CN101278316B - System and method for automatic segmentation of vessels in breast MR sequences - Google Patents
System and method for automatic segmentation of vessels in breast MR sequences Download PDFInfo
- Publication number
- CN101278316B CN101278316B CN2006800367673A CN200680036767A CN101278316B CN 101278316 B CN101278316 B CN 101278316B CN 2006800367673 A CN2006800367673 A CN 2006800367673A CN 200680036767 A CN200680036767 A CN 200680036767A CN 101278316 B CN101278316 B CN 101278316B
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- alpha
- mtd
- msubsup
- Prior art date
- Legal status (The legal status 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 status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 230000011218 segmentation Effects 0.000 title description 19
- 210000000481 breast Anatomy 0.000 title description 11
- 239000011159 matrix material Substances 0.000 claims abstract description 35
- 230000002708 enhancing effect Effects 0.000 abstract 1
- 230000003902 lesion Effects 0.000 description 20
- 230000006870 function Effects 0.000 description 17
- 238000001514 detection method Methods 0.000 description 9
- 230000002792 vascular Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 230000010354 integration Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000002595 magnetic resonance imaging Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 239000002872 contrast media Substances 0.000 description 3
- 230000004807 localization Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000002583 angiography Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000002591 computed tomography Methods 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001493 electron microscopy Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002445 nipple Anatomy 0.000 description 1
- 230000010412 perfusion Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012285 ultrasound imaging Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
- Image Processing (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
A method for segmenting digitized images includes providing (61) a digitized image, selecting (62) a point with a median enhancement greater than a predefined threshold, wherein a contrast enhancing agent was applied to the subject matter of said digitized image prior to acquisition of said image, defining (63) a shape matrix for the selected point in said image from moments of the intensities in a window of points about said selected point, calculating (64) eigenvalues of said shape matrix, determining (65) an eccentricity of a structure underlying said point from said eigenvalues, and segmenting (67) said image based on said eccentricity values, wherein the steps of defining a shape matrix, calculating eigenvalues of said shape matrix, and determining the eccentricity of the underlying structure are repeated (66) for all points in said image.
Description
Cross reference to related U.S. applications
The present application claims priority from U.S. provisional application Serial No. 60/764,122, "Automatic segmentation of vessel branch MR sequences as a false positive detection technique for Automatic deletion and segmentation using", filed on 2006, 2, 1, and Hermosillo et al, and U.S. provisional application Serial No. 60/704,930, filed on 2005, 8, 2, the contents of both U.S. provisional applications are hereby incorporated by reference.
Technical Field
The present invention is directed to segmentation of digitized medical images.
Background
Contrast enhanced MR sequences are powerful diagnostic tools for detecting lesions in the breast. Typically, diagnosis is initiated by identifying suspicious enhancement regions in post-enhancement (post contrast) acquisitions relative to pre-enhancement (pre-contrast) acquisitions. Automating this process is therefore an action that the computer aided detection system needs to perform. A difficulty with such a system is the fact that many non-suspicious structures, in addition to lesions, are enhanced in the enhanced image. Most of these structures are vessels. Vessels are the main type of false positive structures that occur when automatically detecting lesions as areas that are enhanced after contrast agent injection.
Dynamic subtraction of the enhanced T1 weighted images is routinely performed as part of a protocol to assess breast lesions using Magnetic Resonance Imaging (MRI). Since lesions usually contain high vascularity, the perfusion of contrast agent makes the lesion appear brighter than the background, and therefore this modality is quite sensitive. Automatically segmenting lesions can provide radiologists with accurate automatic measurements and make these measurements more consistent from reader to reader. A region growing segmentation algorithm or even a simple thresholding may be used to segment those lesions if not due to the fact that the segmentation penetrates the vessels as a result of their attachment to those lesions. Thus, removing the vessel can ease the segmentation task. On the other hand, automatic detection of lesions requires the ability to distinguish the lesion from various types of normal structures that are also enhanced with contrast agents. These normal structures include the parenchymal tissue of the breast, the vessels, the area under the nipple, and the area around the heart. There has been interest in developing automated methods and the like for segmenting vessel structures in modalities like CT or MR angiography and the like. The literature on this topic is very rich and describes both automated and semi-automated methods, which cover a very broad range of models, assumptions and techniques. In a clinical workflow environment, the extraction of vascular structures should be fully automatic and require no more than a few seconds of computational time. One technique that performs well, can be easily verified with clinical data, and is easy to implement includes the use of moments, which are rarely reported in research literature. Previous moment-based methods include using moment invariants (momenting) to extract and characterize vessels in infrared images of laser-heated skin, using geometric moments to extract vascular structures from large CT datasets, and characterizing the vessels and computing multi-resolution moment filters for extracting linear structures from noisy 2D images.
