CN108564590A - A kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images - Google Patents
A kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images Download PDFInfo
- Publication number
- CN108564590A CN108564590A CN201810361167.9A CN201810361167A CN108564590A CN 108564590 A CN108564590 A CN 108564590A CN 201810361167 A CN201810361167 A CN 201810361167A CN 108564590 A CN108564590 A CN 108564590A
- Authority
- CN
- China
- Prior art keywords
- right ventricle
- segmentation
- magnetic resonance
- image
- segmentation result
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20061—Hough transform
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30048—Heart; Cardiac
Abstract
The present invention provides a kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images, acquires the subject a certain number of heart original magnetic resonance images of people using magnetic resonance imaging system, extracts area-of-interest.Atlas is added in the atlas image that the right ventricle of fixed quantity is chosen in original magnetic resonance image, and atlas image obtains expert's manual segmentation result described in expert's manual segmentation.Atlas image converts to obtain right ventricle coarse segmentation result using the B-spline based on normalized mutual information with target image, coarse segmentation result is merged using COLLATE, log-likelihood estimation first is carried out to partial data, then EM algorithm iterative solution is recycled until restraining, and is carried out correcting process and obtained the final segmentation result of right ventricle.The present invention has higher robustness, and can improve accuracy and the accuracy of fusion, is used for Accurate Segmentation heart right ventricle short axis images.
Description
Technical field
The invention belongs to magnetic resonance image process fields, and in particular to a kind of based on cardiac magnetic resonance film short axis images
Right ventricle multichannel chromatogram dividing method.
Background technology
Cardiac magnetic resonance film short axle is imaged in addition to having no ionization radiation injury, multi-faceted imaging, soft tissue contrast
Outside the advantages that high, moreover it is possible to provide spatial and temporal resolution high dynamic film image.Cardiac magnetic resonance short axle film image one side energy
The structure of enough accurate display hearts, on the other hand can be used in cardiac segmentation and quantization heart function parameter.
Right ventricle is because its myocardial wall is thin, obscure boundary, myocardial structural change greatly, it is low with surrounding tissue contrast the features such as,
As a big difficulty of cardiac segmentation, Accurate Segmentation right ventricle is even more a major challenge in global cardiac segmentation field.
Traditional medicine image segmentation algorithm is to the equal Shortcomings of the segmentation of heart right ventricle：The active contour model condition of convergence
Be difficult to determine, the initial profile of level set algorithm it is very sensitive.Traditional multichannel chromatogram algorithm fusion stage uses Weighted Fusion plan more
Slightly or STAPLE convergence strategies are handled, and to brain hippocampus or liver segmentation can be obtained preferably as a result, but to the right side
Good effect cannot be reached when ventricular segmentation.
Invention content
The technical issues of to solve right ventricle segmentation, the purpose of the present invention is to provide one kind being based on cardiac magnetic resonance film
The right ventricle multichannel chromatogram dividing method of short axis images.
The present invention is achieved by the following technical solutions：
A kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images, includes the following steps：
S1, the subject a certain number of heart original magnetic resonance images of people are acquired using magnetic resonance imaging system, to the original
Beginning magnetic resonance image is pre-processed, and area-of-interest is extracted；
S2, atlas is added in the atlas image of the right ventricle of selection fixed quantity in the original magnetic resonance image, specially
Atlas image obtains expert's manual segmentation result described in family's manual segmentation；
S3, the atlas image is registrated with target image, the registration uses the B samples based on normalized mutual information
Item converts, and transformation parameter is applied in the tag image corresponding with the atlas image, obtains the collection of illustrative plates after stating registration
Image obtains right ventricle coarse segmentation result；
S4, the coarse segmentation result is merged using COLLATE, log-likelihood estimation first is carried out to partial data, then
EM algorithm iterative solution is recycled until restraining, and carries out correcting process and obtains the final segmentation result of right ventricle；
S5, step S1~S4 is repeated, calculates right ventricular volume and different physical signs, choose relevant parameter as objective
Evaluation index is evaluated and is analyzed to experimental result.
Further, the extraction area-of-interest specific method is：Left ventricle centre bit is determined using hough-circle transform
It sets, right ventricle position is determined by the position relationship in short axis images of left and right ventricles and intercepts while including left and right ventricles
Parts of images is as area-of-interest.
