WO2013131420A1 - 医学影像的目标区域边界确定装置和方法 - Google Patents

医学影像的目标区域边界确定装置和方法 Download PDF

Info

Publication number
WO2013131420A1
WO2013131420A1 PCT/CN2013/071129 CN2013071129W WO2013131420A1 WO 2013131420 A1 WO2013131420 A1 WO 2013131420A1 CN 2013071129 W CN2013071129 W CN 2013071129W WO 2013131420 A1 WO2013131420 A1 WO 2013131420A1
Authority
WO
WIPO (PCT)
Prior art keywords
area
sub
threshold
region
medical image
Prior art date
Application number
PCT/CN2013/071129
Other languages
English (en)
French (fr)
Inventor
李澎
袁昕
陈功
Original Assignee
杭州弘恩医疗科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 杭州弘恩医疗科技有限公司 filed Critical 杭州弘恩医疗科技有限公司
Priority to US14/383,060 priority Critical patent/US20150023578A1/en
Publication of WO2013131420A1 publication Critical patent/WO2013131420A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0012Biomedical image inspection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/11Region-based segmentation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/136Segmentation; Edge detection involving thresholding
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/60Analysis of geometric attributes
    • G06T7/62Analysis of geometric attributes of area, perimeter, diameter or volume
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2200/00Indexing scheme for image data processing or generation, in general
    • G06T2200/04Indexing scheme for image data processing or generation, in general involving 3D image data
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10016Video; Image sequence
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10132Ultrasound image
    • G06T2207/101363D ultrasound image
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20092Interactive image processing based on input by user
    • G06T2207/20101Interactive definition of point of interest, landmark or seed
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing
    • G06T2207/30048Heart; Cardiac

