 14 shows a three-dimensional model similar to that of FIG. In the three-dimensional fundus model, a slice plane is shown as an example. The number of frames between the cross sections formed by the slices is in accordance with the above-described rule, and is constituted by a part of a plurality of frames, and matches the thickness of the local region in the slice direction. When an initial image (frame N-1) is selected from a series of images of a living body (a continuous frame in a stereoscopic model of the fundus), for example, the optic disc S in a blood vessel region having a size that can be visually recognized to some extent is captured. (See Figure 15 (A)). In this case, the extraction may be performed manually or may be performed by computer processing. FIG. 15 (B) shows an instrumental region T and a blood vessel U in the fundus V of each frame. Here, the blood vessel U to be extracted is included in the quality region T.

The extraction of the blood vessel region may be performed by reciprocating in a uniaxial direction with respect to the stereoscopic model of the fundus. In this case, as shown in FIG. 16, only the blood vessel region is extracted from the three-dimensional model, and the blood vessel extending in the slice direction is branched. Understand. Therefore, the accuracy of blood vessel extraction can be improved by reciprocating the blood vessel region extraction work.

When the optic disc is set in the image of frame N-1, the region is the vascular region R O I. (Fig. 17 (A)). Then, in the continuous cross-sectional image of the living body, an area at the exact same position as the area extracted in the image of the upper frame N-1 with respect to the image of the next continuous frame N is first selected as the temporary area S (No. 1 7 Figure (B

)

Then, the correct region, that is, the actual blood vessel region R〇 is re-extracted from the tentative state by the extended region growing method of the present invention described later (see FIG. 17 (C)). Thus, the extraction of the blood vessel region R O is completed (see FIG. 17 (D)). The extended region growing method is a blood vessel region R O I extracted from the image of frame N-1. The blood vessel region RO to be extracted is extracted from the image of frame N using the density values based on the !! value and S value, and the next successive I / one N + 1, N + 2, N + 3 · · · · is a method to extract at least from the density values inside and outside the blood vessel region in the past frame.

That is, in each frame, a region at the same position as the blood vessel region extracted in the previous frame is first given as a temporary region, and each boundary pixel of the temporary region is focused on one by one. A process is performed to determine whether the change has occurred inside or outside of the area.

In FIG. 18, attention is first focused on a local region of 7 _× 7 _× 4 pixels centered on a blood vessel boundary pixel Px. In this case, the Z direction is indicated by N, N-1, N-2, and N-3, but the processing direction of blood vessel region extraction is downward in the figure. During its local region, H values of the pixels that are determined to vascular region up to this point, the concentration median S value (median value) to calculate the M _in. At the same time, the median value M of the H value and S value of the pixel determined to be outside the blood vessel region up to this point. _{ut is} also calculated. … (Four )

Pixel value of the pixel P _x determines whether studied vascular region or outside close to either aspect of the blood vessel region and outside in the local region. In fact, H values of the inner and outer vessels regions, respectively calculated the density difference d between the concentration representative value M and the pixel P _x by S value is regarded as characteristic of the person and the pixel P _x difference was small similar . Thus, taking a P _x vascular region內, either extravascular region. The following expressions (5) to (6) express this by a calculation expression. d _in (= δ (Ρ _0? Μ _ίη ))

… (5) d _out (= 5 (P ₀ , M _out ))

… (6)

The above processing is performed on all the boundary pixels of the blood vessel region. This process is automatically repeated until the capture into and out of the blood vessel region is completed. Completion of the capture at this time means that the setting of the correct blood vessel region boundary is completed by repeating the capture or the removal of the pixel into the blood vessel region.

In this manner, after the blood vessel extraction processing for the image of frame N is completed, the same processing is performed on the images of frames N + 1, N + 2,. Execute

The configuration and operation will be described below based on the above principle. First, the configuration will be described. FIG. 19 is a block diagram showing a configuration of an image processing apparatus according to an embodiment of the present invention. The image processing apparatus in FIG. 19 constitutes the image processing apparatus according to the present invention, and is connected to the image processing apparatus in FIG. 19 through a recording medium such as a floppy (registered trademark) disk, a CD, a DVD, or a communication line. The supplied program shall include the hollow H organ blood vessel extraction program according to the present invention.

