CN101138498A - An image processing method based on magnetic resonance three-dimensional renogram - Google Patents

Abstract

The invention relates to an image processing method based on a magnetic resonance three-dimensional nephrogram, which comprises the following steps: (1) reading the magnetic resonance three-dimensional scanning image of the kidney from a magnetic resonance workstation; (2) determining the number of each phase image and background phase, corticomedullary boundary phase, and whole kidney phase; (3) Resect the relevant blood vessels and renal pelvis in the whole kidney phase image; (4) Identify the outline of the whole kidney; (5) Match the background phase with The cortex and medulla are divided into phases and silhouetted; (6) segment the cortex and medulla of the kidney; (7) match images of all phases. The present invention proposes a new kidney dynamic enhanced image segmentation and matching scheme, which can more accurately correct the movement of the kidneys caused by breathing during the MRI image acquisition process, and can accurately segment the cortex and medulla. Practice has proved that the image processing method of the present invention can obtain a reasonable three-dimensional magnetic resonance renogram. The invention can be widely applied in the image processing of the magnetic resonance renogram.

Description

An image processing method based on magnetic resonance three-dimensional renogram

technical field

The invention relates to a magnetic resonance image processing method, in particular to an image processing method based on a magnetic resonance three-dimensional renogram.

Background technique

MRI (Magnetic Resonance Imaging, Magnetic Resonance Imaging) dynamic contrast-enhanced scan, as a safe way to evaluate renal function, can be used to evaluate renal perfusion and filtration functions, and has important clinical application prospects. During dynamic contrast-enhanced renal MR (magnetic resonance) scanning, MRI contrast agents pass through the cortex, medulla, and collection system of the kidney in sequence, and the changes in signal intensity of contrast agents in different regions relative to time can reflect some functions of the region. The time-signal intensity curve formed by the change of signal intensity is called MRR (Magnetic Resonance Renography), also called magnetic resonance renography. Magnetic resonance renography, with high spatial resolution, can clearly distinguish the renal cortex and medulla, and the examination process has no radiation damage to the patient. If the MRR curves of the cortex and medulla can be analyzed separately or comprehensively, and the corresponding relationship between the signal intensity of the image and the concentration of the contrast agent can be established, then it is possible to obtain important indicators such as the glomerular filtration rate, and then establish a set of quantitative parameters. Analyze the system to obtain the index parameters corresponding to different types of kidney diseases, and provide an important basis for the diagnosis of kidney diseases.

During the kidney dynamic contrast-enhanced scanning process of collecting MRR, the patient's abdomen can be scanned repeatedly and quickly to obtain multi-temporal data and draw MRI renogram. The total scanning time can reach 2 to 4 minutes. In the magnetic resonance images collected at different times, since the kidneys will be displaced to a certain extent with the respiratory movement of the subject, the MRR automatically drawn on this basis will have obvious partial volume effects, which will affect the accuracy of the MRR . To obtain accurate MRR, it is necessary to register dynamically enhanced images in different phases. Hospitals now mostly use medical image post-processing software attached to MRI workstations. Up to now, no workstations have been able to perform targeted and specialized processing on renal perfusion images, especially the correction of respiratory motion and the precise division of renal parenchyma. ; Even if the achievements in this area are abroad, they only stay at the research level. Since the image movement caused by respiration cannot be corrected and the cortex and medulla can not be automatically and efficiently segmented, the renogram error is relatively large, and it is difficult to provide valuable clinical information, which brings difficulties to follow-up research.

At present, doctors can only obtain the region of interest and perform data analysis by manually delineating a single image. However, the three-dimensional renal perfusion with LAVA (liver acquisition with volume acceleration) sequential scanning The number of images is huge. Foreign studies have shown that if the image registration and segmentation are all performed manually by researchers, it takes about 2 to 3 hours to process an image of a patient. For clinical applications, this kind of workload does not allow time. Also too big. Therefore, the efficient and accurate identification and correction of the movement caused by breathing, as well as the accurate segmentation of the cortex and medulla, have become a very key topic in the current MRI renography research.

Contents of the invention

In view of the above problems, the purpose of the present invention is to provide a method that can more accurately correct the movement of the kidneys due to breathing, and can accurately segment the cortex and medulla, thereby obtaining a reasonable three-dimensional magnetic resonance renogram, which is useful for the diagnosis of kidney diseases. An image processing method based on magnetic resonance three-dimensional renography that provides an important basis.