Definitions

  • the present invention relates to a method and apparatus for determining a target region boundary of a medical image, and determining a physiological parameter using the determined boundary of the target region. More specifically, the present invention relates to the determination of cardiac physiology parameters based on real ultrasound image data. Background technique
  • Medical imaging has become an integral part of modern medicine, and its application throughout clinical work is not only widely used for disease diagnosis, but also plays an important role in planning design, program implementation, and efficacy evaluation of surgery and radiation therapy.
  • medical images can be divided into two parts: an anatomical image and a functional image.
  • Anatomical images primarily describe human morphology information, including X-ray transmission imaging, CT, MRI, US, and others.
  • the use of computer technology for quantitative analysis of medical images has become an important technical improvement direction, to increase the objectivity of diagnosis, and easier to grasp the operation, can reduce the experience of readers.
  • Dependency exclude differences in judgment between different readers.
  • the art desires to more accurately quantify the physiological parameters of the heart based on cardiac image photographic sequences, such as ventricular volume, myocardial mass, cardiac wall thickening, cardiac ejection fraction (EF value), and the like. Accurate acquisition of cardiac ejection fraction is of great significance. According to the cardiac ejection fraction, the ejection ability of the heart can be estimated, which is an important parameter for judging cardiac function.
  • 3D ultrasound is a non-invasive imaging technique. It has the characteristics of fast imaging speed and low cost in the exploration of heart diseases. Therefore, it is the most widely used in the diagnosis and treatment of heart disease.
  • Analysis of cardiac chamber volume, ejection fraction, myocardial volume and mass in 3D ultrasound images is an important basis for diagnosis.
  • echocardiography contains a lot of noise, and the intimal and myocardial edges of the heart chamber are irregular (especially the diseased heart chamber and myocardium), it is difficult to calculate the relevant quantitative calculations.
  • One of the difficulties is how to accurately obtain the boundaries of the endocardium and how to accurately measure the irregular changes in the heart.
  • the field has long been committed to improving the accuracy and operability of ultrasound images to acquire physiological parameters.
  • the current cardiac ejection fraction (EF value) method which is currently used clinically, defines some control points in an interactive manner and uses mathematical modeling to approximate a cardiac cavity using a series of simulated geometries. inaccurate.
  • JP2002085404 entitled “ultrasonic imaging processor” » teaches to divide the heart chamber
  • EP 123617 teaches the use of segmented curves to describe the heart chamber.
  • JP2008073423 teaches interpolating a reference contour of more than 50 image sets to obtain an approximate cavity.
  • EP1998671 (A1) teaches the use of the mouse to point out several control points and match a template to achieve automatic segmentation.
  • EP 2030042 (A1) teaches a manual marking of a small number of control points, combined with a trained template to obtain an endocardium.
  • the prior model is a model based on statistics, which means that the data set to be analyzed obeys an unknown probability distribution and has a definite relationship with the data set of a known sample. In order to find this unknown distribution, it is necessary to calculate the probability distribution of the observed sample data set. This previously calculated probability distribution or parameter is called the prior model.
  • the diseased heart is generally no longer a heart chamber that can be estimated using the above model, compared to the normal heart chamber.
  • the heart chamber of the diseased heart has an unpredictable shape change, and the endocardium is irregular (eg, tumor occupying ventricular aneurysm, thickened heart wall). Changes in the shape of the heart chamber lead to reduced ejection function, heart valve insufficiency and other symptoms.
  • the prior shape model of the heart cavity is obtained by pre-calculating the multi-frame image, and the heart cavity on the current image is corrected by comparing with the approximate geometric model of the heart cavity on the current image.
  • this type of prior model is calculated from normal cardiac calculations. In practical clinical applications, it is difficult to ensure accurate results for diseased hearts.
  • Paragios N A level set approach for shape-driven segmentation and tracking of the left ventricle. Medical Imaging, 2003, Page(s): 773 776 uses the level set algorithm as the main body of the left ventricular segmentation algorithm, plus the use of a large number of Knowledge, that is, the correct left ventricular segmentation results. Use a priori experience combined with the characteristics of the image itself to develop a speed function for the level set and a restricted area. Thereby achieving the purpose of left ventricular segmentation.
  • CN101404931A (Application No. CN200780009898.7), entitled “Quantitative Ultrasound Diagnosis by Myocardial Function", teaches manually setting control points first, and then connecting control points by curves according to image gradients, thereby achieving the purpose of approximate tracing.
  • CN101454688A (Application No. C ⁇ 07 18854.0), entitled “Quantification and Display of Cardiac Wall Thickness”, discloses the distance, wall thickness variation and strain of the designated myocardial location point by speckle tracking. No single myocardium was obtained. This technique uses image gradients to determine the endocardial border, which is inaccurate if the image noise increases. Because the epicardium has no clear gradient, it is often missing and inaccurate when it is automatically determined.
  • the patent provides a tool to manually adjust the two boundaries at the beginning and end of the cardiac cycle, and then automatically set the points to be tracked between the two boundaries, they are located on the myocardium, and then record the pixels around each point as The speckle pattern, the speckle pattern between different frames for maximum correlation block matching, so that the motion of each point can be tracked.
  • speckle tracking is susceptible to noise
  • the diseased myocardium Compared with normal myocardium, the diseased myocardium has lesions such as dilatation, contractility, and hypertrophy, which ultimately affects its ability to contract, specifically in the change of its elastic deformation parameters. In terms of geometry, it will change with normal myocardium, and thus may have irregular boundaries.
  • the present invention is directed to an apparatus and method for more efficient and accurate image processing and calculation based on existing medical imaging techniques to improve and improve the volume and ejection fraction of the heart chamber.
  • the accuracy of related physiological parameters such as myocardial volume and quality, thus helping to make correct and timely diagnosis during clinical treatment.
  • a first aspect of the invention relates to a target region boundary determining device for a medical image And configured to distinguish a target region boundary according to a corresponding physical quantitative feature reflected by a tissue distribution in the medical image, the device comprising:
  • the sub-area is set as follows: dividing the area to be analyzed into a plurality of adjacent sub-areas overlapping each other, and each adjacent sub-area The stacked or non-overlapping regions collectively and completely cover the target area.
  • the device wherein the position of the selected target area is centered, the radius r is set, a circular area is defined, and the physical quantitative feature distribution in the circular area is analyzed to determine Threshold.
  • the medical image is a medical image of the heart, selecting a position point of the heart chamber as a center, defining a circular area with a radius of 5 mm, and calculating a pixel gray in the circular area
  • the average value of the degree values is taken as the threshold value; and, if the pixel gray level in a sub-area is smaller than the threshold value, the sub-area is marked as the target area, that is, the area of the heart chamber.
  • the medical image is a medical image of the heart
  • the area to be analyzed is divided into a series of overlapping circular regions, wherein the circular region is the sub-region,
  • the radius of the circle is lmm
  • the distance between the centers of the circles is also 1 mm
  • the average value of the gray areas of each of the sub-areas is calculated. If the average value is greater than the threshold, the pixels in the sub-area are marked. It is a myocardial area, otherwise it is marked as a non-myocardial area.
  • the device according to the eighth aspect of the present invention wherein the medical image is a cardiac image, and the target region is any one of the heart chambers or the myocardium.
  • the gradation of the pixel or voxel is preferred.
  • the average gray scale is a feature measurement method, and its measurement speed is faster.
  • the gradient distribution of the region can also be examined, which is another simple and efficient feature measurement.
  • the above object of the invention is to determine the boundary of a medical image target area by adopting a more accurate and effective method.
  • the invention can be used to process more accurate quantitative physiological parameters when processing real 3D ultrasonic medical images.
  • Real 3D ultrasound medical images are 3D images directly generated by 3D ultrasound probes. In ultrasound 3D images, boundary determination of endocardial borders, etc., is of great importance in determining cardiac-related physiological parameters.
  • the inventors of the present invention have been inspired by tissue perfusion methods commonly employed in medical imaging.
  • the contrast agent diffuses in the cavity and the gap of the human tissue, and when the digitized image is formed, the contrast between the anatomical tissue and the contrast agent data is generated, which makes the image operator easy. Identify the boundaries of various anatomical structures to provide a reliable basis for further quantitative diagnosis.
  • the inventor proposed a virtual tissue perfusion model by means of computer technology.
  • the sp using the physical quantitative characteristics reflected by the pixels contained in the image, helps the person skilled in the art to effectively identify the boundaries of the tissue to obtain a more accurate tissue boundary; further, to obtain the quantitative physiological parameters of the heart based on the accurate boundary.
  • the present invention utilizes computer technology to extract interest tissue boundaries from digitized images. There is a clear contrast between the pixels or voxels around the boundaries of the tissue of interest, but the boundaries are unclear due to the effects of granular noise.
  • the inventors have specifically examined the characteristics of the pixels in the image, and set a cell or sub-region in the region to be analyzed, which is filled with pixels inherent to the padding sub-region of the minimum basic unit, and therefore, is assumed to be a "pixel padding unit".
  • a point investigation point is set on the area to be analyzed, and a circular or elliptical sub-area around the point is a unit or a pixel filling unit, and the sub-areas overlap each other, and the pixel value or the voxel value in the sub-area is analyzed.
  • the distribution feature from which a fixed or non-fixed threshold is derived, and according to this threshold, each pixel or voxel in the area around each inspection point is marked to obtain the tissue region of interest, its boundary It is the boundary of the organization of interest.
  • the design algorithm can be used to set the inspection point, and a plurality of circular or elliptical regions of different scales or sizes can be used to further analyze the distribution of pixel values or voxel values. Further refine the boundaries of the tissue area of interest.
  • the present invention utilizes computer technology to directly utilize the regional characteristics of different related images by utilizing the tissue in the image, for example, the difference in physical characteristics between the heart chambers of the heart, and the differences in tissue characteristics reflected in the medical images.
  • the operator selects the approximate middle position of the area according to experience, and uses computer technology to determine the physical characteristics of the area, for example, the average value of the gray scale, the gradient value, etc., and divides the area and the boundary into two categories by threshold comparison, BP, The effect of image binarization is achieved, thereby distinguishing the boundaries.
  • This distinction is more objective and accurate, avoiding the limitations of the prior model segmentation of the heart chamber and myocardial methods.
  • FIG. 1 is a schematic diagram of approximate heart cavity segmentation results on a typical conventional image processing device
  • FIG. 2 is a schematic illustration of an embodiment of the present invention employing interactive selection of a target heart chamber
  • Figure 3A illustrates the boundary of the heart chamber marked by the method of the present invention
  • FIG. 3B is a schematic diagram of a heart cavity volume change curve of each frame of a time series, from which the maximum volume TM « and the minimum volume «TM of each frame image in each cardiac cycle can be seen ;
  • Figure 4 is a flow chart of a particular embodiment of the invention. detailed description
  • boundary processing proposed by the present invention for an organization of interest or region can have many different applications.
  • the description of the specific embodiments is intended to assist those skilled in the art to understand the invention and should not be construed as limiting.
  • the analysis is mainly made by taking the pixel gradation as a physical quantitative feature as an example.
  • Other suitable physical quantitative features can also be applied to the present invention.
  • the boundary processing of the present invention includes the following steps:
  • the preliminary markup obtains one or more connected areas, and then integrates the area of interest (ROI), or the target area, for further processing, that is, only the connected area containing the operator's mouse click is retained, and other areas are Discard, or say, unmark other areas. This gives the result of the initial split.
  • ROI area of interest
  • the boundary refinement process ⁇ first marks the boundary of the segmented region separately, and then arranges the pixel padding cells on the boundary, and sets the pixel padding cells to cover smaller regions. , can be half the size of the pixel-filled cells in the first step, they still need to overlap each other. Also on these areas Quantitative features, such as average grayscale or gradient, are calculated, and a threshold is obtained. Each pixel filling unit is marked according to the threshold, and an OR operation is performed with the initial divided region result, and the refined regional result is obtained.
  • further refinement can be performed, for example:
  • the operator repeats step 3 according to clinical needs, and can further refine the boundaries by further reducing the size of the pixel fill unit until a satisfactory result is obtained.
  • the so-called three-dimensional data is made up of the previous sliced images.
