CN102038501B - Correcting method and device of background phase - Google Patents
Correcting method and device of background phase Download PDFInfo
- Publication number
- CN102038501B CN102038501B CN 200910180891 CN200910180891A CN102038501B CN 102038501 B CN102038501 B CN 102038501B CN 200910180891 CN200910180891 CN 200910180891 CN 200910180891 A CN200910180891 A CN 200910180891A CN 102038501 B CN102038501 B CN 102038501B
- Authority
- CN
- China
- Prior art keywords
- static
- phase
- plane
- phase image
- static pixels
- 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.)
- Active
Links
Images
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a correcting method and device of background phase; the method comprises the steps of: after obtaining a phase image of a speed code, selecting a static pixel point corresponding to a static organization in the phase image; performing plane fitting according to the selected static pixel to obtain a plane reflecting a background phase offset error of the static pixel; subtracting the phase image with the plane obtained by the fitting, so as to obtain a phase image corrected by the background phase. By the method of the present invention, the static pixel corresponding to the static organization in the image is selected by kernel clustering algorithm; the robustness of the kernel clustering algorithm is quite good, so that the static pixel can be selected stably, and the correction to the background phase is realized; simultaneously, by the method of the present invention, sampling to a single static body model is not needed, and operations during clinical use are greatly convenient.
Description
Technical field
(magnetic resonance angiography, MRA) technology particularly relate to a kind of background phase to phase-contrast MRA and carry out gauged method and apparatus to the present invention relates to magnetic resonance angiography.
Background technology
In the quantitative application of magnetic resonance film (MR CINE) phase-contrast speed, the phase image of velocity encoded cine can bring out the field owing to the eddy current of not compensated and cause distortion, thereby phase shift errors occurs, and this error is identified as the non-zero speed of static tissue again.Owing to the volume flow of trunk is estimated to be determined by time and spatial integration, less flow velocity skew also may cause average external volume flow, stroke volume and kinemic result significant error to occur.Most of scanning devices all are designed to remedy eddy current and bring out the field, but have occurred high-power gradient system in the recent period, and the problem of phase shift errors highlights once more.
The someone adopts the method for post processing to proofread and correct the phase deviation that this vortex flow is brought out now.More common method is to utilize near static tissue zone, tested position to come side-play amount is estimated.Static tissue is meant actionless basically tissue in the process of measuring, such as people's skeleton.Since the spatial variations of side-play amount, and the necessary close area-of-interest in this static tissue zone (Region Of Interest, ROI).But the method is not suitable near the mobile quantitative measurement of the trunk the heart, because lack static tissue near the heart.At this problem, a kind of solution is to independent static phantom sampling, to determine the phase deviation on the vessel position.Yet because layer orientation can have influence on phase deviation, this sampling must repeat different situations, thereby makes sampling need the cost plenty of time, and the data volume of sampling is bigger, operates also more complicated.Therefore, this solution inconvenience is used in clinical practice.
Summary of the invention
The invention provides a kind of bearing calibration of background phase, utilize the kernel clustering algorithm to select static pixels, and carry out background correction automatically, conveniently in clinical practice, use.
The present invention also provides a kind of correcting unit of background phase, utilizes the kernel clustering algorithm to select static pixels, and carries out background correction automatically, conveniently uses in clinical practice.
A kind of bearing calibration of background phase comprises:
After obtaining the phase image of velocity encoded cine, utilize the kernel clustering method to select the static pixels point corresponding in the described phase image with static tissue;
Carry out plane fitting according to selected static pixels, obtain reflecting the plane of static pixels background phase offset error;
Described phase image is deducted the plane that institute's match obtains, obtain through the phase image after the background phase correction.
Wherein, the described method of kernel clustering of utilizing selects the static pixels corresponding with static tissue in the described phase image to comprise:
With the pixel in the described phase image as pixel samples;
Utilize Mercer nuclear that described pixel samples is mapped to high-dimensional feature space;
In feature space, described pixel samples is sorted, and select the part sample of eigenvalue minimum;
The part sample of described eigenvalue minimum is shone upon back image space, obtain the static pixels point.
Wherein, after the static pixels point in having selected described phase image, further comprise: selected static pixels point is screened.
Wherein, describedly carry out plane fitting according to selected static pixels, obtain reflecting that the plane of static pixels background phase offset error comprises: as initial point, carry out plane fitting with selected static pixels, obtain reflecting the plane of static pixels background phase error.
Wherein, adopt least-squares algorithm to carry out plane fitting.