To achieve the above object, the present invention adopts the following technical solutions: an image processing method based on a magnetic resonance three-dimensional nephrogram, which includes the following steps: (1) reading the magnetic resonance three-dimensional scanning image of the kidney from the magnetic resonance workstation; (2) ) Determining the number of images in each phase, the background phase, the corticomedullary boundary phase, and the whole kidney phase; (3) resecting the relevant blood vessels and renal pelvis regions in the phase images of the whole kidney; (4) identifying the outline of the whole kidney; (5) Match the background phase with the corticomedullary boundary phase, and silhouette; (6) Segment the cortex and medulla of the kidney; (7) Match images of all phases.

In step (4), the steps of obtaining the contour of the whole kidney using the region growing method are as follows: ① The user selects one or several seed points P1, P2, P3, ..., Pn that fall within the kidney region in the whole kidney phase; ② Take a point Pi from all seed points as the starting point of this region growth; ③ calculate the growth threshold T of the seed point; ④ calculate the difference between the gray value of each point in the 26 neighborhood of the current seed point and I, if the difference is less than The growth threshold of the current seed point, merge this point into the result area, and count it as a new seed point; ⑤Repeat steps ②, ③, ④ until the result area cannot grow.

Use the linear calculation method to determine the growth threshold T of the seed point, let T=I*k, where I is the average gray value of the result area obtained by the current area growth, and k is the slope set by the user.

Wherein in step (5), the matching background phase and the corticomedullary boundary phase use the cortical area as a template, and the traditional two-dimensional correlation matching based on the correlation coefficient is extended to three dimensions, and the calculation formula of the three-dimensional correlation coefficient is as follows :

$\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; r</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}\text{<span class="notranslate"> \&rsqb;\&lsqb;w</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; w</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \{</span>\underset{\mathrm{<span\; class="notranslate">\; r</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; \&rsqb;</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\underset{\mathrm{<span\; class="notranslate">\; r</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; w</span>}}<span\; class="notranslate">\; \&rsqb;</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}{<span\; class="notranslate">\; \}</span>}^{\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$

为已知图像皮质区域象素的灰度均值，x，y，z为肾在待匹配三维图像内平移的坐标，w(x+r，y+s，z+t)为皮质平移后在待匹配图像内对应点的灰度值，

为其均值。ρ(x，y，z)为计算所得的相关系数，ρ(x，y，z)∈[-1，1]。Among them, r, s, t represent the three-dimensional coordinates of the pixels in the cortical area, f(r, s, t) is the gray value of the pixel in the cortical area of the known image,

is the average gray value of the pixels in the cortical area of the known image, x, y, z are the translation coordinates of the kidney in the three-dimensional image to be matched, w(x+r, y+s, z+t) is the coordinates of the kidney to be matched after translation Match the gray value of the corresponding point in the image,

for its mean value. ρ(x, y, z) is the calculated correlation coefficient, ρ(x, y, z)∈[-1, 1].

Because the present invention adopts the above technical scheme, it has the following advantages: 1. Since the present invention can more accurately correct the movement of the kidneys caused by breathing in the MRI image acquisition process, and can accurately segment the cortex and medulla, the image space resolution The higher the rate, the more accurate the MRR. 2. Since the present invention can separately or comprehensively analyze the MRR curves of the cortex and medulla, and establish a corresponding relationship between the signal intensity of the image and the concentration of the contrast agent, the quantitative analysis of the MRR diagram becomes possible, providing an important basis for kidney diagnosis. 3. Since the present invention can efficiently and accurately identify and correct movement caused by breathing, and accurately segment cortex and medulla, time can be saved and work intensity can be reduced. 4. Since the present invention proposes a cutting method for the situation that the image is three-dimensional and there are many levels to be processed, the complexity of this step can be greatly simplified.

Description of drawings

Fig. 1 is a flowchart of the present invention

Figure 2 is the initial interface diagram of the running program

Figure 3 is the interface diagram for selecting the image directory to be processed

Figure 4 is the interface diagram for setting the phase length

Figure 5 is an interface diagram of the time-sharing phase display image set structure

Fig. 6 is the interface diagram of inputting the specific phase number (background, CMD, whole kidney) selected by the user

Figure 7 is an interface diagram for confirming that the input phase number is correct

Figure 8 is the pop-up image window

Figure 9 is a schematic diagram of the confirmed and modified kidney outline

Figure 10 is a schematic diagram of the segmentation results of the corticomedullary region

Figure 11 is a schematic diagram of the contrast agent concentration change curve

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings and embodiments.

As shown in Figure 1, an image processing method of a magnetic resonance three-dimensional renogram based on region growth and correlation matching in the present invention, its specific operation process is as follows:

The first step: run the program, the program interface appears (as shown in Figure 2), call the image processing module, select the image directory to be processed (as shown in Figure 3), read the image, the MRI image obtained from the magnetic resonance workstation It is in DICOM (Digital Imaging Communications in Medicine) format.