  • the stacking of the boundaries of the previously obtained slice is represented as a surface in the 3D data.
  • the voxel fill unit has the same settings as when you performed the third step last time, that is, with the same radius setting, they still need to overlap each other.
  • quantitative features are calculated based on the pixel gray value on these regions, and a threshold is obtained, each pixel filling unit is marked according to the threshold value, and an OR operation is performed with the regional result obtained in the last third step, and the combined regional result is obtained.
  • Regarding the processing of the heart chamber boundary it is basically the same as the above description, and further, it is necessary to add the following processing in step 2 thereof:
  • step 2 On the area obtained in step 2, mark the boundary separately, and then divide the boundary into a series of overlapping circular areas, the center of the circle is the point on the boundary, and the radius is the radius of the circular area in the first step. Half of it. The average value of the pixel gray value on each circular area and the average value of the pixel gray gradient mode are calculated. Then calculate the average of these values to get two thresholds:
  • n is the number of circular regions.
  • the grayscale average reflects the mean of the grayscale mean;
  • the mean value of the mode reflects the mean of the mean of the gradient mode, and the change of the number of pixels in the analysis area. It reflects the size of the pixel change in this area. As the boundary, the value will become larger, and less than this value indicates that it is still within the boundary.
  • It is marked that the condition is that the gray level mean of a sub-area is less than the threshold of the gray mean, and the gradient mode mean is also smaller than the W value of the gradient mode mean.
  • the pixels in the circular area are marked as heart chamber regions, otherwise marked as non-cardiac chamber regions.
  • the heart chamber area marked in this step and the heart chamber area marked in the second step are ORed together to obtain a refined heart chamber area.
  • the operator repeats step 3 according to clinical needs.
  • the radius of the circular area used each time is half of the radius of the last used circular area, to further refine the boundary until the result on the satisfactory 2D slice is obtained. until.
  • the 2D slice map is stacked into 3D data
  • the fourth step is that the heart cavity regions on each 2D slice map are simultaneously stacked into 3D regions.
  • the center of the sphere is the point on the boundary surface, and the radius is the last used circular area in the fourth step. radius.
  • the average value of the voxel gray value on each spherical region and the average value of the voxel gray gradient mode are calculated.
  • the average of the gray scale and the gradient mode average of the pixels are obtained by calculating the average of these values.
  • n is the number of spherical regions.
  • the grayscale average and the gradient mode average of each spherical region are examined, and the pixels in the spherical region are marked as the cardiac cavity region, otherwise marked as the non-cardiac cavity region.
  • the heart chamber area marked in this step and the heart chamber area marked in the fourth step are ORed, and the refined 3D area of the heart chamber is obtained.
  • the present invention is applied to real three-dimensional (3D) ultrasound image data processing for a patient's palpitations, which is used in this embodiment to acquire a cardiac chamber volume and an ejection fraction.
  • Step 1 Obtain medical image data of the patient by using an ultrasound imaging device.
  • the heart region is scanned using a real 3D ultrasound probe to obtain multiple time series of 3D ultrasound images.
  • Each time series contains a series of frames, and one or more complete cardiac cycles are recorded, each frame containing There are 3D voxel data composed of multiple slices.
  • the imaging devices used are, for example, the Siemens SC2000 echocardiograph and the Philips IE33.
  • Step 2 Extract the contour of the heart cavity in all slice images of all frames in the real 3D ultrasound image time series.
  • 5 8 time series are scanned, one time sequence has ⁇ 44 frames, one frame has 256 slice images, and each image has a size of 256*256 pixels.
  • the slice image is selected based on the heart chamber containing the most interesting and exposed.
  • the location of the mouse click can be clearly determined and is clearly within the heart chamber.
  • the operator clicks on the slice map with the mouse, and the position of the click is required to be inside the heart cavity of interest.
  • the width direction is the X axis
  • the positive direction is the right direction
  • the height direction is the y axis
  • the positive direction is the downward direction
  • the X, y coordinates thus obtained.
  • the purpose of setting coordinates is to describe the position of each pixel or voxel in space, which is uniquely determined by the coordinates (x, y) or (x, y, z).
  • the purpose of using coordinates is mainly used to determine the adjacency relationship between pixels or voxels (8 neighborhoods or 4 neighborhoods on 2D images, 6 neighborhoods and 26 neighborhoods on 3D images), for setting The determination of the extent of the filling unit, as well as the marking of the heart chamber of interest (the perfusion area covering the heart chamber of interest forms a connected abutment relationship between them after being labeled, so that a single heart chamber can be separated).
  • an automatic association processing unit may be additionally provided.
  • a slice When a slice is clicked, all slices of the 3D image of the frame are automatically associated, and only one slice is required for each frame, and other slices are automatically processed.
  • the image range of an ultrasound contains the region of interest and noise (non-interest zone), the only region that is not ideal. Due to the limitations of the actual effect, the operator asks to confirm (click) the region of interest as the entire technology.
  • the center of the heart chamber as the center, define a circular area with a radius of 5 mm, according to the voxel resolution of the 3D ultrasound image (ie the distance between the voxel center points in the three directions of x, y, z) , in mm, converted to the range of the circular area in pixels, and the average value of the pixel gray value in the circular area is calculated as a model parameter, that is, the threshold parameter t.
  • each circular area can be regarded as a sub-area of pixel fill of an image.
  • the distribution of pixel values in each circular region is analyzed, and the cardiac cavity is marked by the threshold segmentation method according to the threshold parameter t, that is, each circular region is marked as a cardiac cavity region and a non-cardiac cavity region, respectively.
  • thresholds are divided for all pixels of the slice using the threshold calculated according to step b). Since the pixel gray value of the region where the heart chamber is located is low, it is necessary to mark the pixel smaller than the threshold in the slice map as the cardiac cavity region.
  • the slice map is first divided into a series of overlapping circular regions as sub-regions or pixel-filled regions, the radius of the circle is 5 mm, and the distance between the centers of the circles is also 5 mm, according to step b) The method in the method is converted to a circular area range in pixels. Then, the average value of the gray scales of all the pixels in the area is calculated.
  • the pixels in the circular area are marked as the cardiac cavity area, otherwise they are marked as non-cardiac cavity areas.
  • the connected graph is checked in the 8-neighbor manner of the marker map, and the connected domain containing the position of the cardiac chamber marked by the operator is taken as the segmentation result of the cardiac chamber of interest. . Finally, the same threshold segmentation is performed for all slices on all frames of an image time series.
  • Step 3 Calculate the volume of the heart chamber and the EF value based on the marked area of the heart chamber.
  • each pixel On the marked heart cavity area, use the neighborhood check method to determine whether each pixel is an inner point or a boundary point. If it is a boundary point, it is marked as white, and other points are marked as black, thereby obtaining an irregular endocardial boundary. .
  • N is the total number of pixels on the boundary
  • miller is the maximum value of the grayscale gradient mode of the pixel on the boundary
  • / min is the minimum value of the grayscale gradient mode of the pixel on the boundary
  • Heart chamber volume Where S is the total number of slices on the image of the frame, « « « ⁇ ,. is the number of pixels in the endocardial border of each slice, ⁇ 2, is the number of pixels on the endocardial border of each slice, and Is the distance between the voxel center points in one of the x, y, and z directions of a frame of images, in mm.
  • Steps 1 and 2 of Embodiment 4 are the same as Embodiment 1 described above, and thus will not be specifically described.
  • steps a), b), c) in step 2 are repeated to identify other cardiac chamber regions on the slice map for use in the cardiac exclusion step in subsequent myocardial segmentation.
  • Other areas of the heart chamber refer to similar segmentation operations on other incompletely exposed, unclear heart chambers, with the goal of marking all heart chambers so as not to affect the division of the heart muscle.
  • This step is an additional pre-treatment step before myocardial segmentation, with the goal of eliminating all heart chambers.
  • Step 3 Extract the myocardial contour from all slice images of all frames in the real 3D ultrasound image time series.
  • a) Use the mouse to click on multiple myocardial locations of interest.
  • the operator clicks on the slice map with the mouse, and the position of the click is required to be near the edge inside the myocardial (target myocardial) of interest.
  • the X-coordinate and y-coordinate of the position point are recorded with the upper left corner of the image as the origin. There may be more than one myocardial location point of interest.
  • the pixel at the point of the myocardial position clicked with the mouse does not reflect the distribution range of the pixel gray value in the myocardium, and the pixel average value in a neighborhood around the selected position point can be used to obtain a more accurate gray value.
  • Estimated distribution Therefore, with the center of the myocardial position as the center, with a radius of 1 mm, define a circular area, according to the voxel resolution of the 3D ultrasound image (ie, the distance between the voxel center points in the three directions of x, y, z, In mm, converted to the range of the circular area in pixels, the average value of the pixel gray value in the circular area is calculated as a model parameter, that is, the threshold parameter t.
  • the myocardium is marked by the method of threshold segmentation.
  • This step is based on the threshold parameter t calculated in step b, threshold-dividing all the pixels of the slice, and excluding the pixels of all the regions where the heart chamber is obtained in step 2 and the additional step.
  • the slice map is first divided into a series of overlapping circular regions, that is, pixel filling units (cells).
  • the radius of the circle is lmm, and the distance between the centers of the circles is also lmm, which is converted into the area of the circular area in pixels according to the method in step b.
  • the average value of the gray scales of all the pixels in the region is calculated. If the average value is larger than the threshold value t, the pixel points in the circular region are marked as myocardial regions, otherwise they are marked as non-myocardial regions.
  • the connected map is examined in a manner of 8 neighborhoods, and the connected domain containing the myocardial position points marked by the operator is used as the segmentation result of the myocardial of interest. Finally, the same threshold segmentation is performed for all slices on all frames of an image time series.
  • Step 4 Calculate myocardial volume and mass based on the marked myocardial area. a) The boundaries of the individual myocardium are obtained from each of the marked myocardial regions.
  • the neighborhood check method is used to determine whether each pixel is an inner point or a boundary point. If it is a boundary point, it is marked as white, and other points are marked as black, thereby obtaining an irregular myocardial boundary.
  • Each pixel on the myocardial boundary calculates a weight that affects the number of pixels on the border of the myocardium.
  • is the total number of pixels on the border of the myocardium
  • is the maximum value of the gray scale gradient mode of the pixel on the myocardial boundary
  • / min is the minimum value of the gray scale gradient model of the pixel on the myocardial boundary
  • / is the each on the myocardial boundary Pixel grayscale gradient mode.
  • p is the average density of the myocardium obtained from clinical trials and is the volume of a certain myocardial of interest on the frame image.
  • the above formula for calculating the volume takes into account the uncertainty of the boundary voxels on the precise description of the boundary. Therefore, these voxels are not directly used as a volume element to participate in the volume calculation, but are multiplied by a weighted value and then participate.
  • the accumulation of volume reflects its ambiguity and can more accurately reflect the actual volume of the heart chamber or myocardium.
  • the volume parameter in the formula for calculating EF is obtained using the method of the present invention.
  • a formula for calculating myocardial mass wherein the volume parameter is obtained using the method of the present invention can be performed on 2D slices or on 3D voxel data, and can be extended to any high-dimensional data processing.
  • the geometry of the filling unit is a circle, and the pixel intensity data in the circular area is examined.
  • the geometry of the filling sub-area is a sphere, and the voxel intensity data in the sphere is examined.
  • the processing on 2D is a preliminary process and is further refined/optimized on 3D.
  • the division of the neighborhoods is overlapped by the principle of full coverage.
  • the circular area around each set point is one of the elements of the invention. You can use different shapes flexibly; Pixels Filled areas (sub-areas) are the total collection of circular sub-areas around each set point.
  • the boundary processing of the present invention can also be processed on three-dimensional data, with reference to the above-described two-dimensional processing embodiment. For example, you can change the geometry of the segmentation processing area from a circle to a sphere.
  • the invention can also be applied to other types of image data processing, such as CT, MRI, PET, SPECT, etc., to segment and identify the anatomical tissue of interest therein, and to calculate relevant physiological parameters.
  • the anatomical tissue of interest in the image has a certain degree of contrast with the surrounding tissue, and is irregular, and is suitable for segmentation by applying the present invention.
  • the present invention is applicable to both normal tissue conditions and tissue lesions. It will be apparent to those skilled in the art that ⁇ RTIgt; ⁇ / RTI> ⁇ RTIgt; ⁇ / RTI> ⁇ RTIgt; ⁇ / RTI> ⁇ RTIgt; Accordingly, the present invention is intended to cover various modifications and alternatives

