KR101182729B1 - Medical image processing methods for blood vessel - Google Patents

Abstract

I) Adaptive partitioning of blood vessels and non-vessels based on the thresholds after setting regional thresholds based on the average or standard deviation of image pixels in each region Local binarization processing step; ii) a morphology calculation processing step of performing a morphology calculation to correct noise included in the image in which the adaptive local binarization processing step is completed; iii) a thinning processing step of extracting a blood vessel line by thinning the image in which the morphology calculation processing step is completed; iv) determining the direction of the blood vessel line extracted in each region in the image where the thinning process is completed, and then analyzing the gray profile of the blood vessel orthogonal to the blood vessel direction to measure the thickness of the blood vessel; And v) an adaptive Gabor filtering process for selectively applying a Gabor filter most suitable for each region in consideration of the direction and thickness of the blood vessel measured in step iv).

Description

Medical Image Processing Methods for Blood Vessel

The present invention relates to a medical vascular image processing method, and more specifically, by applying a Gabor filter in consideration of the direction and thickness of the processed vascular image after the binarization process, morphology calculation process and thinning process, The present invention relates to a medical blood vessel image processing method for observing a blood vessel included in a blood vessel image more precisely and clearly.

In the blood vessel image photographed by using an infrared light or an infrared camera, the blood vessel part is darkly displayed, and the non-vascular part is brightly displayed.

This phenomenon is caused by the absorption of the near-infrared wavelength illumination used in the blood vessel imaging to hemoglobin contained in the blood in the blood vessels.

In particular, in order to use the blood vessel image photographed by the above method in the medical imaging field, it is important to accurately separate blood vessels and non-vessels, and provide a clear image quality to display a region of fine vessels that cannot be visually observed. shall.

In addition, when judging whether the blood vessels of the junction region is correctly bonded or when the blood vessels are blocked, during the splicing operation due to the cutting of the finger, the blood vessel image of the finger or the hand should be clear to perform the treatment more easily.

On the other hand, methods such as histogram smoothing, histogram stretching, and median filtering are used as methods for improving image quality such as portrait and landscape images. In this case, the detailed shape of the blood vessel may not be displayed due to the structural shape of the blood vessel, and only a thick blood vessel degree may be displayed.

In addition, as an example of processing blood vessel images, Shi (Z. Shi., W. Yiding, and W. Yunhong, "Extracting Hand Vein Patterns from Low-quality Images: A New Biometric Technique Using Low-cost Devices," In Proc of the Fourth International Conf. on Image and Graphics, 2007), a matched filter and a Wiener filter to remove image noise as much as possible, taking into account the noise of an imaging device that prevents accurate separation of blood vessels and non-vessels. (wiener filter) and average filter (average filter) were used as a pretreatment step, through which separation of blood vessels and non-vessels was attempted.

However, the method may blur the processed image by filtering by using a plurality of filters in the pre-processing step, which may cause an incorrect separation of blood vessels and non-vessels in the separation of blood vessels and non-vessels.

To overcome these problems, Lingyu (W. Lingyu and L. Graham. “Gray-scale Skeletonization of Thermal Vein Patterns Using the Watershed Algorithm in Vein Pattern Biometrics,” in Proc. Of Int. Conf. On Computational Intelligence and Security, 2006 .) Proposed a method of extracting the skeleton of the image by replacing the vascular and non-vascular separation stages of the image with a watershed algorithm.

However, this method has a problem in that when two blood vessels are located close to each other, it is difficult to distinguish them.

Thus, Miura (N. Miura, N. Akio, and M. Takafumi, “Extraction of Finger-Vein Patterns Using Maximum Curvature Points in Image Profiles,” IEICE Transactions on Information and Systems, vol. E90-D, no. 8, 1185-1194, 2007.) discloses a preprocessing method for effectively extracting features of blood vessel images and applying them to real-time retrieval in view of the above problems.

However, since the image noise that is not removed in the preprocessing process has a great influence on the separation of blood vessels and non-vessels, the Miura image processing method has a problem of showing an inaccurate image processing result after image thinning.

