CN102397070A - A fully automatic method for segmenting and quantifying the left ventricle in cardiac magnetic resonance images - Google Patents

Abstract

The invention discloses a method for fully automatically segmenting and quantifying the left ventricle of a cardiac magnetic resonance image, which comprises the following specific steps: firstly, automatically denoising and edge enhancement processing are carried out on the 4D cardiac magnetic resonance image; automatically and preliminarily determining the center of the left ventricle by utilizing Hough transform, taking the center as an initial seed point to implement a region growing technology to provide a whole blood region of the left ventricle, and taking the centroid of the region as the center of the left ventricle of the current layer; finding the center of each layer by using a seed propagation technology; taking the center of the left ventricle of each layer as an initial seed point, automatically extracting the blood area of the left ventricle of each layer based on an iterative descent threshold region growing technology and calculating the area of the blood area of the left ventricle of each layer; then automatically segmenting the top of the left ventricle according to the space-time continuity of the area of the left ventricle and calculating the area of the top of the left ventricle; positioning the bottom of the left ventricle according to the space-time continuity of the area and the shape of the left ventricle, automatically segmenting the bottom of the left ventricle of the heart by adopting a region growing technology constrained by the shape of the space-time continuous left ventricle, and calculating the area of the bottom of the left ventricle; and finally, realizing the integral segmentation of the left ventricle image. The invention is a fully automatic process without any manual intervention.

Description

A fully automatic method for segmenting and quantifying the left ventricle in cardiac magnetic resonance images

technical field

The present invention belongs to the technical field of magnetic resonance imaging, and specifically refers to a method for automatically and accurately positioning the left ventricle in a 4D cardiac magnetic resonance image, and automatically identifying the bottom and top positions of the left ventricle, thereby realizing fully automatic segmentation and quantification of the left ventricle in the entire cardiac magnetic resonance image method.

Background technique

In recent years, heart disease has become the number one killer that endangers human health. In order to improve the quality of life and reduce the mortality rate of heart disease, a large number of technologies have been developed for clinical prospective diagnosis and treatment of heart disease. Accurate evaluation of heart disease with the help of medical imaging technology has become a routine and effective method for clinical diagnosis of heart disease and formulation of treatment plans. Magnetic resonance imaging has the characteristics of no damage, high soft tissue contrast, wide field of view, and arbitrary tomographic imaging. It has been widely used in clinical medical diagnosis. According to research, cardiac magnetic resonance imaging has high precision and good repeatability, and has been widely used in clinical practice. It has become the gold standard for estimating cardiac function, evaluating myocardial variation, and detecting myocardial scars and congenital heart diseases. The use of cardiac magnetic resonance images can not only observe the morphological structure of the heart, but also estimate the functional state of the heart, which can help doctors make correct judgments on the structure and function of the heart.

Since the left ventricle is the pump body of the systemic blood circulation, left ventricular function indicators, such as left ventricular systolic volume (including end-diastolic volume and end-systolic volume, are the basic indicators for evaluating ventricular shape and function), left ventricular stroke volume (from The volume of blood pumped by the left ventricle into the aorta through the aortic valve is an important indicator reflecting the strength and speed of cardiac contraction) and the left ventricular ejection fraction (the ratio of left ventricular stroke volume to end-diastolic volume, which is an evaluation of heart pump function It is an important reference for clinical diagnosis of heart disease and its curative effect. Quantifying left ventricular function index has become a routine method for clinical diagnosis and treatment of heart disease.

Segmentation of the left ventricle is a prerequisite for quantifying left ventricular function indicators. In standard clinical practice, ventricular segmentation is manually delineated by experienced physicians. However, the amount of clinical image data is large, and manual segmentation is generally only for end-diastolic and end-systolic images, and there are subjective differences when depicting complex myocardial structures such as trabecular muscles and papillary muscles. It can be seen that manually dividing the ventricle is very time-consuming, has low work efficiency, high labor intensity, and poor repeatability. Therefore, automatic efficient and accurate segmentation and quantification of the left ventricle has always been the focus and hotspot of current research.