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)
  • Image Processing (AREA)

Abstract

一种医学影像的目标区域的边界确定装置,用于根据医学影像中组织分布所反映出的与其相对应的物理定量特征来区分目标区域的边界,其包括:交互单元,操作人员经由交互单元在医学影像上选择目标区域;阈值设定单元,其确定所选的目标区域中的物理定量特征的阈值;阈值分割单元,其将待分析区域或者说感兴趣区域划分成子区域,或在待分析区域中填充影像单位(像素),其中,该待分析区域包含有目标区域,子区域即在图像分割出的子区域。阈值分割单元将子区域的物理定量特征的参数平均值与阈值比较,根据比较结果标记各子区域。该装置及其方法物理意义明确,算法简单有效,特别适应临床各种病变心脏的特殊情况处理。

Description

医学影像的目标区域边界确定装置和方法 技术領域
本发明涉及医学影像的目标区域边界确定方法和装置, 以及利 用确定的目标区域边界来确定生理参数。更具体地, 本发明涉及基于 真实超声影像数据测定心赃生理参数。 背景技术
医学成像已经成为现代医疗不可或缺的一部分, 其应用贯穿整 个临床工作, 不仅广泛用于疾病诊断, 而且在外科手术和放射治疗等 的计划设计、 方案实施以及疗效评估方面发挥着重要作用。 目前, 医 学图像可以分为解剖图像和功能图像两个部分。解剖图像主要描述人 体形态信息, 包括 X射线透射成像、 CT、 MRI、 US等。
特别是在现代心脏病的诊断与治疗方面, 利用计算机技术对医 学影像进行定量分析成为重要的技术改进方向,用以增加诊断的客观 性, 并且更容易掌握操作, 能够减少对阅片人的经验依赖, 排除不同 阅片人之间的判断差异。进一步, 本领域渴望基于心脏图像摄影序列 更准确获知心脏的量化的生理参数, 例如, 心室的容积、 心肌质量、 心腔壁增厚、 心脏射血分数 (EF值) 等等。 准确获取心脏射血分数具 有重要意义, 根据心脏射血分数可以推算心脏的射血能力, 是判断心 功能的重要参数。
3D超声是一种无探伤的影像检查技术, 在心脏疾病的探査中, 其 具有成像速度快且成本低的特点, 因此, 在心脏病诊断与治疗方面应 用最为广泛。 在 3D超声影像中分析心腔容积、 射血分数、 心肌的体积 和质量等生理参数是进行诊断的重要依据。 但是, 由于超声心动图含 有大量噪声, 且心腔的内膜和心肌的边缘是不规则的 (尤其是发生病 变的心腔和心肌) , 从而给相关定量计算带来了困难。 其中的困难之 一, 是如何准确地得到心内膜的边界, 以及如何准确针对心脏的不规 则的变化进行计量。 本领域长期致力于提高超声影像获取生理参数的 准确性和可操作性。 - 目前在临床上较为普便使用的心脏射血分数(EF值〉测定方法是 以交互的方式定义一些控制点, 并通过数学建模, 使用一系列模拟的 几何形状来逼近心腔, 因而是很不准确的。
多件专利公开文献采用上述手段。例如, JP2002085404, 题为《超 声波图像处理器》 (ultrasonic imaging processor) » 教导将心腔分为
20段来近似统计其容积。 EP123617,教导使用分段的曲线来描述心腔。
JP2008073423 , 教导用 50多个影像集合的参考轮廓来插值得到近似心 腔。 EP1998671(A1), 教导利用鼠标点出几个控制点, 和一个模板匹配 达到自动分割。 EP2030042 (A1 )教导了一种手工标记少量控制点, 结 合训练出的模板得到心内膜。
常规技术中, 较多采用先验模型处理数据, 以获得具有复杂形状, 例如心脏和心肌等的与体积或容积相关的生理参数。
关于先验模型, 是基于统计学的一个模型, 指要分析的数据集合 服从某种未知概率分布, 并且和一个已知样本的数据集合之间有确定 的联系。 为了求出这个未知分布, 需要在已经样本数据集合上计算其 服从的概率分布, 这个能事先算出的概率分布或参数就被称为是先验 模型。
病变的心脏和正常的心腔相比, 通常来说, 不再是一个能用上述 模型估箅的心腔。 病变心脏的心腔具有不可预测的形状改变, 而且心 内膜不规则 (如: 肿瘤占位室壁瘤、 心壁增厚) 。 心腔形状的改变导 致射血功能减低, 心瓣膜功能不全等症状。
在临床应用方面, 已有预先计算多帧影像后得到心腔的先验形状 模型, 通过和当前影像上心腔的近似几何模型对比, 修正得到当前影 像上的心腔。 但是, 这类先验模型是根据正常的心脏计算得到的, 在 实际的临床应用中, 对于病变的心脏, 该方法难以保证获得准确的结 果。
参见 Hansson M, Fundana , Brandt S.S, Gudmundsson P, Convex spatio-temporal segmentation of the endocardium in ultrasound data using distribution and shape priors. Biomedical Imaging; From Nano to Macro, 2011 , Page(s): 626 629。 该文献提出了使用机器学习和形态学结合的 方法来做心腔分割, 提出使用瑞利分布为基础建立一个概率模型, 该 模型用来计箅当前区域属于心腔内部的概率和当前区域属于心腔外部 的概率。 然后使用大量的超声图像数据来训练该模型, 得到概率模型 中的各参数估计值。 最后使用该概率模型计算出来的概率作为先验, 结合先验的心腔形态学模型来做新图像中心腔的分割。
Paragios N. A level set approach for shape-driven segmentation and tracking of the left ventricle. Medical Imaging, 2003, Page(s): 773 776 采用的是水平集算法作为左心室分割算法的主体, 加之使用大量 的先验知识, 也就是己知正确的左心室分割结果。 使用先验经验结合 图像本身特点制定水平集的速度函数以及限制区域。 从而达到左心室 分割的目的。
Hamarneh G, Gustavsson T. Combining snakes and active shape models for segmenting the human left ventricle in echocardiographic images. Computers in Cardiology 2000 Digital Object Identifier: 10.1109/CIC.2000.898469 Publication Year: 2000 , Page(s): 115 118 使用 snake模型来进行左心室的分割,该方法需要医生手动划分大量心 脏超声图像中的左心室的轮廓轨迹作为一个训练样本, 然后使用这些 数据来定义一系列离散余弦变换系数 (DCT eoeffidents)。 在使用 snake 做新的左心室分割的时候, 寻找到 snake 初始坐标的离散余弦变换系 数, 然后使用先验经验中的离散余弦变换系数作为外力部分对活动轮 廓迭代至能量最小化。
其他的相关专利文献, 例如, 关于中国专利公开号 CN1777898A, 申请号 200480010928,2, 题为 《无创式左心室的容积测定》 , 其涉及 处理 MR图像, 基于心脏 3D图形中的心内膜轮廓来估计 LV容积。 这 些轮廓由人工指定或半自动导出。 以这些轮廓所包围面积内的强度变 化来估计 LV容积。其中教导,基于图像像素之间的差异(即图像梯度), 采用人工描迹来标识边界点, 因此易受成像噪声的影响, 造成不准确。 进一步, 把这确定轮廓直接应用到其他的时间幀上, 虽然经过自动修 正, 仍然会进一步引入误差。
关于心肌测量的常规技术, 目前在临床上较多使用的心肌分割方 法是基于斑点紋理分析, 其同样需要以交互的方式定义一些控制点, 运用拟合曲线的方法, 得到心肌的近似轮廓, 因而是很不准确的。 同 样地, 在临床上还预先计算多帧影像后得到心肌的先验形状模型, 通 过和当前影像上心肌的近似几何模型对比》 修正得到当前影像上的心 肌。 但是, 如上文提到的, 先验模型是根据正常的心脏计算得到的, 在实际的临床应用中, 对于病变的心脏该方法同样难以获得准确的结 果。
CN101404931A (申请号 CN200780009898.7) , 题为 "借助心肌 机能的量化的超声诊断" , 教导手工先设置控制点, 再根据图像梯度 用曲线连接控制点, 从而达到近似描迹的目的。
CN101454688A (申请号 C誦 07議 18854.0) , 题为 "心腔壁增 厚的量化和显示" , 披露了散斑跟踪指定心肌位置点的距离、 壁厚度 变化以及应变。 也没有获得单个的心肌。 该技术是使用图像梯度确定 心内膜边界, 如果图像噪声增大, 则不准确。 心外膜因为没有明确的 梯度, 因此它在自动确定时, 往往边界会缺失, 且不准确。 所以该专 利提供了一个工具, 在心动周期开始和结束时人工调整这两条边界, 然后再在两条边界间自动设置需要跟踪的点, 它们位于心肌上, 然后 记录每个点周围的像素作为散斑图案, 不同帧之间的散斑圉案进行最 大相关性块匹配, 从而能跟踪每个点的运动。 这样的散斑跟踪容易受 到噪声影响
相关论文》 Alessandrini, M, Dietenbeck, T. Barbosa, D. D'hooge, J. Basset, O. Speciale, N. Friboulet, D. Bernard,0. Segmentation of the full myocardium in echocardiography using constrained level-sets. Computing In Cardiology. 2010,披露了将传统的 level- set方法和先验的形态学方 法结合, 将图像中的点标上 level- set 能量和形态学能量两个属性, 最 后将两个能量属性值加权相加, 得到各像素点的能量值。 在算法初始 化时, 人工在图像上点 6个点 (5个点在心外膜上, 一个点在 心内膜 上) , 对心内膜和心外膜上的点分别建立值为 0 的演化函数, 然后对 图像上所有点计算两个演化函数的值, 分别得到两条演化曲线。 分割 的是心肌层。 相关论文, Alessandrini, M. Friboulet, D, Basset, O. D'hooge, J. Bernard, O. Level-set segmentation of myocardium and epicardium in ultrasound images using localized Bhattacharyya distance. Ultrasonics Symposium (IUS). 2009,披露的算法使用基于瑞利分布的巴氏距离作为 leveret算法在演化时候的能量限制, 在算法初始化时, 人工在图像上 点 6个点 (5个点在心外膜上, 一个点在心内膜上) , 对心内膜和心外 膜上的点分别建立演化函数。 分割的是心肌层。
相关论文 T. Dietenbeck, M. Alessandrini, D, Barbosa, J, D, hooge, D. Friboulet, O. Bernard, Detection of the whole myocardium in 2D-echocardiography for multiple orientations using a geometrically constrained level-set. Medical Image Analysis. 2011 : 该 文 在 ( Segmentation of the Full Myocardiumin Echocardiography Using Constrained Level-Sets ) 的基础上增加了厚度因素作为 level-set的能量 约束条件, 用于防止心内膜和心外膜两条进化曲线在演化过程中由于 相 同的因素导致两条曲线的融合。 为了保证算法在短轴和长轴等图像 上的正确应用, 在使用该算法前需要手动指定两个点确定三尖瓣 的位 置用来保证算法的正确执行。 分割的是心肌层。