Further, the log-likelihood estimation uses formula：
θ=argmaxθLnf (D, T, C | θ)
Wherein, θ indicates that the corresponding performance level of the coarse segmentation result is estimated, and f (D, T, C | θ) indicate that right ventricle completely counts
According to the probability mass distribution function of corresponding random vector, D is by the atlas image after the coarse segmentation result and the registration
The matrix that constitutes of pixel, T indicates the unknown true segmentation of the pixel as a result, C indicates the consistency of each pixel
Or confusion.
Further, the EM algorithm iterative solution uses formula：
Q(θ|θ(0)) ≡ E [ln f (D, T, C | θ) | D, θ(0)=∑TF (D, T, C | θ) f (D, T, C | θ(0))
Wherein, Q (θ | θ(0)) indicate θ=θ(0)The conditional expectation of Shi Suoshu partial data log-likelihood functions, θ (0) are indicated
The initial value of the corresponding performance level estimation of the coarse segmentation result.
Further, the correcting process method is：To right ventricle top easily occur 1~2 layer of segmentation errors into
Row modified result when middle part image result centre coordinate is in top layer initial segmentation result, carries out shape constraining, effectively prevent letting out
Dew；When point coordinates is not in top layer images initial segmentation result in the image result of middle part, manual segmentation is carried out.
Further, the physical signs include end-diastolic volume EDV, end-systolic volume ESV, Ejection and
Stroke output SV.
Further, the relevant parameter includes accuracy, correlation and consistency.
Further, the accuracy index includes Hausdorff distance HD, using formula：
HD (A, B)=max (maxa∈A(minb∈BD (a, b)), maxb∈B(mina∈AD (a, b)))
Wherein, A indicates expert's manual segmentation as a result, B indicates that the final segmentation result, d (a, b) are Euclid
Distance, wherein a are the points in the atlas image, and b is the point in the target image, and HD (A, B) indicates the atlas image
With the asymmetric difference of maximum of the target image, the institute for reflecting expert's manual segmentation result described in right ventricle and the present invention
State the distance difference of final segmentation result.
Further, the accuracy index includes matrix similarity DM, using formula：
Wherein, AaIndicate the volume size of the final segmentation result, AmIndicate the volume of expert's manual segmentation result
Size, A∩=Aa∩AmIndicate the volume size of overlapping region, DM (Aa, Am) indicate the final segmentation result and the expert
The registration of manual segmentation result.
Compared with prior art, the present invention has the advantages that：
The present invention provides a kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images, using more
COLLATE blending algorithms merge a series of right ventricle coarse segmentation results under collection of illustrative plates frame.The present invention is in STAPLE algorithms
On the basis of be added to pixel consistency probability in right ventricle image, there is higher robustness compared with STAPLE algorithms, and can
Accuracy and the accuracy for improving fusion, can be with Accurate Segmentation heart right ventricle short axis images.
Description of the drawings
Fig. 1 is right ventricle diastasis short axle bottom image segmentation result figure；
Fig. 2 is image segmentation result figure in the middle part of right ventricle diastasis short axle；
Fig. 3 is right ventricle diastasis short axle top image segmentation result figure；
Fig. 4 is right ventricular end systolic short axle bottom image segmentation result figure；
Fig. 5 is image segmentation result figure in the middle part of right ventricular end systolic short axle；
Fig. 6 is right ventricular end systolic short axle top image segmentation result figure；
Wherein, a indicates the area-of-interest in original image, including complete Ventricular；B indicates that expert divides manually
It cuts as a result, i.e. goldstandard；C indicates the segmentation result that the method for the present invention obtains；D indicates segmentation result of the present invention in area-of-interest
In label；
Fig. 7 is the distance difference comparison diagram of the final segmentation result of right ventricle and expert's manual segmentation result；
Fig. 8 is the similarity comparison figure of the final segmentation result of right ventricle and expert's manual segmentation result.
Specific implementation mode
Elaborate below to the embodiment of the present invention, the present embodiment with the technical scheme is that according to development,
Give detailed embodiment and specific operating process.