The image processing apparatus according to the present embodiment has an internal path as shown in FIG. 19, for example. Connect the communication interface 12, the CPU 13, the ROM 14, the RAMI 5, the display 16, the keyboard mouse 17, the drive 18, and the hard disk 19 to the interface 11 to transmit address signals, control signals, data, etc. The configuration for realizing the blood vessel extraction according to the embodiment is provided.

In FIG. 19, the communication interface 12 has a function of connecting to a communication network such as the Internet, for example, and the hollow organ blood vessel extraction program according to the present invention can be downloaded. The CPU 13 controls the entire apparatus by using 0 S stored in 10; 14, and has a function of executing processing based on various application programs stored in the hard disk 19.

The ROM 14 stores programs for controlling the entire apparatus, such as OS, etc., and has a function of supplying these to the CPU 13. The RAMI 5 has a memory function used as a work area when the CPU 13 executes various programs. In the RAMI 5, for example, stacks A and B are set at the time of region extraction.

The display 16 has a function of displaying menus, statuses, display transitions, and the like associated with various processes of the CPU 13. The keyboard / mouse 17 includes keys for inputting data such as characters, numbers, and symbols, cursor keys for operating point positions on the screen, and function keys. A mouse is provided to control the point position.

The drive 18 is a drive unit for executing installation work from a recording medium such as a CD or a DVD in which various programs and data are recorded. Note that a floppy (registered trademark) disk drive function may be added.

The hard disk 19 is an external storage device that stores a program 19A, a memory 19B, a three-dimensional fundus model 19C, flag information 19D, and the like. The program 19A is equivalent to a program stored from the communication interface 12, the drive 18, and the like described above in an executable form. The memory 19B is a storage unit for storing files such as execution results of various programs. The fundus three-dimensional model 19C is information constructed by the method described above, and is a data file read through the communication interface 12, the drive 18 and the like. The flag information 19D represents, for example, RGB with one flag in correspondence with each frame of the fundus stereoscopic model 19C by coordinates.

Subsequently, the present embodiment will be functionally described. FIG. 20 is a block diagram illustrating functions according to the present embodiment, and FIG. 21 is a diagram illustrating an example of display transition according to the present embodiment. The functions of the present invention include an extension condition setting unit 100, an initial blood vessel region setting unit 101, a temporary region setting unit 102, a median density value calculation unit 103, a pixel value acquisition unit 104 near the boundary, It is composed of a region inside / outside determination unit 105, a blood vessel region extraction unit 106, a display image formation unit 107, and the like.

According to the above configuration, the condition of the extended region growing is set by the extension condition setting unit 100 so as to expand each section. The image of the frame N-1 is extracted from the fundus three-dimensional model 19C by the initial blood vessel region setting unit 101, and the initial blood vessel region (optic nerve head) is set by manual or computer processing. The median density calculating unit 103 calculates the median density from the density values of each pixel in the area set by the blood vessel area setting unit 101.

The temporary area setting unit 102 sets a temporary area at the same position as the blood vessel area set by the initial blood vessel area setting unit 101 with respect to the next frame N following the frame N-1. Then, the pixel value of the pixel near the boundary inside and outside the temporary area is obtained by the pixel value obtaining unit 104 near the boundary. However, under the control of the extension condition setting unit 100, the processing shifts to the next successive frame after the extension region growing on the same cross section is completed. Therefore, in the same cross section, the blood vessel region is determined while shifting the local region sideways. However, the measurement area is the target of blood vessel extraction, and the area outside that area is excluded as a processing target.

The region inside / outside determination unit 105 determines the inside / outside of the blood vessel region by the extended region gliding method based on the pixel near the boundary, the pixel value, and the density center calculated by the density median value calculation unit 103 as described above. Is performed. Note that the local area of 7 X 7 X 4 The setting is started after the extraction of blood vessels for 4 frames in the past is completed. Therefore, if the extraction start frame is N, a local region from frame N + 3 to 7 × 7 × 4 is adopted.

Then, the vascular region extracting unit 106 performs extraction by expanding or expanding / contracting the vascular region based on the inside / outside determination of the vascular region, and the result of the extraction is reflected in the flag information 19D.