The use of geometric moments to extract image structure differs among the methods proposed in the literature. Often, the moment of inertia is calculated on a binarized image (binarized image) obtained after thresholding. The problem with this is that the threshold is often difficult to select and may not allow detection of small vessels, as a low threshold will cause smaller vessels (which tend to have lower intensity) to merge with adjacent structures. Another problem with thresholding is that structures become "pixilated", i.e. exhibit sharp edges that make the calculation of their shape inaccurate relative to the true shape of the underlying structure.
An alternative to thresholding is to calculate moments using the image intensity function f as a density function. However, in a region where the signal-to-noise (SN) ratio is low, it is difficult to establish a threshold value regarding the eccentricity of the fitting ellipse to detect an elongated structure. For example, fig. 1(a) depicts a MIP of a sub-volume extracted from an actual image surrounding a vascular connection. The top row shows the initial voxel values interpolated using nearest neighbors. The middle row shows the binary image obtained after manual thresholding. The threshold is adjusted to capture both vessels, a task that is difficult to achieve automatically. The pixelation effect of thresholding is significant, which affects the accuracy of the shape descriptor. The third row shows the same sub-block using a more complex interpolation scheme.
Disclosure of Invention
Exemplary embodiments of the invention described herein generally include systems and methods for automatically detecting bright tubular structures, and automatically segmenting vessels in a breast MR sequence based on geometric moments used to extract tubular structures from images. The method according to an embodiment of the invention is based on eigenvalues of the shape tensor and harmonically eliminates the need to threshold the image, where the structure can be reliably recovered at very low signal-to-noise (SN) ratios. The method according to embodiments of the invention does not depend on the first order image derivatives, like the method based on the eigenvalues of the average structure tensor, or on the second order image derivatives, like the method based on the eigenvalues of the Hessian (Hessian), and avoids the smoothing of the output inherent to the method based on the Hessian or the structure tensor. The method according to embodiments of the invention can run fast, requiring only a few seconds per sequence. The motion corrected breast MR sequence based on the test results shows that the method according to embodiments of the invention reliably segments vessels while keeping the lesions intact, and is stronger than differential techniques in terms of sensitivity and localization accuracy, and less sensitive to scale selection parameters.
In accordance with an aspect of the present invention, there is provided a method for segmenting a digitized image, the method comprising: providing a digitized image, the digitized image comprising a plurality of intensities corresponding to a domain of points on a three-dimensional grid; defining a shape matrix for a selected point in the image from moments of intensity in a window of points surrounding the selected point; calculating eigenvalues of the shape matrix; determining an eccentricity of a structure below the point from the feature value; and segmenting the image based on the eccentricity values, wherein the steps of defining a shape matrix, calculating eigenvalues of the shape matrix, and determining eccentricity of underlying structures are repeated for all points in the image.
In accordance with another aspect of the invention, the selected points have a median enhancement greater than a predetermined threshold, wherein a contrast-enhancing agent is applied to the subject matter (objectmate) of the digitized image prior to acquiring the image.
In accordance with another aspect of the invention, a median enhancement is calculated by taking the difference of the median value of the contrast enhanced image and the median value of the enhanced image before enhancement and normalizing the difference to within a predetermined range.
In accordance with another aspect of the present invention, the shape matrix SαIs defined as
Wherein
Wherein the moment mp，q，r，αIs defined as
Where w is a window function with tight support (contact support) p, q, r μ 0, and α μ 1.
In accordance with another aspect of the invention, the integral is computed by summing over a limited neighborhood around each point.
In accordance with yet another aspect of the invention, the window function is defined by:
wherein v isx、vy、vzIs the image dot spacing, Nx、Ny、NzIs a defined non-negative integer in which the window size contains the largest diameter of interest.