相对正常心肌而言, 病变的心肌具有扩张性、 收缩性、 肥大型等 的病变, 最终影响其收縮能力, 具体表现在它的弹性形变参数的改变 上。 而在几何形态上, 和正常心肌相比, 也会随之发生变化, 因而可 能产生有不规则的边界。
因此, 本领域迫切的需求进一步改进利用图像处理获取与心脏相 关的量化参数, 以进一步提高测量精确度以及可操作性。 发明内容
鉴于上述现有技术存在的缺点,本发明旨在基于现有的医学影像 技术, 寻求更为有效和准确的图像处理和计算的装置和方法, 以改善 和提高关于心腔的容积、 射血分数、 心肌体积和质量等相关的生理参 数的准确性, 从而在帮助临床处理过程中做出正确及时的诊断。
本发明的第一方面, 涉及一种医学影像的目标区域边界确定装 置, 用于根据该医学影像中组织分布所反映出的相对应的物理定量特 征来区分目标区域边界, 该装置包括:
交互单元, 操作人员经由交互单元在医学影像上选择目标区域; 阈值设定单元,其确定所选的目标区域中的物理定量特征的阈值; 阈值分割单元, 其将至少包含目标区域局部的待分析区域分割成 子区域, 以及, 将各子区域的物理定量特征的参数平均值与阈值比较, 根据比较结果标记各子区域。
基于第一方面的本发明第二方面提供的装置, 其中所述物理定量 特征包括像素灰度, 像素梯度, 体素灰度, 或体素强度。
基于上述方面的本发明第三方面所述的装置, 其中, 所述子区域 按照如下方式设置: 将待分析区域划分成相互交叠的多个相邻的子区 域, 各相邻的子区域交叠或非交叠区域共同并完全覆盖目标区域。
基于上述方面的本发明第四方面的装置, 以所选的目标区域内部 的位置点为圆心, 设定半径 r, 定义一个圆形区域, 分析该圆形区域内 的物理定量特征分布, 以确定阈值。
基于上述方面的本发明第五方面的装置, 其中医学影像为心脏的 医学影像, 选择心腔的一个位置点为圆心, 以 5mm为半径, 定义一个 圆形区域, 计算该圆形区域内像素灰度值的平均值作为所述阈值; 以 及, 如果一子区域中像素灰度小于所述阈值, 则将该子区域标记为所 述目标区域, 即所述心腔的区域。
基于上述方面的本发明第六方面的装置, 其中, 医学影像为心脏 的医学影像, 将待分析区域划分为一系列相互交叠的圆形区域, 该圆 形区域为所述的子区域, 该圆形的半径是 lmm, 各圆形之间圆心的距 离也是 1mm, 计算出各所述子区域灰度平均值, 如果该平均值大于所 述阈值, 则将该子区域内的像素点都标记为心肌区域, 否则都标记为 非心肌区域。
基于上述方面的本发明第七方面的装置, 其中将子区域设置为球 体, 并将球体内的平均体素灰度或体素梯度与阈值参数比较并做标记。
基于上述方面的本发明第八方面的装置, 其中, 其中的医学影像 为心脏影像, 以及, 目标区域是任意一心腔、 或者心肌。 基于上述方 ϊ I本发明第九方面的装置, 其中的医学影像为 3D 超声影像。 .
另外, 本发明还包括一种医学影像的目标区域边界确定方法, 其 中, 根据该医学影像中组织分布反映出的物理定量特征区分目标区域 边界, 该方法包括如下步骤:
选择目标区域,
设定所述目标区域中的物理定量特征的阈值,
. 将至少包含目标区域局部的待分析区域分割成子区域,
将各子区域的物理定量特征的参数平均值与阈值比较, 根据比 较结果标记各子区域。 本发明的上述方面, 基于成像对象组织分布的一种物理性质反映 在影像中的一种定量特征, 针对目标区域中典型区域, 例如目标区域 的中间部分的局部区域的这种定量特征, 设定阈值参数, 通过阈值分 割的方法判断各子区域与阈值比较的结果, 从而将各子区域分为两类, 用以区分影像的目标区域边界。
关于定量特征, 优选像素或体素的灰度。 平均灰度是一种特征测 量方式, 其测定速度较快。 此外, 也可考察区域的梯度分布, 是另一 种简单高效的特征测量。
上述发明的目的在于, 通过采用更准确和有效的方法确定医学 影像目标区域边界。 本发明应用于处理真实 3D超声波医学影像时, 可以获取更为准确的量化生理参数。 真实 3D超声波医学影像是指由 3D超声探头直接生成的 3D影像。 在超声波 3D影像中, 心内膜边界 等的边界确定在测定心脏相关生理参数方面具有重要的意义。
本发明的发明人受到医学成像中常采用的组织灌注方法的启 发。现有技术中,熟知通过造影剂在人体组织的空腔以及间隙内扩散, 在形成数字化的影像时,解剖组织和造影剂的数据之间就产生了明显 的对比度, 从而能使得影像操作员容易识别各种解剖组织的边界, 为 进一步定量诊断提供可靠的依据。
发明人提出了借助于计算机技术实现一种虚拟的组织灌注的模 式, sp , 利用图像包含的像素所反映的物理定量特性, 帮助本领域技 术人员有效识别组织的边界, 以获取更为准确的组织边界; 进一步, 基于准确的边界获取心脏的量化生理参数。
更具体而言, 本发明利用计算机技术, 从数字化影像中提取感 兴趣组织边界。所说的感兴趣的组织边界周围的像素或体素之间有明 显的对比度, 但边界会受颗粒状噪声的影响而变得不清晰。发明人具 体考察图像中像素的特点, 在待分析区域中设置单元或子区域, 该单 元中填充着最小基本单位的填充子区域固有的像素, 因此, 假想其为 "像素填充单元"。 在待分析区域上设置点考察点, 该点周围一个圆 形或者椭圆形子区域即为一个单元或像素填充单元,各子区域之间相 互交叠, 分析子区域内的像素值或者体素值的分布特征, 从中推算出 一个固定的或者不固定的阈值, 根据这个阈值, 再对每个考察点周围 区域内的每个像素或者体素进行标记, 从而得到感兴趣的组织区域, 它的边界就是感兴趣的组织的边界。在已经标记好的组织区域上, 还 可以再重新用设计的算法来设置考察点,并使用多种不同尺度或大小 的圆形或者椭圆形区域来进一步分析像素值或体素值的分布规律,进 一步细化感兴趣组织区域的边界。
更进一步说明, 本发明利用计算机技术, 利用影像中的组织, 例如, 心脏心腔之间物理特性的不同, 以及其反映于医学影像中的组 织特性区别, 直接利用不同相关图像中的区域特性, 由操作者根据经 验选取该区域的大致中间位置, 利用计算机技术确定该区域物理特 性, 例如, 灰度的平均值、 梯度值等, 通过阈值比较将该区域和边界 处区分成两类, BP , 达成图像二值化的效果, 从而区分出边界。 这种 区分方式更为客观准确,避免先验模型分割心腔和心肌方法的局限性。
上述说明不希望使本发明拘泥于任何理论局限, 仅仅为了使本 领域技术人员更容易理解本发明。 .
下面参照附图和具体实施方式进一步说明, 以使本领域技术人 员更容易理解本发明并了解本发明的优点和其他的目的。
附图说明
为了更完整地理解本发明, 参见以下说明及附图, 其中; 图 1 一种典型常规技术图像处理装置上的近似心腔分割结果的 示意图;
图 2是示意本发明一个实施例中采用交互式选择目标心腔; 图 3A示出本发明方法在标记出的心腔边界;
图 3B是一个时间序列各帧的心腔容积变化曲线示意图,从图中 可以看出每个心动周期中各帧影像的最大容积 ™«和最小容积 «™; 以及
图 4本发明一个具体实施例的流程图。 具体实施方式
本发明提出的针对感兴趣组织或区域 (目标区域) 的边界处理, 可以有多种不同的应用。 通过具体实施方式的说明在于帮助本领域技 术人员理解本发明, 而不应当构成对本发明的限定。
具体实施方式的描述中, 主要以像素灰度为物理定量特征为例进 行分析。 本发明也可以应用其他合适的物理定量特征。
在一种实施方式中, 本发明的边界处理包括如下步骤:
1.首先将医学影像的切片图划分为一系列相互交叠的圆形区域作 为覆盖待分析区域的小的子区域, 并将其定义为单元, 这些单元看成 是由影像的像素填充的单元, 因为, 这些单元中充满了图像的像素。 使这样划分的圆形区域覆盖全图, 在各圆形区域上根据像素灰度值计 算定量特征, 并确定阈值, 根据阈值将的各单元初步标记出来, 即, 根据阈值对各单元进行区分。
2. 初步的标记得到一个或多个连通的区域, 接着综合感兴趣的区 域 (ROI) , 或者说是目标区域, 进一步处理, 即只有包含了操作员鼠 标点击的连通区域被保留, 其他区域都抛弃, 或者说, 将其他的区域 取消标记。 这样得到初次分割的结果。
3. 得到初次分割的区域结果后, 接着进一步对边界细化处理 β 首 先把分割后区域的边界单独标记出来, 然后在边界上布置像素填充单 元, 将这些像素填充单元设置为覆盖更小的区域, 可以是第一步中像 素填充单元的一半大小, 它们仍然需要相互交叠。 同样在这些区域上 计算定量特征, 如平均灰度或梯度等, 并得到阈值, 根据阈值将各像 素填充单元进行标记, 并和初次分割的区域结果进行 "或" 操作, 合 并得到细化的区域结果。 此外, 还可以进行进一步的细化处理, 例如:
操作员根据临床需要重复步骤 3 , 可以通过进一步减小像素填充 单元的尺寸来进一步细化边界, 直到得到满意结果为止。
此外, 还可以直接在三维数据上进行最后一次细化边界的处理。 所谓三维数据就是由前面的切片图堆积而成的。 同样前面得到的切片 图的边界的堆积在 3D数据中表现为一张曲面。在这张曲面上布置体素 填充单元, 体素填充单元与最后一次执行第三步时的设置相同, 即, 具有相同的半径设置, 它们仍然需要相互交叠。 同样在这些区域上根 据像素灰度值计算定量特征, 并得到阈值, 根据阈值标记各像素填充 单元, 并和最后一次第三步得到的区域结果进行 "或" 操作, 合并得 到细化的区域结果。 关于处理心腔边界, 与上述说明基本相同, 进一步, 需要在其中 的步骤 2中增加如下处理:
( 1 )在切片图像中初步标记的处理步骤与前述相同, 但是, 在选 取心腔区域时, 这一步只观察灰度平均值。
(2 )综合操作员鼠标点击感兴趣区域, 得到初步的分割区域; 同 详细步骤中的按 8 邻域连通域把含有鼠标点击的那个区域单独分离出 来的表述;
( 3 )在步骤 2得到的区域上, 把边界单独标记出来, 然后将边界 划分为一系列相互交叠的圆形区域, 圆心都是边界上的点, 半径为第 一步中圆形区域半径的一半。 计算每个圆形区域上的像素灰度值的平 均值、 像素灰度梯度模的平均值。 再通过计算这些数值的平均值得到 两个阈值:
其中, n 为圆形区域的个数。 接着检查每个圆形区域的灰度平均 值和像素的梯度模平均值。 灰度平均值反映的是灰度均值的均值; 梯 度模平均值反映的是梯度模均值的均值, 分析区像数变化的大小, 它 反映这个区域像素变化的大小, 作为边界这个值会变大, 而小于此值 说明它还在边界内, 应当被标记出来, 条件就是某个子区域的灰度均 值小于灰度均值的阈值, 并且梯度模均值也小于梯度模均值的 W值。 则把该圆形区域内的像素标记为心腔区域, 否则标记为非心腔区域。 再把本步骤标出的心腔区域和第二步标出的心腔区域进行 "或"操作, 合并得到细化的心腔区域。
( 4) 操作员根据临床需要重复步骤 3, 每次使用的圆形区域半径 都是上次使用的圆形区域半径的一半, 来进一歩细化边界, 直到得到 满意的 2D切片图上的结果为止。
( 5 ) 在该幀的 3D数据上进行最后一次细化边界的处理。 将 2D 切片图堆积成 3D数据, 第四步得到各 2D切片图上的心腔区域同时被 堆积成 3D区域。 先把 3D区域的边界曲面单独标记出来, 然后将边界 曲面划分为一系列相互交叠的球形区域, 球心都是边界曲面上的点, 半径为第四步中最后一次使用的圆形区域的半径。 计算每个球形区域 上的体素灰度值的平均值、 体素灰度梯度模的平均值。 再通过计算这 些数值的平均值得到灰度平均值和像素的梯度模平均值。
其中, n 为球形区域的个数。 接着检查每个球形区域的灰度平均 值和梯度模平均值, 把该球形区域内的像素标记为心腔区域, 否则标 记为非心腔区域。 再把本步骤标出的心腔区域和第四步标出的心腔区 域进行 "或"操作, 合并得到细化后的心腔 3D区域。 实施例 1
本发明应用于针对患者心赃的真实三维(3D)超声影像数据处理, 在本实施例中用于获取心腔容积以及射血分数。