In addition, since the method attempts to separate blood vessels and non-vessels using the histogram curvature of the image, there is a problem in that the adjacent vessels cannot be effectively distinguished.

The present invention provides a medical vascular image processing method for clarifying the distinction between blood vessels and non-vessels by removing noise signals of blood vessel images and amplifying signals of blood vessel portions, so that not only thick blood vessels but also invisible thin blood vessels can be clearly displayed. do.

The present invention

i) An adaptive regional binarization process that divides blood vessel images into regions and sets regional thresholds based on the average or standard deviation values of the image pixels in each region, and then classifies and binarizes the vessels and non-vessels based on the thresholds. step;

ii) a morphology calculation processing step of performing a morphology calculation to correct noise included in the image in which the adaptive local binarization processing step is completed;

iii) a thinning processing step of extracting a blood vessel line by thinning the image in which the morphology calculation processing step is completed;

iv) determining the direction of the blood vessel line extracted in each region in the image where the thinning process is completed, and then analyzing the gray profile of the blood vessel orthogonal to the blood vessel direction to measure the thickness of the blood vessel; And

and v) an adaptive Gabor filtering process for selectively applying a Gabor filter most suitable for each region in consideration of the direction and thickness of the blood vessel measured in step iv).

Medical vascular image processing method according to the present invention can be used for the medical diagnosis and treatment of vascular diseases by obtaining a high-resolution quality vascular image.

In addition, the present invention can clarify the distinction between blood vessels and non-vascular vessels, and can provide a blood vessel image capable of clearly displaying not only thick blood vessels but also invisible thin vessels.

1 is a flow chart showing a medical blood vessel image processing method according to the present invention,
2 is a view showing an image processed according to the medical blood vessel image processing method of the present invention,
3 is a view showing a spatial region form of a Gabor filter according to the present invention;
4 is a view showing the frequency domain form of the Gabor filter according to the present invention.

I) Adaptive partitioning of blood vessels and non-vessels based on the thresholds after setting regional thresholds based on the average or standard deviation of image pixels in each region Local binarization processing step; ii) a morphology calculation processing step of performing a morphology calculation to correct noise included in the image in which the adaptive local binarization processing step is completed; iii) a thinning processing step of extracting a blood vessel line by thinning the image in which the morphology calculation processing step is completed; iv) determining the direction of the blood vessel line extracted in each region in the image where the thinning process is completed, and then analyzing the gray profile of the blood vessel orthogonal to the blood vessel direction to measure the thickness of the blood vessel; And v) an adaptive Gabor filtering process for selectively applying a Gabor filter most suitable for each region in consideration of the direction and thickness of the blood vessel measured in step iv).

The blood vessel image processing method, preferably medical blood vessel image processing method according to the present invention when the doctor photographs the blood vessels of the body, such as hands, fingers, back of the hand, fingers, toes or both for the purpose of treating or treating a patient, By processing the photographed blood vessel image to be able to observe the blood vessel contained in the image more precisely and clearly.

In this case, the body is not limited to a hand, a finger, a back of a hand, a finger, a toe, or both, but may include other parts according to a user's selection.

In addition, the blood vessel image refers to blood vessel images such as magnetic resonance imaging (MRI) and computed tomography (CT) photographed using contrast media, as well as blood vessel images photographed by infrared illumination and infrared cameras.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. However, the following description is only for the purpose of specifically describing the present invention, and the scope of the present invention is not limited by the following description.

1 is a flow chart showing a medical blood vessel image processing method according to the present invention, Figure 2 is a view showing an image processed according to the medical blood vessel image processing method of the present invention, Figure 3 is a spatial region of the Gabor filter according to the present invention Fig. 4 is a diagram showing a frequency domain form of the Gabor filter according to the present invention.