In the next frame N + 1 which is interrupted, the three-dimensional local region by frames N-2, N-1 and N is a target when the median density calculation unit 103 calculates the median density. . At this time, the temporary area setting unit 102 captures the blood vessel area of the frame N from the flag information 19D, and sets the temporary area according to the blood vessel area. Thereafter, similarly, blood vessel extraction is performed by the extended region glowing method.

When the extracted blood vessel region is displayed, display data is generated by the display image forming unit 107 based on the flag information 19D and the biological color image information 19C. As shown in (A) to (F), the display transition of the vascular region ROI can be observed.

Next, the operation will be briefly described. FIG. 22 is a flowchart illustrating the operation according to the present embodiment, and FIG. 23 is a diagram illustrating management of flag information according to the present embodiment. The processing shown in FIG. 22 corresponds to the hollow organ blood vessel extraction program according to the present invention in the program 19A. However, the display follows the functional blocks shown in FIG. The extended region growing performed in the following process shall be performed, for example, in units of four frames corresponding to the thickness between the cross sections.

First, the first frame image is input (step S I), and the HSV conversion process is executed (step S 2). Then, the starting point of the blood vessel region to be extracted (optical nipple, etc.) is specified (step S3).

At this stage, since the next frame image exists (step S4), a continuous next frame image is input (step S5). At this time, the input frame The flag information corresponding to the camera image is reset (step S6).

For the input frame image, an extraction start region is specified as a temporary region at the same position as the blood vessel region specified in the previous frame image (step S7). One pixel near the boundary of the extraction start area is selected (step S8) and stored in the stack A (see FIG. 19) (step S9).

At this stage, since the stack A is not empty (step S10), one pixel of interest P is extracted from the power of the stack A (step S11), and the above-described extended region glowing processing using the local region is executed. (Step S12).

If the target pixel P is in the blood vessel region (step S13), a flag is set for the target image P (step S14), and pixels outside the extraction start region near the target pixel P are set. Is selected (step S15). Thereafter, the selected pixel is stored in the stack A (step S9), and the same processing is repeated. The flag information is managed by coordinates for each pixel for each frame, as shown in FIG. For example, when the coordinates of frame No. 100 (X / YX is determined to be inside the blood vessel region, the flag is set to “1.” The coordinates of frame No. 101 (Χ ₂ , Υ ₂ , ₂ ) Is determined to be within the blood vessel region because of the flag "1".

If the pixel of interest 外 is outside the region in step S13, one pixel in the blood vessel region is selected near the pixel of interest Ρ (step S16), and the selected pixel is stored in the stack Β (step S13). 17). Then, the empty state of the stack Α is confirmed (step S18). The processing related to the stack A targets the inside of the blood vessel region.

If the stack A is not empty (NO in step S18), the process returns to step S11, and the same process is repeated. On the other hand, if the stack A is empty (YES route in step S18), the empty state of the stack B is further confirmed (step S19).

If the stack B is not empty (NO route in step S19), the processing is stopped. If the stack B is empty (YES route in step S19), the processing returns to step S4 because the processing for one input frame image is completed. It is determined whether there is a next frame image.

In step S20, one pixel is taken out of the stack B as the pixel of interest P, and the above-mentioned extended region glowing process is performed on the pixel of interest P (step S21). If the pixel of interest P is outside the blood vessel region (step S22), one pixel in the blood vessel region is selected near the pixel of interest P to ask for further judgment (step S22). 3) The selected pixel is stored in the stack B (step S17), and the same processing is repeated thereafter. Since the processing relating to the stack B is processing outside the blood vessel region, if it is determined in step S22 that the pixel of interest is within the blood vessel region, the processing returns to step S19.

As described above, according to the first embodiment, it is possible to quantitatively measure a measurement target with respect to a hollow organ such as an eyeball and to automate mouth-past blood vessel extraction. In particular, the extended region glowing uses the three-dimensional local space around the pixel of interest, and uses the median density inside and outside the blood vessel region in the three-dimensional local space around the pixel of interest to extract the area. In addition, it is possible to take into account the determination results of several consecutive frames before the frame to be extracted. As a result, it is possible to realize automatic segmentation by extended region glowing.