DICOM files refer to medical files stored in accordance with the DICOM standard. The DICOM file header contains relevant information identifying the data set, in other words, a lot of information about the image is stored in the file header. Since the order of filenames of DICOM images is not consistent with the order of images in the scan, it is a prerequisite for us to get the correct image number to process the image, and this information is stored in a tag called InstanceNumber. By reading the value of InstanceNumber, the images can be sorted correctly and ready for the next step of processing.

Step 2: The user inputs parameters related to the currently processed image sequence, including the number of images in each phase, the background phase, the CMD (corticomedullary differentiation) phase, the whole kidney phase, and the mesonephros in each phase start and end images. The specific operation steps are as follows:

1. Input phase length (as shown in Figure 4);

2. Time-sharing phase display image set structure (as shown in Figure 5);

3. Input the specific time phase number selected by the user, the phase number includes background, CMD and whole kidney (as shown in Figure 6);

4. Confirm that the input phase number is correct (as shown in Figure 7).

Step 3: In the image window that pops up, select a point in the central area of the kidney by double-clicking the mouse (as shown in Figure 8).

Step 4: Confirm and modify the contour of the kidney obtained, and perform the operation of removing the renal pelvis if necessary (as shown in Figure 9). In some aspects of the whole kidney phase, due to the presence of contrast medium in the renal blood vessels, renal pelvis, and calyces, these tissues and cortex are simultaneously developed; Some parts are closely related to the kidney, which brings great difficulties to the recognition of the contour of the kidney. Therefore, before the contour recognition of the whole kidney, it is necessary to separate the areas that may affect the segmentation results, especially the renal pelvis and blood vessels with the renal parenchyma. separate.

After performing the necessary cleanups, a region growing method is proposed to obtain the contour of the whole kidney, and the specific steps are as follows:

1. The user selects one or several seed points P1, P2, P3, ..., Pn in the whole kidney phase (seed points should fall within the kidney area);

2. Select a point Pi from all the seed points as the starting point of this regional growth;

3. Calculate the growth threshold T of the seed point: In the traditional method, T is a fixed threshold, which needs to be adjusted several times to obtain the ideal result when processing different images. The present invention adopts a method for linearly calculating the discrimination threshold, so that T =I*k, where I is the average gray value of the resulting region obtained by growing the current region, and k is the slope set by the user. Inside the kidney region, the average gray value of the region is large, and the discrimination threshold is therefore relatively large, which can combine some dark spots caused by noise; while at the edge of the kidney region, the gray value of the pixel decreases rapidly, and the average value of the region decreases. As a result, the discrimination threshold is also sharply reduced, which can effectively prevent overgrowth.

4. Calculate the difference between each point and the gray value of I in the 26 neighborhoods of the current seed point (the set of 26 adjacent pixels in the image space), if the difference is less than the growth threshold of the current seed point, put This point is merged into the result area and counted as a new seed point.

5. Repeat operations 2, 3, and 4 until the resulting area cannot grow.

Step 5: The user confirms the results of corticomedullary region segmentation, including silhouette, pre-matching, corticomedullary region segmentation and contour saturation adjustment (as shown in Figure 10).

The image silhouette is to subtract the images of the CMD phase and the background phase of the same section, so that the influence of the background signal can be removed in the CMD phase, the display of the contrast agent signal is highlighted, and the contrast of the image is improved, which is beneficial to Subsequent identification and segmentation of cortex and medulla.

The signal intensities of the cortex and medulla vary continuously at different times due to the different concentrations of contrast agents, and the trends of cortical and medulla brightness are quite different. The traditional correlation coefficient matching method uses an m×n sub-image as a template, which is suitable for static images, so this method is not applicable to the matching of kidney dynamic enhancement images. Considering that during the entire dynamic enhancement process, the signal intensity of the entire cortical area changes relatively uniformly with the change of the contrast agent concentration, so the traditional template matching method is improved, using the cortical area as a template, and the traditional correlation-based The two-dimensional correlation matching of coefficients is extended to three dimensions, further enhancing the accuracy of the matching results. During image matching, the images with the same tangent plane at different times are taken out according to the moving template to obtain two corresponding pixel point sets, and the correlation coefficient of the two point sets is calculated by using the correlation coefficient calculation formula, and the image template displacement corresponding to the maximum correlation coefficient to adjust the original image to complete the image matching between the two time phases.