In accordance with yet another aspect of the invention, the method includes calculating the moments using nearest neighbor interpolation and correcting the shape matrix according to
Wherein v isx、vy、vzIs the image dot pitch.
In accordance with yet another aspect of the invention, the method includes computing the moments using tri-linear interpolation.
In accordance with yet another aspect of the present invention, α ═ 1, and the shape matrix is corrected in accordance with the following equation
Wherein v isx、vy、vzIs the image dot pitch.
In accordance with another aspect of the present invention, a program storage device readable by a computer, tangibly embodying a program of instructions executable by the computer to perform method steps for segmenting a digitized image is provided.
Drawings
Fig. 1(a) shows a MIP of a sub-volume extracted from a real image around a vascular junction according to an embodiment of the invention.
FIG. 1(b) illustrates a simulated vessel and its detection without thresholding using moment of inertia according to an embodiment of the invention.
FIG. 2 shows basis functions for 1D linear interpolation in accordance with an embodiment of the present invention.
Fig. 3(a) - (c) show segmentation of a large lesion in accordance with an embodiment of the present invention.
Fig. 4(a) - (c) illustrate segmentation of multiple small lesions in accordance with an embodiment of the present invention.
Fig. 5 illustrates segmentation of vascular structures in breast MRI using shape tensors in accordance with an embodiment of the present invention.
FIG. 6 shows a flow diagram of a method of moment-based segmentation in accordance with an embodiment of the invention.
FIG. 7 is a block diagram of an exemplary computer system for implementing a moment-based segmentation method in accordance with an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the invention described herein generally include systems and methods for automatically detecting bright tubular structures, and their automatic segmentation of vessels in a breast MR sequence. A method according to an embodiment of the invention is based on eigenvalues of the shape tensor. This method can be compared with the method based on eigenvalues of the average sea race and the method based on eigenvalues of the average structure tensor. A sea race defined in terms of second order derivatives can be considered as a structural descriptor of the second order. Similarly, the structure tensor is a first order structure descriptor. The shape tensor can be viewed as a structural descriptor of the zeroth order.
As used herein, the term "image" refers to multi-dimensional data composed of discrete image elements (e.g., pixels of a 2D image and voxels of a 3D image). The image may be a medical image of the subject collected, for example, by computer tomography, magnetic resonance imaging, ultrasound imaging, or any other medical imaging system known to those skilled in the art. The image may also be provided by a non-medical environment, such as by a remote sensing system, electron microscopy, or the like. Although the image may be considered to be from R3To R, but the method of the invention is not limited to such images and can be applied to images of any dimension, such as 2D pictures or 3D volumes. For 2-dimensional or 3-dimensional images, the domain of the image is typically a 2-dimensional or 3-dimensional rectangular array, where each pixel or voxel can be addressed with reference to a set of 2 or 3 mutually orthogonal axes. The term "digital" or "digitized" as used herein shall mean an image or volume (as appropriate) in a digital or digitized format obtained by a digital acquisition system or by conversion from an analog image.
The method according to an embodiment of the invention exerts an influence on the image intensity by calculating the second order geometrical moments of the underlying (bright) structures. The method can be used to determine the initial position of the targetThe enhanced image after enhancement of (2) applies a threshold value to the obtained binarized image, but the method may be applied without this threshold value. The eigenvalues of the second order geometric moments are classical tools of shape characterization in object recognition. However, these feature values have never been used as a filter for extracting an image structure. Given a binary image, a small sub-volume around each pixel (whose size is related to the structure of interest) is considered, and the shape tensor at that location is defined as the second moment of the location of the bright voxel relative to the center of the sub-volume. For voxels whose central pixel is bright and sufficiently close to the center of the underlying shape, the eigenvalues of the shape tensor are computed and the value λ is calculated1-λ2/(λ1+λ2) Giving the filter a response of λ2＞λ1Is the maximum eigenvalue.