In particular, the starting point of the cut plane is the circumscribed surface of the optic disc (where the optic nerve and vascular force S enters the eyeball from outside the eyeball) where the blood vessels are the thickest and all blood vessels are connected. High effects can be obtained. This makes it possible to reduce the number of inversions in one axis direction of the blood vessel extraction operation.

At this time, it is assumed that the criterion for determining the blood vessel region of the hollow organ is a three-dimensional space, and the characteristics of the hollow organ, that is, the eyeball is transparent at the fundus, can specify the blood vessel extraction range. By specifying the extraction range in this way, it is possible to realize robustness and high-speed blood vessel extraction. In this way, for example, the region can be accurately extracted in response to a change in the color of the same organ, and the region can be accurately extracted without being affected even if foreign substances are mixed in the same organ. . In addition, it is possible to eliminate the condition for stopping the arbitrary parameter と expansion, which was a problem in the conventional region growing, and to avoid fixing the criterion. Further, the present invention is a method for automatically extracting blood vessels in a three-dimensional fundus model. According to this method, it is possible to accurately extract a fundus blood vessel with high accuracy, and to reduce the calculation cost. It becomes possible to reduce. This makes it possible to extract the blood vessels from the patient-specific fundus model from fundus photographs and ultrasound images currently taken in ophthalmology.

Since the three-dimensional coordinates of the extracted blood vessels can be obtained from the three-dimensional fundus model, it can be obtained as digital data. Using this blood vessel extraction region as a criterion, it becomes possible to provide a diagnostic assistance system that detects abnormalities in the blood vessels of the fundus oculi and informs a doctor.

Also, by recording the change over time of the same patient, it becomes possible to present abnormal values of the patient as numerical data. Its application scope is not limited to diseases in the ophthalmology field, but also extends to the circulatory and medical fields related to blood vessels, as well as public health, because the fundus blood vessels are the only blood vessels that can be seen in the human body. Furthermore, it is also possible to present data based on analysis to diagnosis and treatment systems using simulation technology that has been studied in recent years.

In Embodiment 1 described above, the number of times of repetition of extraction (reciprocation in one axis direction) may be set arbitrarily, or the amount of extension or the amount of extension (operation in one axis direction) may be set. Until the end).

It is also possible to prevent a small area from being extracted by using the expansion amount of the local area as a criterion. This makes it possible to control the extraction of fine blood vessels by the vascular system. By developing this and setting multiple types of flags according to the amount of expansion, it is also possible to use a flag that identifies thick blood vessels and thin blood vessels. Further, according to the first embodiment, although the target blood vessel is in contact with the fundus oculi, the target area does not exist at the center or outside of the fundus oculi model. Therefore, by defining the extended range from the fundus model, the amount of calculation can be drastically reduced.

According to the first embodiment, after extracting the details of the blood vessel region, the center line of the blood vessel is calculated, the cross section orthogonal to the axis is obtained, and the thickness of the blood vessel is determined by the size of the cross section. Is also possible. In this method, since the extraction is performed based on the cut plane (for example, 2.5 dimensions), the memory space can be reduced. By the way, in Embodiment 1 described above, the median density is adopted as the data of the criterion. However, the present invention is not limited to this, and the same effect is obtained from the power density average value where the median density is more preferable. It is possible to obtain

Also, in Embodiment 1 described above, the three-dimensional local region is described as having a size of 7 × 7 × 4, but the present invention is not limited to this, and for example, 3 × 3 × 2 It may be smaller or larger than the above size. By having multiple sizes such as large, medium and small, it is also possible to weight each one. Furthermore, in the first embodiment described above, a color image is taken as an example, but application to a black-and-white image is also possible.

In Embodiment 1 described above, the blood vessel region is manually extracted from the frame N-1 corresponding to the initial image. However, the present invention is not limited to this. May be extracted.

Embodiment 2

Now, the present invention is not limited to Embodiment 1 described above, and various modifications are possible. Therefore, a second embodiment of the present invention will be described below. Note that the principle is the same as that of the above-described first embodiment, and since there is a common function, only the differences will be described below.

In the second embodiment, the optic disc is the most effective as the starting point. However, a plurality of optic discs are designated, and the overlapping thereof is used to improve the mouth-paste property. It becomes possible. It is possible to improve the certainty by having the data of the extension area for each start point and then judging the data for each junction point by a statistical method (such as and).