So far, the left ventricle segmentation algorithm has been developed relatively mature. These algorithms include traditional edge extraction, region growing, and some methods based on specific theories, such as level set algorithm, genetic algorithm and so on. However, these algorithms are almost only effective when segmenting the left ventricle completely surrounded by the myocardium, and there are generally some problems as follows: the left ventricle cannot be automatically positioned accurately and effectively, the bottom and top of the left ventricle cannot be automatically identified, and the bottom of the left ventricle cannot be automatically extracted blood volume. That is to say, before segmenting and quantifying the functional parameters of the left ventricle, it is still necessary to manually locate the left ventricle, manually determine the positions of the bottom and top of the left ventricle, manually close the bottom left ventricle, etc., and manually correct some images with poor segmentation effects . In addition, although some commercial software has been used in clinical practice to replace the traditional purely manual segmentation, due to the limitations of the algorithm, a large amount of manual intervention is still required to complete the quantification of left ventricular function indicators. Therefore, there is still an urgent clinical need for a reliable and effective fully automatic method to further improve the accuracy and repeatability of cardiac function parameters, improve work efficiency, and reduce work intensity.

Contents of the invention

The purpose of the present invention is to address the shortcomings of the above-mentioned prior art, to provide a method for automatically positioning the left ventricle in a 4D cardiac magnetic resonance image, and automatically identifying the bottom and top positions of the left ventricle, thereby realizing fully automatic segmentation and quantification of the left ventricle in cardiac magnetic resonance images method.

Realization of the present invention is accomplished by the following technical solutions:

A method for fully automatic segmentation and quantification of the left ventricle of cardiac magnetic resonance images, characterized in that the specific implementation steps of the method are:

(1) Perform automatic denoising and edge enhancement processing on 4D cardiac magnetic resonance images;

(2) Preliminarily determine the center of the left ventricle, and use the center as the initial seed point to implement the regional growth technique to propose the center of mass of the left ventricular voxel blood area, and determine it as the center of the left ventricle at the current level; implement the seed propagation technique to estimate the remaining level left ventricular center;

(3) With the center of the left ventricle determined in step (2) as the initial seed point, the region growing technique based on the iterative descending threshold is used to automatically propose the left ventricle blood region and calculate the corresponding area for the middle layer of the left ventricle;

(4) With the center of the left ventricle determined in step (2) as the starting seed point, the region growing technology based on iterative falling threshold is used to automatically extract the blood region of each layer image at all time phases for the middle layer of the left ventricle and calculate Corresponding area, automatically locate the top of the left ventricle of the heart according to the temporal and spatial continuity of the area change of the extracted area, and correct and estimate the area of the top of the left ventricle;

(5) Locate the bottom of the left ventricle according to the space-time continuity of the area and shape of the left ventricle, and automatically segment the bottom of the left ventricle of the heart using the shape-constrained region growing technology and automatically correct the segmented area;

(6) Calculate the left ventricular function index according to the area of the left ventricle: including left ventricular systolic volume, stroke volume, ejection fraction, and draw the filling curve.

Preferably, the processing of the 4D cardiac magnetic resonance images in the step (1) adopts anisotropic diffusion method.

Preferably, in the step (2), before the whole voxel blood area of the left ventricle is proposed, Hough transform is performed on the end-diastolic and end-systolic subtraction images of the middle level of the left ventricle.

Preferably, the seed propagation technique in the step (2) is based on the initial seed point of the left ventricle center that is close to the level that has been determined as the current level, implements the region growing technique to extract the whole blood voxel left ventricle region, and converts the The regional center of gravity is determined as the center of the left ventricle in the current slice.

Preferably, in the step (4), the center of the left ventricle is multiplied from the level of the middle of the left ventricle along the direction of the top of the left ventricle, and the center is used as the starting seed point to automatically extract the area growth technology based on the iterative falling threshold method. Blood regions in each image layer at all time phases.

The continuous falling threshold of the threshold can be obtained by the following formula:

th=u _b /r

In the above formula, μ _b is the mean value of the blood area, r is a variable, the initial value is 1.0, and it increases with a step size of 0.05.

