JP2001070247A - Detection method of blood vessel crossing part in fundus image - Google Patents

Abstract

(57) Abstract: The present invention provides an image processing method for detecting a blood vessel crossing part that cannot be detected by a conventional blood vessel crossing part detection method that relies on blood vessel extraction. A fundus image is input (102), an average value of pixel values on the outer periphery of the filter is calculated using a square or circular filter (103), and the number of existing blood vessels is determined from the result. A determination step (104), a step (105) of determining the presence or absence of a blood vessel in the center of the filter, and a step (10) of determining the presence of a blood vessel crossing part in the filter region based on the determination results.
6) a blood vessel crossing detection method.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for providing information to assist a doctor's diagnosis by automatically detecting a crossing portion of a blood vessel in a fundus image used in a health checkup or a medical checkup. is there.

[0002]

2. Description of the Related Art In various medical fields, inspections using medical image equipment have been conventionally performed. In particular, a large amount of images are generated in various medical examinations and the like. Fundus examination is very important for the diagnosis of hypertension and arteriosclerosis, which often show signs in blood vessels. One of the important judgment items in the diagnosis based on the fundus is analysis of the state of the intersection of blood vessels. At present, doctors often look at fundus images and recognize crossing points of blood vessels to make diagnoses. On the other hand, as a detection method by a computer, there is a method of detecting a blood vessel crossing portion from an image obtained by extracting a blood vessel and thinning the image.

[0003]

In a group examination or the like, many people undergo a fundus examination, so that the burden on a doctor who diagnoses a fundus image increases. In addition, a normal fundus photograph is a small image, and care must be taken for diagnosis. Looking back on the current situation, it is important to provide an environment in which a doctor can concentrate on a portion to be focused on in terms of diagnostic ability, maintenance and improvement of efficiency. by the way,
As one of the important judgment items in the diagnosis of the fundus image,
There is an analysis of the intersection of blood vessels.

In the prior art, there are techniques for detecting a blood vessel crossing portion from a fundus image. In these methods, a blood vessel is extracted and then thinning processing is performed to extract the blood vessel crossing portion. In most cases, a blood vessel crossing portion is detected based on extracted blood vessel information. However, the degree of difficulty in extracting blood vessels increases as the fundus image shows a sign of abnormality, and in particular, the blood vessels around the blood vessel crossing part tend to be harder to identify as the image shows a sign showing an abnormality in the blood vessel crossing part. An object of the present invention is to detect a blood vessel crossing portion that cannot be detected by a conventional method from a fundus image because a blood vessel is hardly detected normally in this way.

[0005]

SUMMARY OF THE INVENTION According to the present invention, a blood vessel cannot be accurately detected in an image obtained by shading a fundus image when the area around the blood vessel intersection is unclear. An inability to cross blood vessels is detected by applying a cross blood vessel detection filter that performs the processing shown in FIG.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In a case where a blood vessel crossing portion cannot be detected due to inaccurate blood vessel extraction, a blood vessel crossing detection filter is created in the present invention to detect a blood vessel crossing portion. FIG. 1 is a flowchart of a process performed in the blood vessel crossing detection filter. This blood vessel crossing detection filter has a structure as described below.

FIG. 2 is a view showing an example of the structure of a blood vessel crossing part detection filter. The blood vessel crossing detection filter shown in FIG. 2 is a square filter of N × N pixels, and this area is further divided into n × n square small areas. N and n are positive integers. In the blood vessel crossing detection filter, a region shown by oblique lines in FIG. 2 is defined as an outer peripheral portion of the filter. The size of the blood vessel crossing detection filter is set so that the crossing region to be detected fits in the filter, and the size of the small square region in the filter is determined to be substantially equal to the width of the blood vessel. FIG. 3 shows an example of setting the size of the blood vessel crossing detection filter.

A blood vessel crossing part is detected by applying a blood vessel crossing part detection filter to a fundus image formed as a gray scale image. The processing in the blood vessel crossing detection filter will be described below.

First, a pixel value in an area to which a blood vessel crossing detection filter is applied is input from a gray scale fundus image as an input to the filter. Then, the average value of the pixel values is calculated for each square small area in the blood vessel crossing detection filter. Next, the presence or absence of a blood vessel is determined in each square small area on the outer peripheral portion of the blood vessel crossing detection filter. The presence or absence of a blood vessel in the square small area is determined by performing a threshold process by taking the difference between the average pixel value of the square square area of interest and the adjacent square small area. After this processing is performed on all the square small areas in the outer peripheral portion of the blood vessel crossing detection filter, how many square small areas where blood vessels exist are detected. However, at this time, when it is determined that blood vessels exist in the continuous square small areas, they are counted as one place. In addition, the average pixel value of the entire square small area determined to have these blood vessels is obtained, and compared with the average pixel value of the square small area at the center of the blood vessel crossing detection filter, to determine whether or not a blood vessel exists there. Is determined.