As shown in Figures 1 to 4, the medical vascular image processing method according to the present invention i) partitioning the vascular image to the region and then set the regional threshold based on the average or standard deviation value of the image pixel in each region An adaptive local binarization process of classifying and binarizing blood vessels and non-vessels based on the threshold value; ii) a morphology calculation processing step of performing a morphology calculation to correct noise included in the image in which the adaptive local binarization processing step is completed; iii) a thinning processing step of extracting a blood vessel line by thinning the image in which the morphology calculation processing step is completed; iv) determining the direction of the blood vessel line extracted in each region in the image where the thinning process is completed, and then analyzing the gray profile of the blood vessel orthogonal to the blood vessel direction to measure the thickness of the blood vessel; And v) an adaptive Gabor filtering process for selectively applying a Gabor filter most suitable for each region in consideration of the direction and thickness of the blood vessel measured in step iv).

The binarization, preferably adaptive local binarization processing step according to the present invention divides a blood vessel image, preferably a photographed blood vessel image into regions, and then thresholds the region based on an average or standard deviation value of image pixels in each region. After the value is set, blood vessels and non-vessels are classified based on the threshold value, and are not particularly limited as long as it is a normal binarization step for this purpose.

The blood vessel image refers to blood vessel images such as magnetic resonance imaging (MRI) and computed tomography (CT) photographed using contrast agents, as well as blood vessel images photographed by infrared illumination and infrared cameras.

In addition, the blood vessel image according to the present invention is a target image to be processed by sharpening the image, may be referred to as the original blood vessel image.

Here, the original blood vessel image includes a dark portion and a brightly expressed portion, as shown in ① of FIG. 2, wherein the dark portion represents blood vessels and the bright portion represents non-vascular portions.

In particular, the adaptive local binarization processing step according to the present invention is to more effectively classify blood vessels, blood vessels and portions of the original blood vessel image, and divides the entire photographed image, that is, the entire original image into regions of a constant size; After setting a threshold value for each of the divided regions, blood vessels and non-vessels are distinguished based on the threshold values for each region.

Here, when binning the entire image by region and setting a threshold value for each region, the binarized image can be clearly and easily obtained than binarization by setting a threshold value based on the entire image.

Meanwhile, the partitioned area may select a location and size of the area according to a user's selection, and the threshold for each area is set based on an average value or standard deviation value of image pixels in each area.

The morphology operation processing step according to the present invention is to correct the noise included in the image in which the step i) the binarization processing step is completed.

The correction is performed by applying an opening operation that performs an erosion operation after performing an expansion operation, or a closing operation that performs an expansion operation after performing an erosion operation, and a blood vessel is incorrectly broken due to noise included in the binarized image. And remove what the non-vessels are incorrectly marked as blood vessels.

In this case, the non-vessel portion is removed from the binarized image through the correction, and the portion where the vessel is broken can be minimized.

The thinning processing step according to the present invention is to extract blood vessel lines by thinning the image of which the morphology calculation processing step of step ii) is completed, and if the conventional thinning processing method of the art for this purpose is used. Anything may be used.

In the step of measuring the direction and thickness of blood vessels according to the present invention, after determining the direction of the extracted blood vessel line in each region of the image in which the thinning process is completed, the step thickness of blood vessels is analyzed by analyzing the gray profile of the blood vessels orthogonal to the blood vessel direction. As to measure, if it is a conventional method for measuring the direction and thickness of blood vessels for this purpose is not particularly limited.

In particular, as shown in Figure 2, the step of measuring the direction and thickness of the blood vessel according to the present invention is a previous step for performing the adaptive Gabor filtering processing step of step v), the thinning process is a specific step in the image is finished The direction orthogonal to the blood vessel line (FIG. 2, ⑤ blood vessel direction) extracted in the area | region (square shown by (4) of FIG. 2) is grasped | ascertained.

To this end, the direction and thickness measurement of the blood vessel may be performed in the region of the original blood vessel image (the square shown in ① of FIG. 2) at the same position as the region (the square shown in ④ of FIG. 2) in the image where the thinning process is completed. Analyze the gray profile of the vessel in the orthogonal direction extracted from the through to measure the thickness of the vessel.