The three-dimensional correlation coefficient calculation formula used in the present invention is as follows:

$\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; r</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}\text{<span class="notranslate"> \&rsqb;\&lsqb;w</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; w</span>}}<span\; class="notranslate">\; \&rsqb;</span>}{<span\; class="notranslate">\; \{</span>\underset{\mathrm{<span\; class="notranslate">\; r</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}<span\; class="notranslate">\; \&rsqb;</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}\underset{\mathrm{<span\; class="notranslate">\; r</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\underset{\mathrm{<span\; class="notranslate">\; t</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; r</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; w</span>}}<span\; class="notranslate">\; \&rsqb;</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}{<span\; class="notranslate">\; \}</span>}^{\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$

为已知图像皮质区域象素的灰度均值，x，y，z为肾在待匹配三维图像内平移的坐标，w(x+r，y+s，z+t)为皮质平移后在待匹配图像内对应点的灰度值，

为其均值。ρ(x，y，z)为计算所得的相关系数，ρ(x，y，z)∈[-1，1]。Among them, r, s, t represent the three-dimensional coordinates of the pixel in the cortical area, f(r, s, t) is the gray value of the pixel in the cortical area of the known image,

is the average gray value of the pixels in the cortical area of the known image, x, y, z are the translation coordinates of the kidney in the three-dimensional image to be matched, w(x+r, y+s, z+t) is the coordinates of the kidney to be matched after translation Match the gray value of the corresponding point in the image,

for its mean value. ρ(x, y, z) is the calculated correlation coefficient, ρ(x, y, z)∈[-1, 1].

Step 6: The segmentation of cortex and medulla still adopts the region growth segmentation method, and the matched whole kidney region is used as a mask, and only the medulla is grown within the mask. The user selects a suitable seed point in the medulla to obtain the medulla area, and the remaining area can be considered as the cortex. The result of the segmentation of the cortex and medulla makes it possible to accurately match the cortex area as a template.

Step 7: According to the three-dimensional correlation coefficient matching method, using the cortical area as a template, starting from the whole kidney phase, each phase is matched with the nearest matched phase in turn. The results show that each phase is caused by breathing. The motion of is well recognized. The change curve of the contrast agent concentration shows that the curve includes three concentration curves of the whole kidney, the medulla and the cortex (as shown in FIG. 11 ).

As shown in Fig. 1, after the user is satisfied with the obtained graph, the kidney automatic analysis toolbox call returns.

Claims (4)

1. An image processing method based on a magnetic resonance three-dimensional nephrogram comprises the following steps:

(1) Reading a magnetic resonance three-dimensional scanning image of the kidney from a magnetic resonance workstation;

(2) Determining the number of images in each time phase, a background time phase, a cortex and medulla boundary time phase and a whole kidney time phase;

(3) Resecting relevant vessels and a renal pelvis region in a whole kidney phase image;

(4) Identifying a whole kidney contour;

(5) Matching a background time phase with a cortex medulla boundary time phase, and silhouetteing;

(6) Segmenting the cortex and medulla of the kidney;

(7) The images of all phases are matched.

2. An image processing method based on a magnetic resonance three-dimensional kidney image as set forth in claim 1, characterized in that: wherein the step (4) of obtaining the whole kidney contour by adopting a region growing method comprises the following steps:

(1) the user selects one or a plurality of seed points P1, P2, P3, …, pn which fall in the kidney region in the whole kidney time phase;

(2) taking a point Pi from all the seed points as a starting point of the region growth;

(3) calculating the growth threshold value T of the seed point;

(4) calculating the difference of the gray values of each point and I in the neighborhood of the current seed point 26, if the difference is smaller than the growth threshold value of the current seed point, merging the point into a result area, and calculating as a new seed point;

(5) and (4) repeating the operations of the steps (2), (3) and (4) until the result area cannot be increased.

3. An image processing method based on a magnetic resonance three-dimensional kidney image as set forth in claim 2, characterized in that: and judging a seed point growth threshold T by adopting a linear calculation method, and enabling T = I × k, wherein I is an average value of the gray scale of a result region obtained by increasing the current region, and k is a slope set by a user.

4. An image processing method based on a magnetic resonance three-dimensional kidney image as set forth in claim 1, characterized in that: in the step (5), the matching background time phase and the cortical medullary boundary time phase take a cortical region as a template, and the traditional two-dimensional correlation matching based on the correlation coefficient is expanded to three dimensions, and the three-dimensional correlation coefficient calculation formula is as follows:

wherein r, s, t represents the three-dimensional coordinates of the cortical region pixels, f (r, s, t) is the gray scale value of the cortical region pixels of the known image,

the mean value of the gray levels of the pixels of the cortical region of the known image, x, y, z being the mean of the kidney in the three-dimensional image to be matchedThe shifted coordinates w (x + r, y + s, z + t) are the gray values of corresponding points in the image to be matched after the cortex is translated,

is the average thereof. Rho (x, y, z) is the calculated correlation coefficient, and rho (x, y, z) epsilon [ -1,1]。

Images