Several cardiac magnetic resonance short axle film images of the different phase different parts of the embodiment of the present invention carry out right ventricle
Segmentation, obtains the specific implementation process of final segmentation result.Wherein, it is used for the cardiac magnetic resonance film short axle figure of right ventricle segmentation
Picture data source is obtained in magnetic resonance system through SSFP sequence.In experimental data, male 7, women 3, the age covers 14 and arrives
75 years old.Specific imaging parameters：Image size 256 × 256, thickness 6-8mm, interlamellar spacing 2-4mm, every group of data include 6-10 layers,
Every layer of 20-28 phase, including multiple cardiac cycles.
The embodiment of the present invention includes following steps：
S1, using magnetic resonance imaging system acquire subject a certain number of heart original magnetic resonance images of people, to image into
Row pretreatment.Left ventricle center is determined first with hough-circle transform, then by left and right ventricles in short axis images
Position relationship determine right ventricle position and intercept simultaneously include left and right ventricles parts of images as area-of-interest.
S2, selecting structure is clear in the original magnetic resonance image, and smaller right ventricle short axle is influenced by surrounding tissue
Atlas is added in image, and expert's manual segmentation atlas image obtains right ventricle segmentation result, and generates corresponding tag image L1-
Ln。
There is the corresponding tag image divided, registration process to use mutual based on normalization for S3, every atlas image
The B-spline of information converts, and after atlas image is registrated to target image respectively, records corresponding transformation parameter q1-qn, will convert
Parameter is applied in tag image corresponding with atlas image, obtains right ventricle coarse segmentation result L '1-L′n。
S4, the coarse segmentation result is merged using COLLATE, log-likelihood estimation first is carried out to partial data, then
EM algorithm iterative solution is recycled until restraining, and carries out correcting process and obtains the final segmentation result of right ventricle.
Defining has N number of pixel in the atlas image after right ventricle registration, corresponding each pixel has R coarse segmentation knot
Fruit, pixel indicate that coarse segmentation result is indicated with j with i.A series of label L indicate that certain coarse segmentation result will be a series of possible
Standard distributes to all pixel N.D is the matrix of a N*R, is used for description of the R coarse segmentation result to all pixels N,
Middle Dij∈ 0,1 ... L-1 }.Length is the vector T of N, represents the unknown true segmentation of all pixels N as a result, wherein Ti∈
0,1 ... L-1 }.
On the basis of STAPLE algorithms, a vectorial C with N number of element is defined, indicates each in right ventricle image
The consistency or confusion of pixel share F consistent possibility grades.All elements C in the vectori∈ 0,1 ... F-1 }.
Wherein Ci=0 indicates that pixel i has confusion, Ci=1 indicates that pixel i is with uniformity.Consistency and confusion are from opposite
Direction describes the same phenomenon, and pixel i consistency is reduced, and confusion is increased by.
Define the characteristic that θ indicates R right ventricle coarse segmentation result, each pixel θjIt is the hybrid matrix of a L*L, square
A certain pixel is the possibility probability of true segmentation result in each Quantification of elements right ventricle coarse segmentation j in battle array.As ginseng
It examines, " goldstandard " is a unit matrix.Right ventricle partial data is defined as (D, T, C), the Probability Group distribution of partial data
It is defined as f (D, T, C | θ).
Formula is expressed as to the log-likelihood estimation of partial data：
θ=argmaxθLn f (D, T, C | θ)
Wherein, θ=[θ1, θ2..., θR] indicate performance level estimation corresponding with R right ventricle coarse segmentation result；F (D,
T, C | θ) indicate the probability mass distribution function of the corresponding random vector of right ventricle partial data.
The conditional expectation for the log-likelihood function that E steps calculate right ventricle partial data is expressed as formula：
Q(θ|θ(0)) ≡ E [ln f (D, T, C | θ) | D, θ(0)]=ΣTF (D, T, C | θ) f (D, T, C | θ(0))
Wherein, Q (θ | θ(0)) indicate θ=θ(0)When right ventricle partial data log-likelihood function conditional expectation.θ(0)It indicates
The initial value of the corresponding performance level estimation of the initial value of parameter θ, i.e. right ventricle coarse segmentation result.