步骤 1 , 利用超声成像设备获得患者的医学影像资料。 本实施例 中, 使用真实的 3D超声探头对心脏区域扫描, 得到 3D超声影像的多 个时间序列每个时间序列包含一系列的帧, 记录了一个或多个完整的 心动周期, 每个帧包含有多个切片组成的 3D体素数据。使用的成像设 备例如, 西门子 SC2000超声心动图仪和飞利浦 IE33两种型号。 步骤 2 , 在真实 3D超声影像时间序列中所有帧的所有切片影像 中, 提取心腔轮廓。 在具体的实施例中, 一般地, 对一个病人, 扫描 5 8个时间序列, 一个时间序列有 δ~44帧, 一帧有 256个切片图像, 每个图像的大小为 256*256像素。
提取心腔轮廓包括如下步骤:
a)在真实 3D超声影像时间序列的某一帧的某一个切片影像中, 利用鼠标点选感兴趣的心腔位置, 选择目标区域。
进一步具体说明, 选择切片影像的依据为含有感兴趣且暴露最 清晰的心腔。 鼠标点选的位置目视可以明确确定的, 并且明显在心腔 范围之内。
在显示有影像时间序列的某一帧数据的所有切片图的界面上, 操作员利用鼠标在切片图上点击,点击的位置要求是在感兴趣的心腔 的内部。 最后, 以图像左上角为原点, 记录该位置点的 X 坐标和 y 坐标。 本实施例中, 以宽度方向为 X轴, 正方向是向右; 以高度方向 为 y轴, 正方向是向下; 这样得到的 X , y坐标。 设置坐标的目的在 于描述每个像素或者体素在空间的位置,它们由坐标(x,y )或(x,y,z ) 唯一确定。在计算中, 使用坐标的目的主要用于判断像素或体素之间 的邻接关系 (2D影像上有 8邻域或 4邻域, 3D影像上有 6邻域和 26 邻域) , 用于设置填充单元的范围的确定, 以及感兴趣心腔的标 记(覆盖着感兴趣心腔的灌注区域在被标记后它们之间形成连通的邻 接关系, 从而能分离得到单个的心腔) 。
可选择地, 还可以附加设置自动关联处理单元, 只要点击一个 切片, 该幀 3D影像的所有切片自动得到关联处理, 每个幀都只需要 点击一个切片, 其他切片自动处理。
通常情况下, 一个超声的图像范围包含感兴趣区和噪声 (非感 兴趣区) , 并非理想状态的唯一区域, 由于实际效果的局限, 操作员 要求确认(点一下)感兴趣区作为整个技术实现的第一个步骤, 或者 说是 "启动" 步骤。
b ) 以心腔位置点为圆心, 以 r为半径, 定义一个圆形区域, 分 析该区域内的像素灰度分布, 得到一个模型参数 (阈值参数 t) 。 进一步具体说明, 因为用鼠标点击的心腔位置点处的像素, 并 不能反映心腔内的像素灰度值的分布范围,而利用它周围一个邻域内 的像素平均值, 可以得到更为准确的灰度值分布的估计。 因此, 以心 腔位置点为圆心, 以 5mm为半径, 定义一个圆形区域, 根据 3D超 声影像的体素分辨率(即体素中心点之间在 x,y,z三个方向上的距离, 以 mm为单位) , 换算为以像素为单位的圆形区域的范围, 计算该圆 形区域内像素灰度值的平均值, 作为一个模型参数, 即阈值参数 t。
c) 把切片图划分成半径为 r且相互交叠的圆形区域, 从而使这 种圆形区域全面覆盖切片图。这里, 每个圆形区域可以看作是图像的 像素填充的子区域。 进一步, 分析每个圆形区域内像素值的分布, 并 根据阈值参数 t, 利用阈值分割的方法标记出心腔, 即, 将各圆形区 域分别标记为心腔区域和非心腔区域。
在此步骤中, 采用根据步骤 b) 计算出的阈值, 对切片的全部像素 点进行阈值分割。 由于心腔所在区域的像素灰度值较低, 因此, 需要把 切片图中小于阈值的像素标记为心腔区域。 本发明中, 首先将切片图划 分为一系列相互交叠的圆形区域作为子区域或者说像素填充区域, 圆形 的半径是 5mm, 各圆形之间圆心的距离也是 5mm, 按步骤 b) 中的方法 换算为以像素为单位的圆形区域范围。 然后, 计算出区域内所有像素的 灰度平均值, 如果该平均值小于阈值参数 t, 则把该圆形区域内的像素 点都标记为心腔区域, 否则都标记为非心腔区域。 在所有的圆形区域都 处理完以后, 对标记图以 8邻域的方式进行连通域的检查, 把含有操作 员标出的心腔位置点的连通域, 作为感兴趣的心腔的分割结果。 最后, 对一个影像时间序列的所有帧上的所有切片都作同样的阈值分割。
步骤 3 , 根据标记出的心腔区域, 计算心腔容积和 EF值。
a) 根据标记的心腔区域得到心内膜边界。
在标记好的心腔区域上,使用邻域检查法判断每个像素是内点还 是边界点, 如果是边界点, 则标记为白色, 其它点标记为黑色, 从而 得到不规则的心内膜边界。
b) 统计出心内膜边界内的像素总数《纖1。
c)对心内膜边界上的像素, 根据其灰度梯度, 计算得到一个权值, 作用于心内膜边界上的像素数目。 用如下的公式计算出心内膜边界上的像素数目:
N
numl = ^ ! 其中, N是边界上像素的总数目, m„是边界上像素灰度梯度模的 最大值, /min是边界上像素灰度梯度模的最小值, ,.是边界上每个像素灰 度梯度模。
的心腔容积:
Figure imgf000016_0001
其中 S是该帧影像上的切片总数, 《««ίΐ,.是每个切片上心内膜边界 内的像素数目,∞ 2,.是每个切片上心内膜边界上的像素数目, 和 是一帧影像在 x,y,z三个方向上体素中心点之间的距离, 以 mm为单位。
式计算 EF值:
Figure imgf000016_0002
其中: EF值在一个影像时间序列中的每个心动周期计算, max是 该心动周期内各帧影像心腔容积的最大值, nn是该心动周期内各帧影 像心腔容积的最小值。 实施例 2. 计算心肌体积和质量
实施例 4的步骤 1和步骤 2与上述实施例 1相同,因而不再具体说 明。
在完成步骤 1和步骤 2之后, 重复进行步骤 2中的步骤 a) , b) , c ) 处理, 以标出切片图上的其它心腔区域, 用于后续心肌分割中的心 腔排除步骤。 其他心腔区域, 是指在其他暴露得不完整、 不清晰的心腔 上进行相似的分割操作, 目的是把所有的心腔都标记出来, 以免影响到 对心肌的分割。 此步骤为心肌分割前的附加的预处理步 '骤, 目的在于要 排除所有心腔。
步骤 3,在真实 3D超声影像时间序列中所有帧的所有切片影像中, 提取心肌轮廓。
a) 利用鼠标点选多个感兴趣的心肌位置。 在显示有影像时间序列的某一帧数据的所有切片图的界面上,操作 员利用鼠标在切片图上点击, 点击的位置要求是在感兴趣的心肌(目标 心肌) 的内部靠近边缘的地方。 最后, 以图像左上角为原点, 记录该位 置点的 X坐标和 y坐标。 感兴趣的心肌位置点可以有多个。
b) 以每个心肌位置点为圆心, 以 r为半径, 定义一个圆形区域, 分析该区域内的像素灰度分布, 得到一个模型参数 (t) 。
因为用鼠标点击的心肌位置点处的像素,并不能反映心肌内的像素 灰度值的分布范围, 而利用所选位置点周围一个邻域内的像素平均值, 可以得到更为准确的灰度值分布的估计。 因此, 以心肌位置点为圆心, 以 1mm为半径,定义一个圆形区域,根据 3D超声影像的体素分辨率(即 体素中心点之间在 x,y, z三个方向上的距离, 以 mm为单位) , 换算为 以像素为单位的圆形区域的范围, 计算该圆形区域内像素灰度值的平均 值, 作为一个模型参数, 即阈值参数 t。
c) 先在切片图上排除心腔区域, 再把切片图划分成半径为 r且相 互交叠的圆形区域, 作为单元 (像素填充单元) , 分析每个子区域内的 像素值的分布, 并根据阈值参数 t, 利用阈值分割的方法标记出心肌。
这一步是根据步骤 b计算出的阈值参数 t, 对切片的全部像素点进 行阈值分割, 并把步骤 2和附加步骤中得到的所有心腔所在区域的像素 点排除。
由于心肌所在区域的像素灰度值较高, 因此, 需要把切片图中大于 阈值参数 t的像素标记为心肌区域。
处理中首先将切片图划分为一系列相互交叠的圆形区域,该圆形区 域即像素填充单元(单元) 。 该圆形的半径是 lmm, 各圆形之间圆心的 距离也是 lmm, 按步骤 b中的方法换算为以像素为单位的圆形区域范 围。 然后, 计算出区域内所有像素的灰度平均值, 如果该平均值大于阖 值参数 t, 则把该圆形区域内的像素点都标记为心肌区域, 否则都标记 为非心肌区域。 在所有的圆形区域都处理完以后, 对标记图以 8邻域的 方式进行连通域的检查, 把含有操作员标出的心肌位置点的连通域, 作 为感兴趣的心肌的分割结果。 最后, 对一个影像时间序列的所有帧上的 所有切片都作同样的阈值分割。
步骤 4 根据标记出的心肌区域, 计算心肌体积和质量。 a) 根据标记的各个心肌区域得到各个心肌的边界。
在标记好的心肌区域上,使用邻域检查法判断每个像素是内点还是 边界点, 如果是边界点, 则标记为白色, 其它点标记为黑色, 从而得到 不规则的心肌边界。
b ) 各自统计心肌边界内的像素总数《« l。
c ) 各自对心肌边界上的像素, 根据其灰度梯度, 计算得到一个权 值, 作用于心肌边界上的像素数目。
用如下的公式计算出心肌边界上的像素数目:
numl = ^—―
min 其中, ¥是心肌边界上像素的总数目, ^是心肌边界上像素灰度 梯度模的最大值, /min是心肌边界上像素灰度梯度模的最小值, /,是心肌 边界上每个像素灰度梯度模。
<i ) 用如下的公式计算一帧影像上的各心肌体积:
Figure imgf000018_0001
其中 S是该帧影像上的切片总数, 是每个切片上单位心肌边 界内的像素数目,《« 2,是每个切片上各个心肌边界上的像素数目,sx, 和 是一帧影像在 x^z三个方向上体素中心点之间的距离, 以 mm为 单位。
e ) 用如下的公式计算各心肌的质量:
m - pV
其中: p是根据临床实验得到的心肌平均密度, 是该帧影像上 某个感兴趣的心肌的体积。
上述计算体积的公式, 考虑了边界体素在精确描述边界上的不确 定性, 因此对这些体素不是直接当作一个体积元参加体积计算的, 而 是给它乘上一个加权值, 再参与体积的累加, 反映了其具有一定的模 糊性, 能更准确地反映实际的心腔或心肌的体积。
计算 EF的公式中的体积参数是使用本发明的方法得到的。
计算心肌质量的公式其中的体积参数是使用本发明的方法得到 的。 进一步说明, 本发明中提出的填充单元的处理既可以在 2D 切片 上进行, 也可以在 3D体素数据上进行, 更可以推广到任意高维的数据 上处理。 2D时填充单元的几何形状是圆, 考察圆形区域内的像素强度 数据, 3D时填充子区域的几何形状是球体, 考察球体内的体素强度数 据。在 2D上的处理是初步的处理,在 3D上是进一步的细化 /优化处理。
本发明中, 将划分的邻域交叠采用全面覆盖原则。 每个设置点周 围的圆形区域, 是发明的要素之一。 可以灵活使用不同的形状; 像素 填充区域 (子区域) 就是指每个设置点周围的圆形子区域的总集合。
本发明中提到的阈值分割处理,是一种基于区域的图像分割技术》 其基本原理是: 通过设定不同的特征阈值, 把图像像素点分为若干类。 常用的特征包括: 直接来自原始图像的灰度或彩色特征; 由原始灰度 或彩色值变换得到的特征。设原始图像为 f(x, y),按照一定的准则 f(x, y)中找到特征值 T, 将图像分割为两个部分, 分割后的图像 g(x, y)¾ : 若 f(x» y)像素特征值大于 T, 则 g(x, y)取为 0(黑), 反之为 1(白), 即 为通常所说的图像二值化。 也可以让 f(x, y)像素特征值小于 T时 g(x, y)取为 1 , 反之为 0。
本发明的边界处理, 也可以在三维数据上处理的, 参照上述的二 维处理实施例。 例如, 可以将分割处理区域的几何形状从圆形改变成 球体》 考察球体内的体素来做标记。
本发明也可以应用到其它类型的影像数据处理, 例如 CT、 MRI、 PET, SPECT 等, 以对其中感兴趣的解剖组织进行分割和识别, 并计 算相关生理参数。 影像中感兴趣的解剖组织和周围的组织有一定的对 比度, 且不规则, 适合于应用本发明进行分割, 本发明既适用于组织 正常的情形, 也适用于组织病变的情形。 本领域技术人员应当明了, 对本发明已描述的优选具体实施例可 以进行各种改进和变化, 而不偏离本发明的精神或范围。 因此, 本发 明包括在所附权利要求及其等同替换范围之内的各种改进和变化。