[0010] From these results, the output of the blood vessel crossing detection filter is determined. If there are four or more blood vessels in the outer peripheral portion of the blood vessel crossing detection filter, and if there are blood vessels in the central square small area, it will output a conclusion that a blood vessel crossing part exists in this area, Otherwise, it will output a conclusion that there was no vessel intersection. By adjusting the conditions here, it is possible to detect not only the blood vessel crossing part but also the blood vessel bifurcation. In addition, although the blood vessel crossing detection filter has been described as a square, as shown in FIG. 4, the definition of the outer peripheral portion of the filter is changed, and when the filter is deformed into a circular shape or the like, the size of the filter is adjusted and the like. The effect of is obtained.

[0011]

An embodiment of the present invention will be described with reference to the drawings. There is an area as shown in FIG. 5 (a) in the shaded fundus image,
The case where the detection result as shown in FIG. 5B is obtained will be described as an example.

A fundus image such as a polaroid photograph is read and digitized by, for example, a scanner to obtain a color digital image. After that, a shading process is performed. The grayscale image in this embodiment has 256 gradations, and the pixel value increases as the color approaches white and decreases as the color approaches black. The blood vessel region is a low concentration region.

The blood vessel crossing detection filter used in this embodiment has a size of 50 × 50 pixels, and the inside small square area is 10 × 10 pixels. Therefore, the blood vessel crossing detection filter is composed of a 5 × 5 square small area. The determination as to whether or not a blood vessel exists in the outer peripheral portion of the blood vessel crossing detection filter is based on the determination that the difference between the average pixel values in adjacent square small areas is 15 or more and the average pixel value in that case is lower. Is less than or equal to 180, it is determined that a blood vessel exists in the square small area having the lower average pixel value. Further, when the average pixel value of the central square small area is lower than a value obtained by adding 10 to the average value of the pixel values of the square small area in which the blood vessels at the outer periphery of the blood vessel crossing detection filter are present, the central square small area It was determined that a blood vessel was present.

Using such a blood vessel crossing detection filter, an image shown in FIG.
An area surrounded by a square in (b) is detected as a blood vessel intersection. The processing inside the blood vessel crossing detection filter in this region is as shown in FIG. 6, and since there are four blood vessels existing in the outer peripheral part and blood vessels existing in the central part, the detection conditions match. It was detected.

[0015]

As described above, a blood vessel crossing part which cannot be detected by the conventional blood vessel crossing detecting method depending on the blood vessel extraction result can be automatically detected by a computer using the present invention. The position information of the blood vessel crossing part is important information in the fundus diagnosis, and the application to the analysis of the blood vessel crossing part by a computer based on this information can be considered. In particular, since a blood vessel crossing portion where blood vessels become unclear near the blood vessel crossing portion is often a sign of an abnormality, it can be said that detecting such a portion is useful.

[Brief description of the drawings]

FIG. 1 is a flowchart of processing in a blood vessel crossing detection filter.

FIG. 2 is a structural view of a blood vessel crossing detection filter.

FIG. 3 is an example of determining the size of a blood vessel crossing detection filter.

FIG. 4 is a modified example of a blood vessel crossing part detection filter.

FIG. 5 is an embodiment of a blood vessel crossing detection filter.
(A) is an original image, (b) is a blood vessel crossing part detection result.

FIG. 6 is an example of a process of processing in a blood vessel crossing detection filter. (A) is a calculation result of the average pixel value in the square small area, and (b) is a detection result of a blood vessel area existing in the filter.

[Explanation of symbols]

101 Start 102 Pixel value input processing 103 Average value calculation of pixel values in each small square area 104 Processing for determining the number of blood vessels existing on the outer periphery of the filter 105 Processing for determining the presence of blood vessels in the center of the filter 106 ... Processing for judging whether or not a blood vessel crossing portion exists 107. End 201. Blood vessel crossing portion detection filter 202. Square small area in the blood vessel crossing portion detection filter 301. Artery 302. Vein 303.

Continued on the front page (72) Inventor Kazuaki Sugio 11F term of 4-105, Muromura-cho, Ogaki-shi, Gifu 5B057 AA07 CA08 CA12 CA16 CB08 CB12 CB16 CC02 CE06 CH09 DA08 DB02 DB09 DC22

Claims (1)

[Claims] 1. A method for detecting a blood vessel crossing portion, which cannot be detected by a method of detecting a blood vessel crossing portion from an image obtained by extracting a blood vessel from an image of a fundus and thinning the image, to form a square or circular shape. Inputting the pixel value using the filter of the above, the step of determining the number of blood vessels present on the outer peripheral portion of the filter, the step of determining the presence or absence of the blood vessel in the center of the filter, based on these determination results A method for detecting a blood vessel crossing portion in a fundus image, comprising the step of determining the presence of a blood vessel crossing portion in a filter region.