Preferably, in the step (4), when there is a jump in the area of a part of the phase growth region in a certain layer, the following formula is used to estimate the area according to the time continuity:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; ms</span>}}{\overset{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; ms</span>}}{\overset{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>$

Among them, s represents the sth layer, p represents the time phase where the jump occurs, q represents the time phase closest to p without the area jump, and ms represents the number of layers in the middle layer of the left ventricle. Preferably, in the step (4), the area of all phase growth regions of a certain layer undergoes a jump, and the top of the ventricle is regarded as a cone or an ellipse cone, and the area is estimated by the following formula according to the spatial continuity:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 9</span>}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 12</span>}\sqrt{\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}$

Among them, s represents the sth layer, and p represents the time phase when the jump occurs.

Preferably, in the step (5), the center of the left ventricle is multiplied from the middle layer of the left ventricle along the direction of the bottom of the left ventricle, and the center is used as the starting seed point to automatically extract the area growth technology based on the iterative descending threshold method. Blood regions in each image layer at all time phases.

The advantage of the present invention is that the method of the present invention can accurately and quickly realize the automatic positioning and segmentation of the entire left ventricle of the magnetic resonance image of the heart. Compared with the current existing methods, the present invention adopts a fully automatic automatic segmentation and quantification method, which can automatically complete the segmentation of the left ventricle and the calculation of the functional parameters of the left ventricle without any manual intervention. The functional parameters include left ventricle systolic and systolic Volume, ejection fraction, stroke volume, and filling curve, because the previous methods need to manually segment the bottom and top of the heart, which is highly subjective, and the results obtained by different people will vary greatly .

Description of drawings

Figure 1 is a graph showing the volume change of the extracted region;

Fig. 2 is an original image of the middle part of the left ventricle and its segmented image in the embodiment (the left is the original image, the right is the segmented image);

Fig. 3 is the change of the area and the shape of the left ventricle in time and space in the embodiment (the left is spatial continuity, the right is temporal continuity);

Figure 4 is a three-dimensional graph showing the temporal and spatial continuity of the left ventricular area;

Fig. 5 is an original image of the top of the left ventricle and its segmented image (the left is the original image, the right is the segmented image) in the embodiment;

Fig. 6 is an original image of the bottom of the left ventricle and its segmented image in the embodiment (the left is the original image, the right is the segmented image);

Fig. 7 is the filling curve of the left ventricle in the embodiment (the horizontal axis is the phase of the left ventricle, and the vertical axis is the volume of the left ventricle).

Detailed ways

The features of the present invention and other related features will be further described in detail below in conjunction with the accompanying drawings through the embodiments:

The method of the present invention accurately and efficiently quantifies the left ventricle function index of the 4D cardiac magnetic resonance image (CMRI) automatically without any manual intervention. The following example introduces step by step the specific operation process of the method of the present invention to automatically locate the left ventricle, automatically determine the top and bottom positions of the left ventricle, and automatically segment and quantify the left ventricle.

The magnetic resonance imaging data collected in this embodiment is cardiac magnetic resonance imaging data. The data come from GE Signa 1.5T magnetic resonance imaging system, and the imaging sequence selected is SSFP sequence. Specific imaging parameters: TR 3.3-4.5ms, TE 1.1-2.0ms, flip angle 55-60, matrix size 192×192 256×256, image size 256×256, receiving bandwidth 125kHz, field of view (FOV) 290-400×240 -360, slice thickness and slice spacing are 6-8mm and 2-4mm respectively (total 10mm), the left ventricle of each data has 6-10 slices, 20-28 cardiac phases.

(1) Iterative falling threshold method to extract blood

Firstly, the whole blood voxels are extracted by region growing technique from the seed point, and the mean and standard deviation (μ _b and σ _b ) of the whole blood sample region are calculated. Then take μ _b as the initial threshold to implement a series of region growing techniques based on the continuous decrease of the threshold (th) to extract a series of growth regions until the region suddenly breaks out from the myocardium. The continuous falling threshold can be obtained by μ _b and a variable, specifically using the following formula:

th=u _b /r

①

In the above formula, r is a variable with an initial value of 1.0 and increments with a step size of 0.05. Fig. 1(a) depicts that the area of the extracted region changes continuously with the continuously decreasing threshold value until abruptly from the perspective of the region area. The horizontal axis represents the variable r, and the vertical axis represents the area (expressed in the number of voxels). According to the threshold value and σ _b before the volume mutation, an optimal value can be found for segmenting the left ventricle. Fig. 2 shows an original image of the middle part of the left ventricle and the segmented image obtained based on the continuous descending threshold method in the embodiment.