Wherein, described phase image is deducted the plane that institute's match obtains, phase image after obtaining proofreading and correct through background phase comprises: all pixels in the described phase image space are deducted corresponding point on the plane that institute's match obtains, obtain the phase image through phasing
The gauged device of a kind of background phase comprises:
A kernel clustering unit 201 is used for utilizing the kernel clustering algorithm to select the static pixels corresponding with static tissue at the phase image of the velocity encoded cine that has obtained;
A plane fitting unit 202, the static pixels that is used to utilize described kernel clustering unit 201 to select is carried out plane fitting, obtains the plane of a reflection background phase offset error;
A correcting unit 203 is used for described phase image is deducted the plane of the reflection background phase offset error that described plane fitting unit 202 obtains, the phase image after obtaining proofreading and correct through background phase.
Described device also comprises: a screening unit 301 be used for the static pixels corresponding with static tissue that described kernel clustering unit 201 is selected screened, and the static pixels after will screening is sent to described plane fitting unit 202.
As can be seen from the above technical solutions, adopted method and apparatus of the present invention, utilize the kernel clustering algorithm to select in the image and the corresponding static pixels of static tissue, because the robustness of kernel clustering algorithm is very good, so can stably select static pixels, thereby realize correction, adopt the present invention also to need not independent static phantom is sampled simultaneously, greatly facilitate the operation when clinical use background phase.
Description of drawings
To make clearer above-mentioned and other feature and advantage of the present invention of those of ordinary skill in the art by describing the preferred embodiments of the present invention in detail with reference to accompanying drawing below, identical label is represented identical parts, in the accompanying drawing:
Fig. 1 is the gauged method flow diagram of the background phase of the embodiment of the invention;
Fig. 2 is the gauged structure drawing of device of the background phase of the embodiment of the invention;
Fig. 3 is the gauged structure drawing of device of the background phase of another embodiment of the present invention.
The specific embodiment
In order to make technical scheme of the present invention and advantage clearer,, the present invention is further elaborated below in conjunction with drawings and Examples.Should be appreciated that specific embodiment described herein only in order to explanation the present invention, and be not used in qualification the present invention.
The present inventor is in the face of existing technical difficulty, do not stick near original thinking of tested tissue, seeking static tissue or simulation static tissue, but the bearing calibration of the background phase of a novelty has been proposed, this method shows as the hypothesis of steady spatial variations based on background offset, utilizes kernel clustering algorithm (Kernel Clustering Algorithm) to select in the image and the corresponding static pixels of static tissue.Like this, need not independent static phantom is sampled, made things convenient for operation clinically.
Cluster belongs to non-supervised recognition problem, is characterized in that the sample of the input space does not have desired output.Clustering problem makes similar sample be classified as a class according to the tolerance of certain similarity degree.And dissimilar sample belongs to different classes, i.e. the process of the cluster characteristic difference between the sample that places one's entire reliance upon.Relatively more classical clustering method has traditional C Mean Method and fuzzy C-means clustering method.These methods all are not optimized the feature of sample, but directly utilize the feature of sample to carry out cluster.Like this, the effectiveness of above-mentioned these methods depends on the distribution situation of sample to a great extent.
The kernel clustering method has then increased the optimization to sample characteristics.By nuclear the sample of the input space is mapped to high-dimensional feature space.And in feature space, carry out cluster.This method is pervasive, and the kernel clustering method has bigger improvement than classical clustering algorithm on performance, can differentiate preferably by nonlinear mapping, extracts and amplifies useful feature, thereby realize cluster more accurately.
Method of the present invention mainly comprises three steps: at first, behind the phase image that obtains velocity encoded cine, utilize the static pixels in the kernel clustering algorithm selection image; The second, utilize single order fit Plane of selected static pixels structure; At last, from the phase image of velocity encoded cine, deduct the plane of this match, thereby realize phasing.
The present invention will be described in detail below in conjunction with specific embodiment.
Fig. 1 is the gauged method flow diagram of the background phase of the embodiment of the invention.As shown in Figure 1, the method for present embodiment comprises the steps:
According to embodiments of the invention, with in the phase image all pixels as pixel samples, utilize Mercer nuclear that the pixel samples that obtains is mapped to high-dimensional feature space, and in feature space, pixel samples is sorted, and the part sample of selection eigenvalue minimum, and then this part sample shone upon back image space, just obtained the ideal static pixel.
Herein, the static pixels of utilizing the kernel clustering algorithm to obtain is put 15% to 30% of all images pixel normally.
As initial point, adopt the algorithm of plane fitting to carry out plane fitting with selected static pixels point.The plane that obtains by plane fitting is the plane of a reflection background phase offset error.