According to an embodiment of the invention, the geometric 3D moments may be defined as:
where w is a positive and symmetric window function with tight sets p, q, r μ 0 and α μ 1 providing localization. The shape tensor of order alpha is defined from these moments
Wherein
This matrix is symmetric so all its eigenvalues are real. Let three eigenvalues be λ3＞λ2＞λ1μ 0, then the filter response can be defined as
For linear or cylindrical structures, such as vessels, CThreadλ1。
According to an embodiment of the invention, based on Sα(α > 1) feature value 0[ λ1[λ2[λ3The eccentricity of the shape of the lower layer is calculated. As α gets larger, higher intensity values are given more importance, acting almost like thresholding. As shown in the simulation experiment of fig. 1(b), a high value of α can cope with a very low SN ratio, where a synthetic tubular structure with increased uniform noise is detected using a classical (classical) inertia matrix and a shape tensor of α ═ 15.
Fig. 1(b) shows a simulated vessel and its detection using the standard moment of inertia without thresholding and using a shape tensor of α -15. Columns from left to right show: (1) a central slice of the initial synthetic volume, (2) its Maximum Intensity Projection (MIP), (3) a MIP with the volume of the vessel removed by the standard moment method, (4) a MIP of the vessel detected by the moment method, (5) a MIP with the volume of the vessel removed using a shape tensor of α ═ 15,and (6) MIP of the vessel detected using the shape tensor α -15. The six rows represent increasing levels of additive uniform noise, giving the following SN ratios from top to bottom, respectively: (1)56.3, (2)36.7, (3)20.4, (4)11.6, (5)5.5 and (6)0.8 dB. For each algorithm, the threshold on the eccentricity of the shape is the same from row to row. In all cases, for S15The detection standard is <math><mrow>
<mfrac>
<msub>
<mi>λ</mi>
<mn>3</mn>
</msub>
<msub>
<mi>λ</mi>
<mn>2</mn>
</msub>
</mfrac>
<mo>></mo>
<mn>15</mn>
<mo>,</mo>
</mrow></math> And for a value corresponding to S1The detection criterion is <math><mrow>
<mfrac>
<msub>
<mi>λ</mi>
<mn>3</mn>
</msub>
<msub>
<mi>λ</mi>
<mn>2</mn>
</msub>
</mfrac>
<mo>></mo>
<mn>2</mn>
<mo>.</mo>
</mrow></math> This improved detection performance has been noted in practical situations.
In practice, the above integration is usually replaced by a sum over a limited neighborhood around each voxel, since f is only known at the voxel location. The localization function can be given by the following equation for all experimental assumptions:
wherein v isx、vy、vzIs the image voxel spacing, and Nx、Ny、NzIs a non-negative integer defined such that the window size encompasses the largest diameter of interest. Then, given an image, consider a small sub-block around each pixel and define
Where ρ isijkIs the image value at the voxel corresponding to the index i, j, k. Calculating the eigenvalue 0[ lambda ] of the matrix1[λ2[λ3
Wherein the above values are calculatedBut is calculated using the sum moment. The eccentricity or elongation of the understructure can be measured by the classical eccentricity measure ∈ ═ λ (λ)3-λ2)/(λ3+λ2) Measured by taking a value between 0 and 1, or simply by taking the ratio lambda3/λ2(assume λ)2＞0)
Since moment-based methods do not assume the differentiability of the image intensity function f, a simple interpolation scheme, such as nearest neighbor interpolation or trilinear interpolation, can be used to compute the integral of the interpolated function, rather than the sum over the voxel values. It is expected that the values using these integrals have better accuracy, especially in the case of trilinear interpolation. Using the equation
It can be seen that for nearest neighbor interpolation integration, the above matrix
Should be replaced by
In the case of trilinear interpolation, the function f consists of ∑i，j，kρijkgijkGiven, where i, j, k are indices of image voxels, ρijkIs an image value of a voxel, and
then write out <math><mrow>
<munder>
<mo>∫</mo>
<mi>xyz</mi>
</munder>
<mrow>
<mo>(</mo>
<mo>·</mo>
<mo>)</mo>
</mrow>
<mo>≡</mo>
<msubsup>
<mo>∫</mo>
<mrow>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>z</mi>
</msub>
</mrow>
<mrow>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>z</mi>
</msub>
</mrow>
</msubsup>
<msubsup>
<mo>∫</mo>
<mrow>
<mrow>
<mo>(</mo>
<mi>j</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>y</mi>
</msub>
</mrow>
<mrow>
<mrow>
<mo>(</mo>
<mi>j</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>y</mi>
</msub>
</mrow>
</msubsup>
<msubsup>
<mo>∫</mo>
<mrow>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>x</mi>
</msub>
</mrow>
<mrow>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<msub>
<mi>v</mi>
<mi>x</mi>
</msub>
</mrow>
</msubsup>
<mrow>
<mo>(</mo>