In the second embodiment, the work of extracting the blood vessel region from the stereoscopic model of the fundus is not performed on a cross-sectional unit basis with respect to the incision cross-section, and the size of the local region is also arbitrarily determined. An example of the local region is shown in FIG. The local area W shown in FIG. 24 has a size of, for example, 7 × 7 × 7 pixels. As described above, the extended region growing according to the present invention can be performed in an arbitrary direction in the poxel space assigned to each pixel.

In the first embodiment described above, the extended region growing is performed in units of cross sections in the processing of FIG. 22. However, in the second embodiment, there is no unit between the cross sections, and the extended region growing is performed. Will cover the entire data of the solid model ^^.

As shown in FIG. 25, the local area W can move the determination place while expanding the blood vessel area. Therefore, the extended region growing helps the blood vessel extraction to travel along the blood vessel according to the given local region W of any size. Note that, also in the second embodiment, similarly to the above-described embodiment, the target region is a target of blood vessel extraction, and the outside of the target region is excluded as a target to be processed.

In the above, the starting point for extracting the blood vessel region may be any one or a plurality of points. In this case, as described above, the extraction processing for each pixel is separated by flag management, so that even if the tracking of the blood vessel region is divided into a plurality of branches, the processing can proceed simultaneously. Become.

Further, the extended region may be a three-dimensional space in which the size in one axis direction is removed from the force or extended region glowing method which is a three-dimensional space. In other words, the local area is held in a three-dimensional area taking 7 × 7 × 4 as an example, but the local area is not limited to 7 × 7 × 4 and may be of any size.

For the extended area, the fundus blood vessel area is obtained in advance from the three-dimensional fundus model. At this time, the thickness may be defined by the thickness from the fundus, or the portion exceeding the thickness range may not be calculated. By not calculating the part that exceeds the thickness range, the amount of calculation can be drastically reduced.

In addition, it is possible to set a flag separately from the image data set (image file) for the region where the extraction of the blood vessel region is completed. In this case, it is possible to set a flag with a value specified in the image area.

When a plurality of starting points are set, a potential can be given to each starting point, and a blood vessel portion can be determined from information of a plurality of extended starting points.

Then, it is possible to consider the search area when extracting the blood vessel area as a three-dimensional space. In the voxel space of 100 × 100 × 100 × which is usually assumed as a fundus region, the target region is 100 billion voxels, and the computation cost is enormous. Therefore, it is possible to greatly reduce the calculation cost by limiting the calculation area from the shape of the fundus model to the bowl-shaped fundus shape and its thickness in advance.

If the direction of expansion is not specified as in the above-described modification, it is not necessary to move the cut surface with respect to the three-dimensional model, and it is not affected by the blood vessel shape such as returning to a U-shape. There is an advantage that it is not necessary to consider the number of times. In the above description, the fundus has been described as an example, but the present invention is not limited to this, and can be applied to hollow organs such as the stomach. In addition, it can be applied not only to the fundus blood vessels, but also to a method for extracting unique parts (blood vessels, tumors) by combining images with shape information and images from an endoscope. In addition, the present invention is not limited to extraction of blood vessels of hollow organs, but is also applicable to extraction, detection, and diagnosis of tumors and the like on the front surface (back surface) of hollow organs.

Further, the present invention is not limited to the stack described above, and it goes without saying that a queue may be applied as a data holding method. In this case, the data structure is a fast-in first-out data structure, and the region can be extended along the flow of blood vessels. Optimization of data management such as prevention of data volume swelling Both stacks and queues have the effect.

INDUSTRIAL APPLICABILITY As described above, according to the present invention, a hollow organ blood vessel extraction method capable of quantitatively measuring a measurement target with respect to a hollow organ such as an eyeball and automating mouth-bust blood vessel extraction, It is possible to provide a blood vessel extraction processing program and an image processing device.