In addition, as shown in ⑤ of FIG. 2, the step of measuring the direction and thickness of the blood vessel may predict the vessel thickness and the direction of the blood vessel image, that is, the original vessel image, while moving from region to region for the entire thinning image. Accordingly, an optimal Gabor filter having a size and a direction corresponding to the predicted blood vessel thickness and / or direction can be applied.

In the adaptive Gabor filtering step according to the present invention, a Gabor filter is most suitable for each region in consideration of the direction and thickness of the blood vessel in which the measurement of the direction and thickness of the blood vessel of step iv) is completed. Is selectively applied to process an image according to the Gabor filter equation of Equation 1 and the expression in the frequency domain of Equation 2.

Here, (x, y) of Equation 1 represents the (x, y) pixel position in the filter, g (x, y) represents the Gabor filter coefficient value at the (x, y) position.

는 가버 필터의 주파수를 나타내고, 상기

,

는 가버필터 형태의 표준편차를 나타내는 파라미터이다.In addition, the equation 1

Represents the frequency of the Gabor filter,

,

Is a parameter representing the standard deviation of the Gabor filter type.

,

,

,

에 의해 가버 필터의 형태가 결정된다. In addition, in Equation 1

,

,

,

The shape of the Gabor filter is determined by.

는 가버 필터의 방향을 나타내는 것으로 도 2의 ④ 에 따른 세선화 영상에서의 혈관 방향 값을 사용하게 되며, 상기 가버 필터의 주기(period)는 검출된 혈관 두께의 2배인 수치로 설정한다. At this time,

Represents the direction of the Gabor filter, and uses the blood vessel direction value in the thinning image according to ④ of FIG. 2, and sets the period of the Gabor filter to a value that is twice the thickness of the detected blood vessel.

(가버 필터의 주파수)는 "1/주기"의 수식에 의해 산출될 수 있으므로, 결과적으로 "1/(혈관두께의 2배)"로 정해진다.In addition,

(Frequency of Gabor filter) can be calculated by the formula of " 1 / cycle ", and as a result, is determined to be " 1 / (double the vessel thickness) ".

,

)는 가버 필터의 형태를 결정하는 것으로,

는

과 동일한 값으로 사용하며, 가버 필터의 크기는

(

)의 2배 정도 크기 값으로 사용한다.Also, in general, the standard deviation (

,

) Determines the shape of the Gabor filter,

The

Use the same value as, and the size of Gabor filter

(

Use twice the size as).

와

가 자동적으로 결정되며, 나머지 파라미터인

(

)만 결정되면 가버 필터의 크기[

(

)의 2배] 역시 설정되어, 최종적인 가버 필터가 결정된다.Therefore, as shown in ④ and ⑤ of FIG. 2, when the direction and thickness of the blood vessel are calculated through the thinning processing step and the direction and thickness measurement of the blood vessel,

Wow

Is automatically determined and the rest of the parameters

(

) Is determined, the size of the Gabor filter

(

2 times]) is also set to determine the final Gabor filter.

(

)를 산출하기 위하여, 미리 사용자가 학습영상을 이용하여 눈으로 관찰하는 실험을 통해, 원본 혈관영상에서 구한 혈관 두께와 최적의 가버 필터 파라미터

(

)와의 상관관계를 실험적으로 구하여 테이블에 미리 저장하고, 이후 실제 동작과정에서 입력 영상의 혈관 두께가 도 2에 도시된 이진화, 모폴로지 연산, 세선화 처리단계, 혈관의 방향 및 두께 측정단계를 통해 검출되면, 테이블에 저장된 수치를 검색하여 최적의

(

)를 선택할 수 있다.Specifically, the parameters according to the invention

(

In order to calculate), the blood vessel thickness and the optimal Gabor filter parameters obtained from the original blood vessel image by experimenting with the eyes by using the learning image in advance.

(

Experimentally obtain the correlation with), and store it in a table in advance, and then detect the blood vessel thickness of the input image through the binarization, morphology calculation, thinning processing step, blood vessel direction and thickness measurement step shown in FIG. When the value is stored in the table,

(

) Can be selected.