(2) Cardiac MRI images of left ventricular center propagation

In the middle part of the heart, the region growing technique is performed on the current layer with the center of the left ventricle in the adjacent layer as the starting point, and the centroid of the growth area is determined as the center of the left ventricle in the current layer; at the top and bottom of the heart, the center of the left ventricle in the adjacent layers is determined is the center of the left ventricle at the current level.

(3) Segmentation of the top of the left ventricle and its area estimation

The area of the left ventricle varies continuously in time and space. Figure 3 illustrates the spatiotemporal continuum of the area and shape of the left ventricle. Figure 4 shows the spatio-temporal continuity with the values of the left ventricle area in the examples. As shown in Figure 4, the y-axis represents the number of phases, the x-axis represents the number of data layers collected, and the z-axis represents the area of the growth region. In this embodiment, there are 12 layers and 20 time phases in total. The 12 spline curves in the figure represent the 12 layers, and each represents the change of the area of the growth region extracted from each layer over time. The top of the heart is automatically segmented according to the spatial-temporal continuity of the area change of the extracted region, and the area of these positions is estimated according to the jump of the top. There are two cases:

(a) There is a jump in the area of some phase growth regions in a certain layer, and the area is estimated by the following formula according to the time continuity:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; *</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; ms</span>}}{\overset{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; ms</span>}}{\overset{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; q</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>$

②

Among them, s represents the sth layer, p represents the time phase where the jump occurs, q represents the time phase closest to p without the area jump, and ms represents the number of layers in the middle layer of the left ventricle.

(b) The area of all phase growth regions in a certain layer changes abruptly. The top of the ventricle is regarded as a cone or ellipse, and the area is estimated by the following formula according to the spatial continuity:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 9</span>}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 12</span>}\sqrt{\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>}$

③

Here, s represents the sth layer, and p represents the phase of the transition.

Fig. 5 shows an original image of the apex of the left ventricle and the segmented image obtained according to the temporal-spatial continuity in the embodiment.

(4) Location and estimation of the bottom of the left ventricle

The bottom position of the left ventricle was determined according to the temporal and spatial continuity of the area change of the extracted region and the change of the center of gravity. And use the shape-constrained region growing technology to automatically segment the bottom of the left ventricle of the heart and correct the segmented region. Fig. 6 shows an original image of the bottom of the left ventricle and the segmented image obtained according to the temporal-spatial continuity in the embodiment.

(5) Left ventricular filling curve

After the automatic segmentation of the left ventricle is completed, the functional parameters of the left ventricle can be quantified according to the area of the segmented area, including left ventricular systolic and systolic volume, stroke volume, ejection fraction, and the filling curve can be drawn. Fig. 7 depicts the left ventricular filling curve of the embodiment, the horizontal axis is the time phase, and the vertical axis is the left ventricular volume.

The above are only preferred embodiments of the present invention, and it should be pointed out that those skilled in the art can make several similar deformations and improvements without departing from the inventive concept of the present invention. Should be considered within the protection scope of the present invention.

Claims (8)

1. automatically cut apart the method that quantizes cardiac magnetic resonance image left ventricle for one kind, it is characterized in that the practical implementation step of this method is:

(1) 4D cardiac magnetic resonance image is carried out automatic denoising and edge enhancement process;

(2) tentatively confirm the left ventricle center, be that initial seed point is implemented all plain this regional barycenter of blood regions calculating of region growth technique proposition left ventricle with this center, and it is confirmed as current aspect left ventricle center; Implement seminal propagation technology estimated remaining aspect left ventricle center;

(3) be initial seed points with the left ventricle center of confirming in the step (2), adopt region growth technique to propose the left ventricle blood regions automatically and calculate respective area based on the iteration falling-threshold value to left ventricle middle part aspect;