Further, according to the present invention, the criterion of the blood vessel region of the hollow organ is determined in a three-dimensional space, and the extraction range of the blood vessel is specified based on the characteristics of the hollow organ. It is possible to provide a hollow organ blood vessel extraction method, a hollow organ blood vessel extraction processing program, and an image processing device capable of realizing the characteristics. Industrial applicability

INDUSTRIAL APPLICABILITY As described above, the hollow organ blood vessel extraction method, the hollow organ blood vessel extraction processing program, and the image processing apparatus according to the present invention are useful for quantitatively measuring a hollow organ measurement target. Suitable for extraction, robust extraction, detection, and diagnosis of tumors on the surface (back side) of hollow organs.

Claims

The scope of the claims

1. A method for extracting a blood vessel region from a continuous frame of a biological organ cross-sectional image of a hollow organ,

An instrument region for each frame for extracting a blood vessel region is provided to a previously prepared three-dimensional model of a hollow organ, and the blood vessel is extracted based on a pixel value of the instrument region of the first frame in accordance with a criterion for each pixel. Determining a region belonging to a region, and setting a determination criterion in a second frame connected to the first frame based on a pixel value of a region including the determined region in the blood vessel region. Hollow organ blood vessel extraction method.

2. A method for extracting a blood vessel region from a continuous frame of a biological organ cross-sectional image of a hollow organ,

Given a stereoscopic model of a hollow organ prepared in advance, an instrumental area for each frame for extracting a blood vessel area is given, and based on the pixel value of the instrumental area of the first frame according to the criterion for each pixel, Do not belong to the region belonging to the blood vessel region! And determining a criterion in a vascular region in a second frame that is continuous with the first frame based on pixel values of a region including the belonging region in the vascular region. A method for extracting hollow organ blood vessels, comprising: setting a criterion for determining whether a blood vessel region is outside a blood vessel region in a second frame that is continuous with the first frame, based on pixel values of a region including the non-attributed region.

3. At least one vascular region is given to the quality region of the first frame from which the extraction of the vascular region is started, and an initial determination criterion is set based on the pixel value of the vascular region. The hollow organ blood vessel extraction method according to claim 1 or 2, wherein the method is characterized in that:

4. In the current frame from which the blood vessel region is extracted, a temporary region is provided at the same position as the blood vessel region determined up to the previous frame, and the attribution is determined in pixel units along the boundary of the temporary region. 3. The method for extracting hollow organ blood vessels according to claim 1 or 2, wherein:

5. In the temporary region, when a pixel belonging to the blood vessel region is outside the temporary region, the pixel is regarded as a pixel of the blood vessel region of the current frame, and the blood vessel region is determined with respect to the blood vessel region of the previous frame. If a pixel that does not belong to the blood vessel region is within the temporary region, the pixel is regarded as a pixel outside the blood vessel region of the current frame, and the blood vessel region is compared with the blood vessel region of the previous frame. The hollow organ blood vessel extraction method according to claim 4, wherein the method is characterized by reducing the size.

6. In the determination of the blood vessel region, a three-dimensional local region of a predetermined size including each pixel of the current frame is set in correspondence with a pixel unit from a plurality of continuous frames before the current frame from which the blood vessel region is extracted. The hollow organ blood vessel according to claim 1 or 2, wherein the set criterion corresponding to a part of the blood vessel region included in each of the local regions is used while extracting the region. Extraction method.

7. The blood vessel region is determined while matching the thickness of the cross section of the living body with the local region having the predetermined size and expanding the region corresponding to the blood vessel region for each of the living body cross sections. 2. The method for extracting a hollow organ blood vessel according to item 1.

8. The hollow organ blood vessel extraction method according to claim 1, wherein the blood vessel region is determined while expanding the blood vessel region in the entire volume of the three-dimensional model in the Vota cell space.

9. Only the pixels in the blood vessel region are stored as display targets. 3. The hollow organ blood vessel extraction method according to item 1 or 2.

10. The hollow organ blood vessel extraction method according to claim 1, wherein extraction of a blood vessel region from the continuous frame is performed at least reciprocally with respect to a continuous direction of the frame.

11. The method according to claim 1 or 2, wherein the hollow organ is a fundus.

1 2. Specify the optic papilla of the fundus for the instrumented area of the first frame from which the extraction of the blood vessel area is started, and set the initial judgment criteria based on the pixel value of the specified optic papilla. 12. The method for extracting a hollow organ blood vessel according to claim 11, wherein the method is performed.