In an embodiment of the present invention, can adopt the algorithm of multiple plane fitting, for example, can adopt the method for least square algorithm to carry out plane fitting.
In the present embodiment, mention in the above and utilize static pixels that the kernel clustering algorithm obtains the chances are 15% to 30% of all images pixel.In plane fitting, can be with selected whole static pixels points as initial point.
Perhaps, in other embodiments of the invention, also filter out a part in the whole static pixels that can from last step, select as initial point.
The present invention also provides the device of realizing said method.
Fig. 2 is the gauged structure drawing of device of the background phase of the embodiment of the invention.As shown in Figure 2, this device comprises: 201, one plane fitting unit 202, a kernel clustering unit and a correcting unit 203.
Plane fitting unit 202 utilizes selected static pixels to carry out plane fitting, obtains the plane of a reflection background phase offset error, and will reflect that the plane of background phase offset error is sent to correcting unit 203.
Correcting unit 203 deducts the plane that institute's match obtains with phase image, obtains through the phase image after the background phase correction.Specifically, correcting unit 203 deducts corresponding point on the fit Plane with pixels all in the phase image, obtains the phase image through background correction, thereby has realized the correction of background phase.
Fig. 3 is the gauged structure drawing of device of the background phase of another embodiment of the present invention.Difference embodiment illustrated in fig. 3 and embodiment illustrated in fig. 2 only is, has increased a screening unit 301.The static pixels corresponding with static tissue that screening 301 pairs of kernel clustering unit, unit 201 are selected screened, and the static pixels after will screening is sent to plane fitting unit 202.
The present invention is specially adapted to utilize the MR scanning device of high-power gradient system to carry out the quantitative measurement of phase-contrast flow velocity, for example, the main pulmonary artery blood flow is estimated.Utilize this new method, at first can stably select static pixels,, both avoided the artificial not operability of selecting, avoid the inconvenience that independent phantom is scanned again because the robustness of kernel clustering method is very good; Obtain reflecting the plane of background phase deviation then according to selected static pixels, thereby realize correction the image background phase place.
The above only is preferred embodiment of the present invention, not in order to restriction the present invention, all any modifications of being done within the spirit and principles in the present invention, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.
Claims (7)
1. the bearing calibration of a background phase comprises:
Behind the phase image that obtains velocity encoded cine, utilize the kernel clustering method to select static pixels point corresponding in the described phase image with static tissue;
Carry out plane fitting according to selected static pixels point, obtain reflecting the plane of static pixels background phase offset error;
Described phase image is deducted the plane that institute's match obtains, obtain through the phase image after the background phase correction, wherein,
The described method of kernel clustering of utilizing selects static pixels point corresponding with static tissue in the described phase image to comprise:
With the pixel in the described phase image as pixel samples;
Utilize Mercer nuclear that described pixel samples is mapped to high-dimensional feature space;
In described high-dimensional feature space, described pixel samples is sorted, and select the part sample of eigenvalue minimum;
The part sample of described eigenvalue minimum is shone upon back image space, obtain the static pixels point.
2. method according to claim 1 is characterized in that, after the static pixels point in having selected described phase image, further comprises:
Selected static pixels point is screened.
3. method according to claim 1 is characterized in that, describedly carries out plane fitting according to selected static pixels point, obtains reflecting that the plane of static pixels background phase offset error comprises:
As initial point, carry out plane fitting with selected static pixels point, obtain reflecting the plane of static pixels background phase offset error.
4. method according to claim 3 is characterized in that, adopts least-squares algorithm to carry out plane fitting.
5. method according to claim 1 is characterized in that, described phase image is deducted the plane that institute's match obtains, and the phase image after obtaining proofreading and correct through background phase comprises:
All pixels in the described phase image are deducted corresponding point on the plane that institute's match obtains, obtain phase image through phasing.
6. gauged device of background phase comprises:
A kernel clustering unit (201) is used for utilizing the kernel clustering algorithm to select the static pixels corresponding with static tissue at the phase image of the velocity encoded cine that has obtained;
A plane fitting unit (202), the static pixels that is used to utilize described kernel clustering unit (201) to select is carried out plane fitting, obtains the plane of a reflection background phase offset error;
A correcting unit (203) is used for described phase image is deducted the plane of the reflection background phase offset error that described plane fitting unit (202) obtains, the phase image after obtaining proofreading and correct through background phase, wherein,
The described method of kernel clustering of utilizing selects static pixels corresponding with static tissue in the described phase image to comprise:
With the pixel in the described phase image as pixel samples;
Utilize Mercer nuclear that described pixel samples is mapped to high-dimensional feature space;
In described high-dimensional feature space, described pixel samples is sorted, and select the part sample of eigenvalue minimum;
The part sample of described eigenvalue minimum is shone upon back image space, obtain static pixels.