Claims

权 利 要 求 书
1. 一种医学影像的目标区域边界确定装置, 用于根据该医学影像 中组织分布所反映出的相对应的物理定量特征来区分目标区域边界, 该装置包括
交互单元, 操作人员经由交互单元在医学影像上选择目标区域; 阈值设定单元,其确定所选的目标区域中的物理定量特征的阈值; 阈值分割单元, 其将至少包含所述目标区域局部的待分析区域分 割成子区域, 以及, 将各所述子区域的物理定量特征的参数平均值与 所述阈值比较, 根据比较结果标记各所述子区域。
2. 根据权利要求 1所述的边界确定装置, 其中所述物理定量特征 包括像素灰度, 像素梯度, 体素灰度, 或体素强度。
3. 根据权利要求 1所述的权利要求所述的边界确定装置, 其中, 所述子区域按照如下方式设置: 将待分析区域划分成相互交叠的多个 相邻的子区域, 各相邻的子区域交叠或非交叠区域共同并完全覆盖所 述目标区域。
4. 根据权利要求 1至 3中任一项权利要求所述的边界确定装置, 以所选的目标区域内部的位置点为圆心, 设定半径 r, 定义一个圆形区 域, 分析该圆形区域内的物理定量特征分布, 以确定所述阈值。
5. 根据权利要求 1至 3中任一项权利要求所述的边界确定装置, 所述医学影像为心脏的医学影像, 选择心腔的一个位置点为圆心, 以
5mm为半径, 定义一个圆形区域, 计算该圆形区域内像素灰度值的平 均值作为所述阈值; 以及, 如果一子区域中像素灰度小于所述阈值, 则将该子区域标记为所述目标区域, 以及, 该目标区域是所述心腔的 区域。
6. 根据权利要求 1至 3中任一项权利要求所述的边界确定装置, 其中, 所述医学影像为心脏的医学影像, 将待分析区域划分为一系列 相互交叠的圆形区域, 该圆形区域为所述的子区域, 该圆形的半径是 1mm, 各圆形之间圆心的距离也是 lmm, 计算出各所述子区域灰度平 均值, 如果该平均值大于所述阈值, 则将该子区域内的像素点都标记 为心肌区域, 否则都标记为非心肌区域。
7. 根据权利要求 1所述的边界确定装置, 其中将所述子区域设置 为球体, 并将球体内的平均体素灰度或体素梯度与所述阈值参数比较 并做标记。
8. 根据权利要求 1~3中任一项所述的边界确定装置, 其中, 其中 所述的医学影像为心脏影像, 以及, 所述目标区域是任意一心腔、 或 者心肌。
9. 根据权利要求 1 3、 7中任一项所述的边界确定装置, 其中所述 的医学影像为 3D超声影像。
10. 一种医学影像的目标区域边界确定方法, 其中, 根据该医学 影像中组织分布反映出的物理定量特征区分目标区域边界, 该方法包 括如下步骤:
—选择目标区域,
-设定所述目标区域中的物理定量特征的阈值,
将至少包含所述目标区域局部的待分析区域分割成子区域, 将各所述子区域的物理定量特征的参数平均值与所述阈值比 较, 根据比较结果标记各子区域。
11. 根据权利要求 10所述的方法, 其中所述物理定量特征选自像 素灰度, 像素梯度, 体素灰度, 或体素强度。
12. 根据权利要求 10所述的方法,
其中, 在所述分割子区域的步骤中, 包括如下处理: 将待分析区 域划分成相互交叠的多个邻域, 各邻域的交叠或非交叠区域共同并完 全覆盖所述目标区域。
13. 根据权利要求 10 12中任一项权利要求所述的方法,其中, 在 设定阈值的步骤中, 包括如下处理: 以所选的目标区域内部的位置点 为圆心, 设定半径 r, 定义一个圆形区域, 分析该圆形区域内的像素灰 度分布, 以确定所述阈值。
14. 根据权利要求 10 12中任一项权利要求所述的方法, 其中, 其 中, 针对心脏获取所述医学影像, 所述的目标区域为心腔, 选择一个 心腔位置点为圆心, 以 5mm为半径, 定义一个圆形区域, 计算该圆形 区域内像素灰度值的平均值作为所述阈值, 以及, 如果一子区域中像 素灰度小于所述阈值, 则将该子区域标记为心腔区域。
15. 根据权利要求 10 12中任一项权利要求所述的方法, 其中, 其 中, 针对心脏获取所述医学影像, 将待分析区域划分为一系列相互交 叠的圆形区域, 该圆形区域为所述的子区域, 该圆形的半径是 1mm, 各圆形之间圆心的距离也是 lmm, 计算出各子区域灰度平均值, 如果 该平均值大于所述阈值, 则将该子区域内的像素点都标记为心肌区域, 否则都标记为非心肌区域。
16. 根据权利要求 10所述的方法,其中将所述子区域设置为球体, 并将球体内的平均体素灰度或体素梯度与所述阈值参数比较并做标 记。
17. 根据权利要求 10 12, 16中任一项权利要求所述的方法, 其中 所述的医学影像为心脏影像, 以及, 所述的目标区域为任意一个心腔 或者心肌。
18. 根据权利要求 10 12、 16中任一项权利要求所述的方法, 其中 所述的医学影像为 3D超声影像。
PCT/CN2013/071129 2012-03-05 2013-01-30 医学影像的目标区域边界确定装置和方法 WO2013131420A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US14/383,060 US20150023578A1 (en) 2012-03-05 2013-01-30 Device and method for determining border of target region of medical images

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN201210055551.9 2012-03-05
CN201210055551 2012-03-05
CN201210375093.7A CN102920477B (zh) 2012-03-05 2012-09-27 医学影像的目标区域边界确定装置和方法
CN201210375093.7 2012-09-27

Publications (1)

Publication Number Publication Date
WO2013131420A1 true WO2013131420A1 (zh) 2013-09-12

Family

ID=47473378

Family Applications (2)

Application Number Title Priority Date Filing Date
PCT/CN2013/071135 WO2013131421A1 (zh) 2012-03-05 2013-01-30 基于3d医学影像测定生理参数的装置和方法
PCT/CN2013/071129 WO2013131420A1 (zh) 2012-03-05 2013-01-30 医学影像的目标区域边界确定装置和方法

Family Applications Before (1)

Application Number Title Priority Date Filing Date
PCT/CN2013/071135 WO2013131421A1 (zh) 2012-03-05 2013-01-30 基于3d医学影像测定生理参数的装置和方法

Country Status (3)

Country Link
US (2) US20150023577A1 (zh)
CN (2) CN102871686B (zh)
WO (2) WO2013131421A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110288581A (zh) * 2019-06-26 2019-09-27 电子科技大学 一种基于保持形状凸性水平集模型的分割方法
CN116386043A (zh) * 2023-03-27 2023-07-04 北京市神经外科研究所 一种脑神经医疗影像胶质瘤区域快速标记方法及系统

Families Citing this family (44)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102871686B (zh) * 2012-03-05 2015-08-19 杭州弘恩医疗科技有限公司 基于3d医学影像测定生理参数的装置和方法
TWI498832B (zh) * 2013-01-22 2015-09-01 Univ Nat Cheng Kung 估算個體運動學或動力學相關參數的電腦實施方法及系統
CN104720850B (zh) * 2013-12-23 2017-10-03 深圳迈瑞生物医疗电子股份有限公司 一种超声造影成像方法及造影图像的区域检测、显像方法
US9436995B2 (en) * 2014-04-27 2016-09-06 International Business Machines Corporation Discriminating between normal and abnormal left ventricles in echocardiography
CN105232081A (zh) * 2014-07-09 2016-01-13 无锡祥生医学影像有限责任公司 医学超声辅助自动诊断装置及方法
CN104398272B (zh) 2014-10-21 2017-09-19 无锡海斯凯尔医学技术有限公司 选择检测区域的方法及装置及弹性检测系统
DE102015208804A1 (de) * 2015-05-12 2016-11-17 Siemens Healthcare Gmbh Vorrichtung und Verfahren zum rechnergestützten Simulieren von chirurgischen Eingriffen
CN104915924B (zh) * 2015-05-14 2018-01-26 常州迪正雅合电子科技有限公司 一种自动实现三维超声图像定标的方法
CN106408648A (zh) * 2015-08-03 2017-02-15 青岛海信医疗设备股份有限公司 一种医学组织的切片图像三维重建的方法及设备
US10140268B2 (en) * 2015-08-27 2018-11-27 Qualcomm Innovation Center, Inc. Efficient browser composition for tiled-rendering graphics processing units
CN107025633B (zh) * 2016-01-29 2020-11-27 中兴通讯股份有限公司 一种图像处理方法及装置
CN107993234A (zh) * 2016-10-26 2018-05-04 中国科学院深圳先进技术研究院 一种坑区域的提取方法及装置
WO2018091486A1 (en) 2016-11-16 2018-05-24 Ventana Medical Systems, Inc. Convolutional neural networks for locating objects of interest in images of biological samples
CN106683083B (zh) * 2016-12-22 2019-09-13 深圳开立生物医疗科技股份有限公司 肛门括约肌图像处理方法及装置、超声设备
WO2018218479A1 (en) * 2017-05-31 2018-12-06 Edan Instruments, Inc. Systems and methods for adaptive enhancement of vascular imaging
US10430987B1 (en) * 2017-06-09 2019-10-01 Snap Inc. Annotating an image with a texture fill
CN107274428B (zh) * 2017-08-03 2020-06-30 汕头市超声仪器研究所有限公司 基于仿真和实测数据的多目标三维超声图像分割方法
CN108305247B (zh) * 2018-01-17 2022-03-04 中南大学湘雅三医院 一种基于ct图像灰度值检测组织硬度的方法
CN108703770B (zh) * 2018-04-08 2021-10-01 智谷医疗科技(广州)有限公司 心室容积监测设备和方法
CN108553124B (zh) * 2018-04-08 2021-02-02 广州市红十字会医院(暨南大学医学院附属广州红十字会医院) 心室容积监测设备和方法
CN108573514B (zh) * 2018-04-16 2022-05-27 北京市神经外科研究所 一种图像的三维融合方法及装置、计算机存储介质
CN109035261B (zh) * 2018-08-09 2023-01-10 北京市商汤科技开发有限公司 医疗影像处理方法及装置、电子设备及存储介质
CN109846513B (zh) * 2018-12-18 2022-11-25 深圳迈瑞生物医疗电子股份有限公司 超声成像方法、系统和图像测量方法、处理系统及介质
CN110009631A (zh) * 2019-04-15 2019-07-12 唐晓颖 眼底图像的血管质量评估方法、装置、设备和介质
CN110136804B (zh) * 2019-04-25 2021-11-16 深圳向往之医疗科技有限公司 一种心肌质量计算方法、系统及电子设备
CN110400626B (zh) * 2019-07-08 2023-03-24 上海联影智能医疗科技有限公司 图像检测方法、装置、计算机设备和存储介质
DE102019212103A1 (de) * 2019-08-13 2021-02-18 Siemens Healthcare Gmbh Surrogatmarker basierend auf medizinischen Bilddaten
CN110866902A (zh) * 2019-11-06 2020-03-06 湖北中烟工业有限责任公司 一种烟标翘曲变形的检测方法
CN110930450A (zh) * 2019-12-11 2020-03-27 清远职业技术学院 基于图像阈值分割和blob分析方法的煤矸石定位方法
US11521314B2 (en) * 2019-12-31 2022-12-06 Shanghai United Imaging Healthcare Co., Ltd. Systems and methods for image processing
CN111368832B (zh) * 2020-03-05 2023-06-20 推想医疗科技股份有限公司 感兴趣区域的标记方法、装置、设备和存储介质
CN112017152B (zh) * 2020-07-02 2022-09-23 杭州市第一人民医院 心房压迹二维影像的处理方法
CN112037167B (zh) * 2020-07-21 2023-11-24 苏州动影信息科技有限公司 一种基于影像组学和遗传算法的目标区域确定系统
US11636603B2 (en) * 2020-11-03 2023-04-25 Dyad Medical, Inc. System and methods for segmentation and assembly of cardiac MRI images
CN112932535B (zh) * 2021-02-01 2022-10-18 杜国庆 一种医学图像分割及检测方法
CN115482246B (zh) * 2021-05-31 2023-06-16 数坤(上海)医疗科技有限公司 一种图像信息提取方法、装置、电子设备和可读存储介质
CN113570594A (zh) * 2021-08-11 2021-10-29 无锡祥生医疗科技股份有限公司 超声图像中目标组织的监测方法、装置及存储介质
CN114299094B (zh) * 2022-01-05 2022-10-11 哈尔滨工业大学 一种基于块选择与扩展的输液瓶图像感兴趣区域提取方法
CN115147378B (zh) * 2022-07-05 2023-07-25 哈尔滨医科大学 一种ct图像分析提取方法
CN116523924B (zh) * 2023-07-05 2023-08-29 吉林大学第一医院 一种医学实验用数据处理方法及系统
CN116630316B (zh) * 2023-07-24 2023-09-26 山东舜云信息科技有限公司 一种基于视频分析的皮带疲劳检测报警方法及报警系统
CN116862930B (zh) * 2023-09-04 2023-11-28 首都医科大学附属北京天坛医院 适用于多模态的脑血管分割方法、装置、设备和存储介质
CN117201800B (zh) * 2023-09-12 2024-03-19 浙江建达科技股份有限公司 基于空间冗余的医疗检查大数据压缩存储系统
CN117237389B (zh) * 2023-11-14 2024-01-19 深圳市亿康医疗技术有限公司 一种中耳胆脂瘤ct影像分割方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5680471A (en) * 1993-07-27 1997-10-21 Kabushiki Kaisha Toshiba Image processing apparatus and method
US20040267125A1 (en) * 2003-06-26 2004-12-30 Skyba Danny M. Adaptive processing of contrast enhanced ultrasonic diagnostic images
CN101219063A (zh) * 2007-01-12 2008-07-16 深圳迈瑞生物医疗电子股份有限公司 基于二维分析的b图像均衡方法和系统结构
WO2011013346A1 (ja) * 2009-07-29 2011-02-03 パナソニック株式会社 超音波診断装置