13. The hollow organ blood vessel extraction method according to claim 11, wherein the plurality of frames are configured by arranging the biological cross-sectional images on one axis in a central direction of the fundus.

14. The hollow organ blood vessel extraction method according to claim 1, wherein the previously prepared three-dimensional model of the hollow organ is information obtained by coordinate measurement.

15. A program for causing a computer to execute a process for extracting blood vessels from continuous frames of a biological organ cross-sectional image of a hollow organ,

To the computer,

A first step of inputting a quality region for each frame for extracting a blood vessel region with respect to a three-dimensional model of a hollow organ prepared in advance;

Based on the pixel value of the quality area of the first frame according to the judgment criteria for each pixel, A second step of determining a region belonging to the blood vessel region;

Setting a criterion in a second frame that is continuous with the first frame based on a pixel value of a region including the determined region in the blood vessel region. Blood vessel extraction program.

16. A program for causing a computer to execute a process for extracting blood vessels from a continuous frame of a biological organ cross-sectional image of a hollow organ,

To the computer,

A second step of judging a region belonging to the blood vessel region and a region not belonging to the blood vessel region based on a pixel value of the quality region of the first frame in accordance with a determination criterion for each pixel; A determination criterion in a blood vessel region in a second frame continuous with the first frame is set based on a pixel value of a region including the region, and based on a pixel value of a region including the non-attributed region outside the blood vessel region. A third step of setting a criterion outside the blood vessel region in a second frame following the first frame,

A hollow organ blood vessel extraction program characterized by executing

17. The computer further sets at least one vascular region with respect to the quality region of the first frame from which vascular region extraction is started, and sets the pixel value of the at least one vascular region. 17. The hollow organ blood vessel extraction program according to claim 15, wherein a fourth step of setting an initial determination criterion based on the program is executed.

18. In the second step, a temporary area is provided at the same position as the blood vessel area determined up to the previous frame in the current frame from which the blood vessel area is to be extracted. The hollow organ blood vessel extraction program according to claim 15 or 16, further comprising determining belonging to each pixel along a field.

19. In the second step, in the temporary area, when a pixel belonging to the blood vessel area is outside the temporary area, the pixel is regarded as a pixel of the blood vessel area of the current frame, and the blood vessel area is set as the pixel. If a pixel that does not belong to the vascular region and extends to the vascular region of the previous frame is present in the temporary region, the pixel is regarded as a pixel outside the vascular region of the current frame, and the vascular region is set to the previous pixel. 19. The hollow organ blood vessel extraction program according to claim 18, wherein the program includes reducing a blood vessel region of the frame.

20. In the second step, a three-dimensional local region of a predetermined size including each pixel of the current frame corresponding to a pixel unit from a plurality of continuous frames before the current frame from which a blood vessel region is to be extracted is included. The method according to claim 15, further comprising using the set determination criterion corresponding to a part of the blood vessel region included in each of the local regions while extracting. Hollow organ blood vessel extraction program.

21. The thickness of the cross section of the living body matches the local region of the predetermined size, and the second step performs the determination of the blood vessel region while expanding the region corresponding to the blood vessel region for each cross section of the living body. 21. The hollow organ blood vessel extraction program according to claim 20.

22. The second step is to determine a blood vessel region while expanding the blood vessel region in the voxel space with respect to the entire three-dimensional model. 2. A hollow organ blood vessel extraction program according to item 1.

23. The computer according to claim 15, further comprising: causing the computer to execute a fifth step of storing only pixels of the blood vessel region as display targets.

Item 7. The hollow organ blood vessel extraction program according to Item 6.

24. The computer according to claim 15, further comprising causing the computer to execute a sixth step of extracting a blood vessel region from the continuous frame at least reciprocally in a continuous direction of the frame. Or the hollow organ blood vessel extraction program according to item 16.

25. The computer program according to claim 15, wherein the hollow organ is a fundus.

26. The computer further specifies the optic disc of the fundus for the instrumental area of the first frame from which the extraction of the blood vessel area is started, and based on the pixel value of the designated optic disc. 26. The hollow organ blood vessel extraction program according to claim 25, wherein a seventh step of setting an initial determination criterion is performed by executing the seventh step.

27. The hollow organ blood vessel extraction program according to claim 25, wherein the plurality of frames are configured by arranging the cross-sectional images of the living body in a single axis in the center direction of the fundus.