<mo>·</mo>
<mo>)</mo>
</mrow>
<mi>dxdydz</mi>
<mo>:</mo>
</mrow></math>
So that, for a tri-linear interpolation in case of α ═ 1, the matrix
Should be replaced by
In the case of a general shape tensor (α > 1) using trilinear interpolation, where f is more complicatedαBy (sigma)i，j，kρijkgijk)αIt is given. Although the corresponding integral can still be calculated in closed form, the complexity increases significantly. In order to calculate the corresponding shape tensor according to an embodiment of the present invention, it should be noted that g is calculated as in the case where α ═ 1 described aboveijk αThe moment of (a) is no longer useful. Continuing, the moments can be obtained using a less straightforward method but can be generalized to α > 1. This can be done in the 1D case and can be generalized directly to the 2D and 3D cases. Let it be assumed that for k < i-2 or k > i +2, ρkWhen the value is 0, then obtain
The above four integrals can be obtained from the 3 piecewise linear basis functions shown in fig. 2. Referring to the figure, the first basis function gi-1Is defined in the domain (i-2) vxTo ivxUpper, second basis function giIs defined in the domain (i-1) vxTo (i +1) vxUpper, and a third function gi+1Is defined in the field ivxTo (i +2) vxThe above.
The method of calculating the integral can be generalized to α > 1. For example, it is possible to calculate
In a similar manner to that described above,
and is
Although it is possible to find a generalized formula for the 3D case and a given α > 1, the complexity of the final polynomial is quite high for potential accuracy improvement. In the 2D case, the above four integrations become sixteen integrations, and in 3D become sixty-four integrations.
The method according to an embodiment of the invention has been tested on over 100 motion corrected breast MR dynamic sequences. The results obtained show that the vessel can be reliably segmented while leaving the lesion intact. In accordance with an embodiment of the invention, the moments are calculated over a sliding window of fixed size, but only points where the median enhancement is above a given threshold are considered. This threshold may be chosen low enough to detect even small vessels. This is not difficult to set up as it does not rely on calculations, but only speeds up the whole process by processing fewer voxels. The median enhancement is calculated by subtracting the median of the post-enhancement acquisitions from the value of the pre-enhancement acquisitions at each image voxel. This difference is then normalized by applying an affine function so that the final enhancement is in the range 0, 200. Fig. 3(a) - (c), fig. 4(a) - (c) and fig. 5 show several representative examples of the results.
Fig. 3(a) - (c) show segmentation of large lesions, while fig. 4(a) - (c) show segmentation of multiple small lesions. For these figures, panel (panel) (a) shows the original enhanced image that was thresholded, panel (b) shows the detected vessels, and panel (c) shows the lesions from which the vessels have been removed.
Fig. 5 shows the segmentation of the vascular structure in breast MRI using a shape tensor of α -6. The three columns show orthogonal views of the same patient. The first row shows the median enhanced initial MIP. The second row shows the same volume of the vessel being automatically removed. The third row shows the MIP of the removed vessel alone. Note that vessels with very different diameters and enhancement levels are correctly segmented. By taking the eigenvalues of the shape tensor so that λ3/λ2Locations > 3.
In each of these figures, note how small the vessel is segmented correctly and how small the spherical structure remains intact. As a further verification, in 40 cases, vascular structures were extracted according to the method of the present invention, three radiologists observed these structures and marked a total of 75 lesions. In all cases the vessel was correctly segmented and all marked lesions remained intact.
A flow diagram of a moment-based segmentation method in accordance with an embodiment of the present invention is shown in fig. 6. Referring now to the figure, at step 61, an image to be segmented is provided. The shape tensor is computed for the voxel in the image whose median contrast enhancement exceeds a predetermined threshold as determined at step 62. In step 63, the moments from which the shape tensor is defined are computed in a fixed-size window around the selected voxel. At step 64, the eigenvalues of the shape tensor are computed, and at step 65, the eccentricity of the underlying structure is determined. At step 66, the process loops until each voxel has been processed. At step 67, the image is segmented based on the eccentricities derived from the shape tensor.