7. device according to claim 6 is characterized in that, also comprises:
A screening unit (301) be used for the static pixels corresponding with static tissue that described kernel clustering unit (201) is selected screened, and the static pixels after will screening is sent to described plane fitting unit (202).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910180891 CN102038501B (en) | 2009-10-20 | 2009-10-20 | Correcting method and device of background phase |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910180891 CN102038501B (en) | 2009-10-20 | 2009-10-20 | Correcting method and device of background phase |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102038501A CN102038501A (en) | 2011-05-04 |
| CN102038501B true CN102038501B (en) | 2013-07-31 |
Family
ID=43905293
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200910180891 Active CN102038501B (en) | 2009-10-20 | 2009-10-20 | Correcting method and device of background phase |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102038501B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102509270A (en) * | 2011-11-14 | 2012-06-20 | 西安电子科技大学 | Image segmentation method based on mark watershed algorithm and kernel evolution clustering |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101156777A (en) * | 2006-10-06 | 2008-04-09 | Ge医疗系统环球技术有限公司 | Apparatus and method for magnetic resonance imaging |
-
2009
- 2009-10-20 CN CN 200910180891 patent/CN102038501B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101156777A (en) * | 2006-10-06 | 2008-04-09 | Ge医疗系统环球技术有限公司 | Apparatus and method for magnetic resonance imaging |
Non-Patent Citations (1)
| Title |
|---|
| A.Suinesiaputra et al..Optic Flow Computation from Cardiac MR Tagging Using a Multiscale Differential Method A Comparative Study with Velocity-Encoded MRI.《Lecture Notes in Computer Science》.2003,第2878卷483-490. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102038501A (en) | 2011-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sun et al. | Joint CS-MRI reconstruction and segmentation with a unified deep network | |
| Sander et al. | Towards increased trustworthiness of deep learning segmentation methods on cardiac MRI | |
| Paragios | A level set approach for shape-driven segmentation and tracking of the left ventricle | |
| US10915990B2 (en) | Systems and methods for denoising medical images with deep learning network | |
| CA2926958C (en) | Flow analysis in 4d mr image data | |
| Zucker et al. | Free-breathing accelerated cardiac MRI using deep learning: validation in children and young adults | |
| CN109754387A (en) | Medical image lesion detects localization method, device, electronic equipment and storage medium | |
| CN109978872B (en) | White matter microstructure characteristic screening system and method based on white matter fiber tracts | |
| Leng et al. | Imaging 4D morphology and dynamics of mitral annulus in humans using cardiac cine MR feature tracking | |
| Dong et al. | DeU-Net 2.0: Enhanced deformable U-Net for 3D cardiac cine MRI segmentation | |
| Itti et al. | Automatic scan prescription for brain MRI | |
| Ankenbrand et al. | Deep learning‐based cardiac cine segmentation: Transfer learning application to 7T ultrahigh‐field MRI | |
| Mirabella et al. | MRI-based protocol to characterize the relationship between bicuspid aortic valve morphology and hemodynamics | |
| Hameeteman et al. | Carotid wall volume quantification from magnetic resonance images using deformable model fitting and learning-based correction of systematic errors | |
| Stough et al. | Left ventricular and atrial segmentation of 2d echocardiography with convolutional neural networks | |
| Eppenhof et al. | Progressively growing convolutional networks for end-to-end deformable image registration | |
| Gong et al. | Deep learning and multi-contrast-based denoising for low-SNR Arterial Spin Labeling (ASL) MRI | |
| CN102038501B (en) | Correcting method and device of background phase | |
| Morales et al. | Present and future innovations in AI and cardiac MRI | |
| Küstner et al. | Self-supervised motion-corrected image reconstruction network for 4D magnetic resonance imaging of the body trunk | |
| Westenberg et al. | Accurate and reproducible mitral valvular blood flow measurement with three‐directional velocity‐encoded magnetic resonance imaging | |
| Gottwald et al. | Retrospective camera‐based respiratory gating in clinical whole‐heart 4D flow MRI | |
| Ghoul et al. | Attention-aware non-rigid image registration for accelerated MR imaging | |
| CN112101396A (en) | Classification method, classification device, classification equipment and storage medium | |
| Goatman et al. | Computed tomography lung iodine contrast mapping by image registration and subtraction |
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 |