28. The hollow organ blood vessel extraction program according to claim 15, wherein the previously prepared three-dimensional model of the hollow organ is information obtained by coordinate measurement.

2 9. An image processing apparatus,

A storage unit for storing a three-dimensional model of the hollow organ, An input unit for inputting a quality region for each frame for extracting a blood vessel region with respect to the three-dimensional model,

A determining unit that determines a region belonging to the blood vessel region based on a pixel value of an aggravated region of the first frame according to a determination criterion for each pixel;

A setting unit that sets a criterion in a second frame that is continuous with the first frame based on a pixel value of a region including the determined region in the blood vessel region;

An image processing apparatus comprising:

30. An image processing apparatus,

A storage unit for storing a three-dimensional model of the hollow organ,

An input unit for inputting a quality region for each frame for extracting a blood vessel region with respect to the three-dimensional model,

A determination unit that determines a region belonging to the blood vessel region and a region not belonging to the blood vessel region based on a pixel value of the quality region of the first frame according to a determination criterion for each pixel;

A criterion in the vascular region in the second frame that is continuous with the first frame is set based on the pixel value of the region including the region belonging to the blood vessel region, and the non-belonging region outside the blood vessel region is set. An image processing device comprising: a setting unit that sets a determination criterion outside a blood vessel region in a second frame that is continuous with the first frame based on a pixel value of a region included.

3 1. Further, at least one vascular region is set for the instrumented region of the first frame from which the extraction of the vascular region is started, and the initial value is set based on the pixel value of the at least one vascular region. Claims characterized in that it has a unit for setting the criterion of

29. The image processing apparatus according to item 29 or 30.

3 2. The determination unit sets a temporary area at the same position as the blood vessel area determined up to the previous frame in the current frame from which the blood vessel area is to be extracted, and along the boundary of the temporary area. 31. The image processing apparatus according to claim 29, wherein the determination of belonging is performed on a pixel-by-pixel basis.

3. In the temporary region, when the pixel to be attributed to the blood vessel region is out of the temporary region, the determination unit determines the pixel as a pixel of the blood vessel region of the current frame. If a pixel that does not belong to the vascular region and extends to the vascular region of the previous frame is present in the temporary region, the pixel is regarded as a pixel outside the vascular region of the current frame, and the vascular region is set to the 33. The image processing device according to claim 32, wherein the image processing apparatus reduces a blood vessel region of a previous frame.

0

34. The determination unit determines a three-dimensional local area of a predetermined size including each pixel of the current frame from a plurality of consecutive frames before the current frame from which a blood vessel area is to be extracted, in correspondence with each pixel. The image processing according to claim 29, wherein the set determination criterion corresponding to a part of the blood vessel region included in each of the local regions is used during the extraction. apparatus.

35. The thickness of the cross section of the living body is matched with the local region of the predetermined size, and the determination unit determines the vascular region while expanding the region corresponding to the vascular region for each cross section of the living body. The image processing apparatus according to claim 34, wherein the image processing apparatus is characterized in that:

0

36. The method according to claim 29, wherein the determination unit determines the blood vessel region while expanding the blood vessel region in a poxel space for the entire three-dimensional model. The image processing apparatus according to claim 1. 53. The image processing apparatus according to claim 29, further comprising a unit that stores only pixels in the blood vessel region as display targets.

38. The apparatus according to claim 29, further comprising: a unit that extracts a blood vessel region from the continuous frame at least reciprocally in a continuous direction of the frame. Image processing device. 39. The claim 29, wherein the hollow organ is a fundus.

Item 7. The image processing device according to item 0.

40. Further, the optic disc of the fundus is designated for the instrumented area of the first frame from which the extraction of the blood vessel area is started, and the initial criterion is determined based on the pixel value of the designated optic disc. 30. The image processing apparatus according to claim 39, further comprising: a unit for setting a value.

41. The image processing apparatus according to claim 39, wherein the plurality of frames are configured by arranging the cross-sectional images of the living body along one axis in a center direction of the fundus.

42. The image processing device according to claim 29, wherein the three-dimensional model of the hollow organ stored in the storage unit is information obtained by coordinate measurement.