The moment-based approach to extracting local shape information may be compared to higher order image derivative-based approaches. For example, the Gradient Square Tensor (GST) (or structure tensor) has been proposed as a robust method of estimating the dimension of a local structure. This method is based on first order derivatives and may therefore be referred to as first order structural descriptors. Feature values of the sea race also provide local image structure information, as well as principal curvature at the iso-level (isovelel) at a given point. The sea race and principal curvatures are defined in terms of second order derivatives and may thus be referred to as structural descriptors of the second order. The shape tensor can be viewed as a structural descriptor of the zeroth order. The shape tensor is based on integrals and thus has very robust properties for noise compared to methods based on first or second derivatives. In addition, no assumption of any differentiability on the image function is required, which simplifies the modeling. The problem with the shape tensor-based approach is that the connection is not detected. In addition, a better understanding is needed to determine whether the geometry attributes can be computed from the eigenvalues of the shape tensor for α > 1.
It is to be understood that the present invention may be implemented in various forms of hardware, software, firmware, special purpose processes, or a combination thereof. In one embodiment, the present invention may be implemented in software as an application program tangibly embodied on a computer-readable program storage device. The application program may be uploaded to, and executed by, a machine comprising any suitable architecture.
FIG. 7 is a block diagram of an exemplary computer system for implementing a moment-based segmentation method in accordance with an embodiment of the present invention. Referring now to FIG. 7, a computer system 71 for implementing the present invention may include, among other things, a Central Processing Unit (CPU)72, a memory 73, and an input/output (I/O) interface 74. The computer system 71 is typically coupled through the I/O interface 74 to a display 75 and various input devices 76, such as a mouse and a keyboard. The support circuits may include circuits such as cache, power supplies, clock circuits, and a communications bus. The memory 73 may include Random Access Memory (RAM), Read Only Memory (ROM), hard disk drives, tape drives, and the like, or a combination thereof. The present invention may be implemented as a routine 77 stored in memory 73 and executed by the CPU 72 to process signals from the signal source 78. Similarly, the computer system 71 is a general purpose computer system that becomes a specific purpose computer system when executing the routine 77 of the present invention.
The computer system 71 also includes an operating system and microinstruction code. The various processes and functions described herein may either be part of the microinstruction code or part of the application program (or a combination thereof) that is executed via the operating system. In addition, various other peripheral devices may be connected to the computer platform such as an additional data storage device and a printing device.
It is to be further understood that, because some of the constituent system components and method steps depicted in the accompanying figures may be implemented in software, the actual connections between the system components (or the process steps) may differ depending upon the manner in which the present invention is programmed. Given the teachings of the present invention provided herein, one of ordinary skill in the related art will be able to contemplate these and similar implementations or configurations of the present invention.
Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art will appreciate that various modifications and substitutions can be made thereto without departing from the spirit and scope of the present invention as set forth in the appended claims.
Claims (20)
1. A method of segmenting a digitized image, the method comprising the steps of:
providing a digitized image, the digitized image comprising a plurality of intensities corresponding to a domain of points on a three-dimensional grid;
defining a shape matrix for a selected point in the image from moments of intensity in a window of points surrounding the selected point;
calculating eigenvalues of the shape matrix;
determining the eccentricity of the structure under said point from said characteristic values, and
segmenting the image based on the eccentricity values, wherein the steps of defining a shape matrix, calculating eigenvalues of the shape matrix, and determining eccentricity of underlying structures are repeated for all points in the image.
2. The method of claim 1, wherein the selected points have a median enhancement greater than a predetermined threshold, wherein a contrast-enhancing agent is applied to the subject matter of the digitized image prior to acquiring the image.
3. The method of claim 2, wherein the median enhancement is calculated by taking the difference of the median of the contrast enhanced image and the median of the pre-enhanced image and normalizing the difference to within a predetermined range.
4. The method of claim 1, wherein the shape matrix SαIs defined as
Wherein,
wherein the moment mp，q，r，αIs defined as
Where w is a window function with tight support sets p, q, r μ 0 and α μ 1.
5. The method of claim 4, wherein the integral is calculated by summing over a limited neighborhood around each point.
6. The method of claim 4, wherein the window function is defined by
Wherein v isx、vy、vzIs the image dot spacing, Nx、Ny、NzIs a defined non-negative integer in which the window size contains the largest diameter of interest.