(4) be initial seed points with the left ventricle center of confirming in the step (2); Left ventricle middle part aspect adopted based on the region growth technique of iteration falling-threshold value extract each tomographic image automatically in the blood regions that is gone up mutually sometimes and calculate respective area; Space and time continuous property according to extracting the region area variation is located the heart left ventricle top automatically, and proofreaies and correct the area at estimation left ventricle top;

(5) according to left ventricle area and shape space and time continuous property location left ventricle bottom, utilize the region growth technique that receives shape constraining to cut apart automatically bottom the heart left ventricle and and carry out from dynamic(al) correction to the zone that is partitioned into;

(6) calculate the left ventricular function index according to the area of left ventricle: comprise left ventricle volume, whenever rich output, the ejection fraction of contracting that relax, and describe full curve.

2. a kind of method that quantizes cardiac magnetic resonance image left ventricle of automatically cutting apart according to claim 1 is characterized in that, in the said step (1) 4D cardiac magnetic resonance treatment of picture is adopted anisotropic diffusion method.

3. a kind of method that quantizes cardiac magnetic resonance image left ventricle of automatically cutting apart according to claim 2; It is characterized in that; In the said step (2) before proposing all plain blood regions of left ventricle, to left ventricle middle part aspect diastasis and end-systole subtraction image carrying out the Hough conversion.

4. a kind of method that quantizes cardiac magnetic resonance image left ventricle of automatically cutting apart according to claim 2; It is characterized in that; Seminal propagation technology in the said step (2) is to confirm to such an extent that the left ventricle center is the initial seed point of current aspect to close on aspect; Implement region growth technique and extract whole blood voxel left ventricle zone, and this regional barycenter is confirmed as the left ventricle center of current aspect.

5. a kind of method that quantizes cardiac magnetic resonance image left ventricle of automatically cutting apart according to claim 4; It is characterized in that; Adopting in the said step (4) earlier from the left ventricle middle part aspect to carry out the breeding of left ventricle center along the left ventricle top-direction, is that initial seed points is carried out and come to extract automatically each tomographic image in the blood regions that is gone up mutually sometimes based on the region growth technique of iteration falling-threshold value method with this center.

The continuous falling-threshold value of said threshold value can be obtained by following formula:

th＝u _b/r

In the above-mentioned formula, μ _bBe the average of blood regions, r is a variable, and initial value is 1.0, increases progressively with step-length 0.05.

6. a kind of method that quantizes cardiac magnetic resonance image left ventricle of automatically cutting apart according to claim 5; It is characterized in that; Working as certain one deck in the said step (4) has part phase place growth district area generation transition, adopts following formula estimated area according to time continuity:

$A(p,s)=A(q,s)*\underset{i=\mathrm{ms}}{\overset{s-1}{\mathrm{\&Sigma;}}}A(p,i)/\underset{i=\mathrm{ms}}{\overset{s-1}{\mathrm{\&Sigma;}}}A(q,i)$

Wherein, s represents the s layer, and the time phase of transition takes place in the p representative, and the q representative is from the nearest time phase that area transition does not take place of p, and ms represents the number of plies of left ventricle intermediate surface.

7. a kind of method that quantizes cardiac magnetic resonance image left ventricle of automatically cutting apart according to claim 5; It is characterized in that; Transition all takes place in all phase place growth district areas of certain one deck in the said step (4); The ventricle top is regarded as circular cone or elliptic cone, adopts following formula estimated area according to spatial continuity:

$A(p,s)=9A(p,s-2)+4A(p,s-3)-12\sqrt{A(p,s-2)\&times;A(p,s-3)}$

Wherein, s represents the s layer, and the time phase of transition takes place in the p representative.

8. according to claim 6 or 7 described a kind of methods that quantize cardiac magnetic resonance image left ventricle of automatically cutting apart; It is characterized in that; Adopting in the said step (5) earlier from the left ventricle middle part aspect to carry out the breeding of left ventricle center along the left ventricle bottom direction, is that initial seed points is carried out and come to extract automatically each tomographic image in the blood regions that is gone up mutually sometimes based on the region growth technique of iteration falling-threshold value method with this center.

Images