Family Cites Families (26)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5457754A (en) * 1990-08-02 1995-10-10 University Of Cincinnati Method for automatic contour extraction of a cardiac image
JP3461201B2 (ja) * 1993-07-27 2003-10-27 株式会社東芝 画像処理装置及び画像処理方法
US6094508A (en) * 1997-12-08 2000-07-25 Intel Corporation Perceptual thresholding for gradient-based local edge detection
US6217520B1 (en) * 1998-12-02 2001-04-17 Acuson Corporation Diagnostic medical ultrasound system and method for object of interest extraction
WO2002052509A1 (en) * 2000-12-22 2002-07-04 Koninklijke Philips Electronics N.V. Method of analyzing a data set comprising a volumetric representation of an object to be examined
US20130211238A1 (en) * 2001-01-30 2013-08-15 R. Christopher deCharms Methods for physiological monitoring, training, exercise and regulation
ATE470199T1 (de) * 2003-04-24 2010-06-15 Koninkl Philips Electronics Nv Eingriffsfreie links-herzkammervolumenbestimmung
US7310435B2 (en) * 2003-11-25 2007-12-18 General Electric Company Method and apparatus for extracting multi-dimensional structures using dynamic constraints
US7248725B2 (en) * 2004-01-07 2007-07-24 Ramot At Tel Avia University Ltd. Methods and apparatus for analyzing ultrasound images
US7676091B2 (en) * 2004-01-07 2010-03-09 Ramot At Tel Aviv University Ltd. Method and apparatus for analysing ultrasound images
KR100747093B1 (ko) * 2005-01-12 2007-08-07 주식회사 메디슨 초음파 진단 영상을 이용한 대상체의 경계를 자동으로검출하는 방법 및 초음파 진단 시스템
US20070116357A1 (en) * 2005-11-23 2007-05-24 Agfa-Gevaert Method for point-of-interest attraction in digital images
CN101404931A (zh) * 2006-03-20 2009-04-08 皇家飞利浦电子股份有限公司 借助心肌机能的量化的超声诊断
EP2025290A1 (en) * 2006-05-19 2009-02-18 Hitachi Medical Corporation Medical image display device and program
WO2008115830A2 (en) * 2007-03-16 2008-09-25 Cyberheart, Inc. Radiation treatment planning and delivery for moving targets in the heart
US8155405B2 (en) * 2007-04-20 2012-04-10 Siemens Aktiengsellschaft System and method for lesion segmentation in whole body magnetic resonance images
US8369590B2 (en) * 2007-05-21 2013-02-05 Cornell University Method for segmenting objects in images
CN101527047B (zh) * 2008-03-05 2013-02-13 深圳迈瑞生物医疗电子股份有限公司 使用超声图像检测组织边界的方法与装置
US8199994B2 (en) * 2009-03-13 2012-06-12 International Business Machines Corporation Automatic analysis of cardiac M-mode views
US20110150309A1 (en) * 2009-11-27 2011-06-23 University Health Network Method and system for managing imaging data, and associated devices and compounds
DE102010008243B4 (de) * 2010-02-17 2021-02-11 Siemens Healthcare Gmbh Verfahren und Vorrichtung zur Ermittlung der Vaskularität eines sich in einem Körper befindlichen Objektes
WO2011125513A1 (ja) * 2010-03-31 2011-10-13 株式会社 日立メディコ 医用画像診断装置及び、医用画像の計測値再入力方法
EP2658439B1 (en) * 2010-12-29 2020-02-12 Dia Imaging Analysis Ltd. Automatic left ventricular function evaluation
CN102068281B (zh) * 2011-01-20 2012-10-03 深圳大学 一种占位性病变超声图像的处理方法
US20130195323A1 (en) * 2012-01-26 2013-08-01 Danyu Liu System for Generating Object Contours in 3D Medical Image Data
CN102871686B (zh) * 2012-03-05 2015-08-19 杭州弘恩医疗科技有限公司 基于3d医学影像测定生理参数的装置和方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5680471A (en) * 1993-07-27 1997-10-21 Kabushiki Kaisha Toshiba Image processing apparatus and method
US20040267125A1 (en) * 2003-06-26 2004-12-30 Skyba Danny M. Adaptive processing of contrast enhanced ultrasonic diagnostic images
CN101219063A (zh) * 2007-01-12 2008-07-16 深圳迈瑞生物医疗电子股份有限公司 基于二维分析的b图像均衡方法和系统结构
WO2011013346A1 (ja) * 2009-07-29 2011-02-03 パナソニック株式会社 超音波診断装置

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110288581A (zh) * 2019-06-26 2019-09-27 电子科技大学 一种基于保持形状凸性水平集模型的分割方法
CN110288581B (zh) * 2019-06-26 2022-11-04 电子科技大学 一种基于保持形状凸性水平集模型的分割方法
CN116386043A (zh) * 2023-03-27 2023-07-04 北京市神经外科研究所 一种脑神经医疗影像胶质瘤区域快速标记方法及系统

Also Published As

Publication number Publication date
CN102871686B (zh) 2015-08-19
US20150023578A1 (en) 2015-01-22
US20150023577A1 (en) 2015-01-22
CN102920477B (zh) 2015-05-20
WO2013131421A1 (zh) 2013-09-12
CN102920477A (zh) 2013-02-13
CN102871686A (zh) 2013-01-16

Similar Documents

Publication Publication Date Title
WO2013131420A1 (zh) 医学影像的目标区域边界确定装置和方法
Elen et al. Three-dimensional cardiac strain estimation using spatio–temporal elastic registration of ultrasound images: A feasibility study
Chalana et al. A multiple active contour model for cardiac boundary detection on echocardiographic sequences
CN102763135B (zh) 用于自动分割和时间跟踪的方法
US7043063B1 (en) Non-rigid motion image analysis
Hansegård et al. Semi-automated quantification of left ventricular volumes and ejection fraction by real-time three-dimensional echocardiography
Giannini et al. A fully automatic algorithm for segmentation of the breasts in DCE-MR images
US20160140751A1 (en) Automated 3D Reconstruction of the Cardiac Chambers from MRI and Ultrasound
US9406146B2 (en) Quantitative perfusion analysis
US9147258B2 (en) Methods and systems for segmentation in echocardiography
CN108830852B (zh) 三维超声肿瘤辅助测量系统及方法
Linguraru et al. Liver and tumor segmentation and analysis from CT of diseased patients via a generic affine invariant shape parameterization and graph cuts
Huang et al. Automatic contouring for breast tumors in 2-D sonography
US8189885B2 (en) Apparatus and method for computing regional statistical distribution over a mean anatomic space
CN108961278B (zh) 基于影像数据的腹壁肌肉分割的方法及其系统
WO2011106622A1 (en) Automatic quantification of mitral valve dynamics with real-time 3d ultrasound
Almeida et al. Left-atrial segmentation from 3-D ultrasound using B-spline explicit active surfaces with scale uncoupling
KR101128662B1 (ko) 좌심실 내외벽 경계 검출장치 및 그 방법
Mosbech et al. Automatic segmentation of abdominal adipose tissue in MRI
Barbosa et al. Quantification of left ventricular volume and global function using a fast automated segmentation tool: validation in a clinical setting
Sotaquirá et al. Anatomical regurgitant orifice detection and quantification from 3-D echocardiographic images
Gao et al. Morphological analysis of the papillary muscles and the trabeculae
CN112336378B (zh) 一种用于动物超声诊断的m型超声心动图处理方法和系统
Moltz et al. Segmentation of juxtapleural lung nodules in ct scan based on ellipsoid approximation
Pednekar et al. Intensity and morphology-based energy minimization for the automatic segmentation of the myocardium

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 13757832

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 14383060

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 13757832

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205 DATED 11/11/2014)

122 Ep: pct application non-entry in european phase

Ref document number: 13757832

Country of ref document: EP

Kind code of ref document: A1