7. The method of claim 4, further comprising calculating the moments using nearest neighbor interpolation and correcting the shape matrix according to
Wherein v isx、vy、vzIs the image dot pitch.
8. The method of claim 4, further comprising computing the moments using trilinear interpolation.
9. The method of claim 8, wherein α -1, and the shape matrix is corrected according to
Wherein v isx、vy、vzIs the image dot pitch.
10. A method of segmenting a digitized image, the method comprising the steps of:
providing a digitized image, the digitized image comprising a plurality of intensities corresponding to a domain of points on a three-dimensional grid;
defining a shape matrix of a selected point in the image from moments of intensity in a window of points surrounding the selected point, wherein the shape matrix SαIs defined as
Wherein,
wherein the moment mp，q，r，αIs defined as
Where w is a window function with tight-branches p, q, r μ 0 and α μ 1;
calculating eigenvalues of the shape matrix; and
determining an eccentricity of the structure below the point from the feature values.
11. The method of claim 10, further comprising repeating the steps of defining a shape matrix, calculating eigenvalues of the shape matrix, and determining the eccentricity of underlying structures for all points in the image, and segmenting the image based on the eccentricity values.
12. A system for segmenting a digitized image, the system comprising:
means for providing a digitized image, the digitized image comprising a plurality of intensities corresponding to a domain of points on a three-dimensional grid;
means for defining a shape matrix for a selected point in the image from moments of intensity in a window of points surrounding the selected point;
means for calculating eigenvalues of the shape matrix;
means for determining the eccentricity of the structure below said point from said characteristic values, and
means for segmenting the image based on the eccentricity values, wherein the steps of defining a shape matrix, calculating eigenvalues of the shape matrix, and determining eccentricity of underlying structures are repeated for all points in the image.
13. The system of claim 12, wherein the selected points have a median enhancement greater than a predetermined threshold, wherein a contrast-enhancing agent is applied to the subject matter of the digitized image prior to acquiring the image.
14. The system of claim 13, wherein the median enhancement is calculated by taking the difference of the median of the contrast enhanced image and the median of the pre-enhanced image and normalizing the difference to within a predetermined range.
15. The system of claim 12, wherein the shape matrix SαIs defined as
Wherein,
wherein the moment mp，q，r，αIs defined as
Where w is a window function with tight support sets p, q, r μ 0 and α μ 1.
16. The system of claim 15, wherein the integral is calculated by a sum over a limited neighborhood around each point.
17. The system of claim 15, wherein the window function is defined by
Wherein v isx、vy、vzIs the image dot spacing, Nx、Ny、NzIs a defined non-negative integer in which the window size contains the largest diameter of interest.
18. The system of claim 15, further comprising means for calculating the moments using nearest neighbor interpolation and correcting the shape matrix according to,
wherein v isx、vy、vzIs the image dot pitch.
19. The system of claim 15, further comprising means for calculating the moments using trilinear interpolation.
20. The system of claim 19, wherein α -1, and the shape matrix is corrected according to
Wherein v isx、vy、vzIs the image dot pitch.