(

)는 일반적으로 혈관 두께가 클수록 큰

(

)값을 사용하고, 혈관두께가 작을수록 작은

(

)값을 사용한다.Where the parameter

(

) Is generally larger

(

), The smaller the vessel thickness, the smaller

(

Value.

Meanwhile, Equation 2 represents an expression in the frequency domain of Equation 1.

, 와

는 필터 형태의 표준편차로써 각각

,

를 나타낸다.In this case, W in Equation 2 represents the frequency of the filter,

, Wow

Is the standard deviation of the filter type,

,

Indicates.

3 and 4 illustrate the shapes of the Gabor filter in a spatial domain and a frequency domain, respectively.

Two axes on the bottom of FIG. 3 mean x and y coordinates of Equation 1, and a vertical axis of FIG. 3 means g (x, y) of Equation 1.

In addition, in FIG. 4, the two axes on the bottom surface represent u and v coordinates of Equation 2, and the vertical axis represents G (u, v) of Equation 2.

In particular, in the adaptive Gabor filtering step of step v), the Gabor filter coefficients obtained according to Equations 1 and 2 are stored in a table, and each region is considered in consideration of the direction and thickness of the blood vessel measured in step iv). Optionally apply the optimal Gabor filter coefficients.

In this case, the area refers to a selection area of an image partitioned according to a user's selection.

Medical vascular image processing method according to the present invention having such a configuration is not only easy to check whether the blood vessels of the junction site is correctly connected after the body cutting and splicing operation of the fingers, toes, hands, feet, etc., the normal physical condition After recording the blood vessel image acquired in the and compared with the image information afterwards can be measured periodically to determine whether the blockage of blood vessels through this, such as Burger's disease (Buerger's disease, Raynaud's phenomenon) Not only can the disease be diagnosed early, but early detection of connective tissue can lead to early diagnosis of rheumatic disease, and to quickly determine whether the toe vessels are blocked due to diabetes.

As described above, those skilled in the art will understand that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. Therefore, the embodiments described above are all illustrative and should not be considered as limiting. The scope of the present invention should be construed as being included in the scope of the present invention all changes or modifications derived from the meaning and scope of the appended claims rather than the detailed description and equivalent concepts thereof.

Claims (3)

i) An adaptive regional binarization process that divides blood vessel images into regions and sets regional thresholds based on the average or standard deviation values of the image pixels in each region, and then classifies and binarizes the vessels and non-vessels based on the thresholds. step;
ii) a morphology calculation processing step of morphological calculation to correct vascular breakage or non-vascular classification as blood vessels due to noise included in the image of which the adaptive local binarization processing step is completed;
iii) a thinning processing step of extracting a blood vessel line by thinning the image in which the morphology calculation processing step is completed;
iv) determining the direction of the blood vessel line extracted in each region in the image where the thinning process is completed, and then analyzing the gray profile of the blood vessel orthogonal to the blood vessel direction to measure the thickness of the blood vessel; And
and v) an adaptive Gabor filtering process for selectively applying a Gabor filter most suitable for each region in consideration of the direction and thickness of the blood vessel measured in step iv). delete The method of claim 1,
In the adaptive Gabor filtering step of step v), the Gabor filter coefficients obtained according to Equation 1 and Equation 2 are stored in a table, and the optimum Gabor filtering coefficients are optimized for each region in consideration of the direction and thickness of blood vessels measured in step iv). A blood vessel image processing method comprising selectively applying Gabor filter coefficients.

&Quot; (1) "

here,
The (x, y) represents the (x, y) pixel position in the filter,
G (x, y) represents a Gabor filter coefficient value at the (x, y) position,
remind

Represents the frequency of the Gabor filter,
Θ represents the direction of the Gabor filter,
remind

,

Is a parameter representing the standard deviation of the Gabor filter type;

&Quot; (2) "

here,
Equation 2 represents an expression in the frequency domain of Equation 1,
W represents the frequency of the filter,
remind

Wow

Is the standard deviation of the filter type,

,

Indicates.

Images