Applications Claiming Priority (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US70493005P | 2005-08-02 | 2005-08-02 | |
| US60/704,930 | 2005-08-02 | ||
| US76412206P | 2006-02-01 | 2006-02-01 | |
| US60/764,122 | 2006-02-01 | ||
| PCT/US2006/029636 WO2007016442A2 (en) | 2005-08-02 | 2006-07-27 | System and method for automatic segmentation of vessels in breast mr sequences |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101278316A CN101278316A (en) | 2008-10-01 |
| CN101278316B true CN101278316B (en) | 2012-06-06 |
Family
ID=39243794
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2006800367673A Expired - Fee Related CN101278316B (en) | 2005-08-02 | 2006-07-27 | System and method for automatic segmentation of vessels in breast MR sequences |
Country Status (4)
| Country | Link |
|---|---|
| JP (1) | JP5132559B2 (en) |
| CN (1) | CN101278316B (en) |
| AU (1) | AU2006275606B2 (en) |
| CA (1) | CA2617671C (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113657138B (en) * | 2020-05-12 | 2024-05-21 | 哈尔滨工程大学 | Radiation source individual identification method based on equipotential star map |
| CN114897780B (en) * | 2022-04-12 | 2023-04-07 | 南通大学 | MIP sequence-based mesenteric artery blood vessel reconstruction method |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030053669A1 (en) * | 2001-07-18 | 2003-03-20 | Marconi Medical Systems, Inc. | Magnetic resonance angiography method and apparatus |
| DE10083899T1 (en) * | 1999-11-19 | 2002-03-07 | Gen Electric | Method and device for reformatting tubular volumetric bodies |
| US7274810B2 (en) * | 2000-04-11 | 2007-09-25 | Cornell Research Foundation, Inc. | System and method for three-dimensional image rendering and analysis |
-
2006
- 2006-07-27 CN CN2006800367673A patent/CN101278316B/en not_active Expired - Fee Related
- 2006-07-27 JP JP2008525070A patent/JP5132559B2/en not_active Expired - Fee Related
- 2006-07-27 AU AU2006275606A patent/AU2006275606B2/en not_active Ceased
- 2006-07-27 CA CA2617671A patent/CA2617671C/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CA2617671C (en) | 2014-10-07 |
| AU2006275606B2 (en) | 2010-08-05 |
| CN101278316A (en) | 2008-10-01 |
| AU2006275606A1 (en) | 2007-02-08 |
| JP2009502430A (en) | 2009-01-29 |
| JP5132559B2 (en) | 2013-01-30 |
| CA2617671A1 (en) | 2007-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | Handbook of medical image computing and computer assisted intervention | |
| Kostis et al. | Three-dimensional segmentation and growth-rate estimation of small pulmonary nodules in helical CT images | |
| US6766043B2 (en) | Pleural nodule detection from CT thoracic images | |
| US10297022B2 (en) | Method and system for analysis of volumetric data | |
| Ochs et al. | Automated classification of lung bronchovascular anatomy in CT using AdaBoost | |
| US9230320B2 (en) | Computer aided diagnostic system incorporating shape analysis for diagnosing malignant lung nodules | |
| US10388017B2 (en) | Advanced treatment response prediction using clinical parameters and advanced unsupervised machine learning: the contribution scattergram | |
| US7457448B2 (en) | Method and system for wavelet based detection of colon polyps | |
| JP2009512527A (en) | Image registration method, algorithm for performing the image registration method, program for registering an image using the algorithm, and biomedical image handling method for reducing image artifacts due to object movement | |
| WO2019182520A1 (en) | Method and system of segmenting image of abdomen of human into image segments corresponding to fat compartments | |
| Serlie et al. | Electronic cleansing for computed tomography (CT) colonography using a scale-invariant three-material model | |
| US7711164B2 (en) | System and method for automatic segmentation of vessels in breast MR sequences | |
| CN117649400B (en) | Image histology analysis method and system under abnormality detection framework | |
| CN101278316B (en) | System and method for automatic segmentation of vessels in breast MR sequences | |
| Bulpitt et al. | Spiral CT of abdominal aortic aneurysms: comparison of segmentation with an automatic 3D deformable model and interactive segmentation | |
| EP2005389B1 (en) | Automatic cardiac band detection on breast mri | |
| Staal et al. | Automatic rib segmentation in CT data | |
| Campbell et al. | 3-dimensional throat region segmentation from MRI data based on fourier interpolation and 3-dimensional level set methods | |
| Herman et al. | Maximum a posteriori image reconstruction from projections | |
| Gala et al. | Three Dimensional MR Brain Image Reconstruction and Interactive Segmentation Based Tumor Region Extraction Using Active Contour Models | |
| Li et al. | Automatic segmentation of cerebral ischemic lesions from diffusion tensor MR images | |
| Mohd et al. | BOUNDARY DETECTION OF KIDNEY ULTRASOUND IMAGE BASED ON VECTOR GRAPHIC APPROACH | |
| BUlpjtta et al. | Spiral CT of abdominal aortic aneurysms: Comparison of segmentation with an automatic 3D deformable model and interactive segmentation | |
| Hermosillo et al. | Automatic segmentation of vessels in breast MR sequences as a false positive elimination technique for automatic lesion detection and segmentation using the shape tensor |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120606 Termination date: 20180727 |