M step require in the parameter space of θ solve Q (θ | θ(0)) maximum value, i.e., for all θ(0),
There are θ(1)Meet formula：
Q(θ|θ(1))≥Q(θ|θ(0))
Wherein, Q (θ | θ(1)) and Q (θ | θ(0)) θ=θ is indicated respectively(1)With θ=θ(0)When right ventricle partial data log-likelihood
The conditional expectation of function.θ(1)Estimated value be by Q (θ | θ(0)) maximization procedure determine.
COLLATE convergence strategies of the present invention mainly utilize EM algorithm, are divided into E steps and M steps.Integer has been calculated in E steps
According to log-likelihood function conditional expectation, M walk solved function maximize when θ value.Each iterative process is walked by E steps and M
It is alternately repeated progress, until convergence.Iterative convergent process control is at 20 times or so, and syncretizing effect is preferably and the used time is shorter,
Experimentally determined cut-off parameter ε=1*10-7.Right ventricle coarse segmentation result is merged to obtain by COLLATE blending algorithms
The final segmentation result of right ventricle.
For complicated and be not easy the right ventricle top 1-2 tomographic images divided, need to carry out modified result.With its phase
Whether adjacent middle part image result is divided into two kinds of situations with template center's point coordinates as template in top layer initial segmentation result
It is handled：When middle part image result centre coordinate is in top layer initial segmentation result, shape constraining is carried out, effectively prevent letting out
Dew；When point coordinates is not in top layer images initial segmentation result in the image result of middle part, manual segmentation is carried out, due to context
Area is small, and the number of plies is few, and manual segmentation influences processing time while improving precision little.
S5, above step S1~S4 is repeated, obtains the segmentation of structure at all levels scan image in a certain phase of right ventricle first
As a result, calculating every layer of ventricle area by Area length method, then every layer of ventricle area and the product of thickness are calculated, by cumulative
Volume is obtained, such as end-diastolic volume EDV and end-systolic volume ESV.Volume by calculating all phases can draw the right side
Ventricular filling curve.By the segmentation to right ventricle diastasis and end systole image, stroke output can be obtained and penetrated
The physiologic informations such as blood fraction.Pass through objective appraisal index：Accuracy, correlation, consistency etc. evaluate experimental result
And analysis, accuracy index is calculated：Hausdorff distance and matrix similarity.
If Fig. 7 is the final segmentation result and expert's manual segmentation knot under two state of right ventricle diastasis and end-systole
The distance difference comparison diagram of fruit.Horizontal axis represents time point, and the longitudinal axis represents the person of outstanding talent of final segmentation result and expert's manual segmentation result
Si Duofu distances HD.
If Fig. 8 is the final segmentation result and expert's manual segmentation knot under two state of right ventricle diastasis and end-systole
The similarity comparison figure of fruit.Horizontal axis represents time point, and it is similar to expert's manual segmentation result that the longitudinal axis represents final segmentation result
Degree.DM(Aa, Am) indicate that the similarity of the final segmentation result of right ventricle and expert's manual segmentation result, variation range are (complete from 0
Mismatch) to 1 (exactly matching), Duplication is higher, as a result more accurate.
Above example be the application preferred embodiment, those skilled in the art can also on this basis into
The various transformation of row or improvement these transformation or improve under the premise of not departing from the application total design and should all belong to this Shen
Within the scope of please being claimed.
Claims (9)
1. a kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images, includes the following steps：
S1, the subject a certain number of heart original magnetic resonance images of people are acquired using magnetic resonance imaging system, to the original magnetic
Resonance image is pre-processed, and area-of-interest is extracted；
S2, atlas, expert's hand is added in the atlas image of the right ventricle of selection fixed quantity in the original magnetic resonance image
The dynamic segmentation atlas image obtains expert's manual segmentation result；
S3, the atlas image is registrated with target image, the registration is become using the B-spline based on normalized mutual information
It changes, transformation parameter is applied in the tag image corresponding with the atlas image, obtain the atlas image after stating registration,
Obtain right ventricle coarse segmentation result；
S4, the coarse segmentation result is merged using COLLATE, log-likelihood estimation first is carried out to partial data, it is then sharp again
It is iteratively solved with EM algorithm until restraining, and carry out correcting process and obtain the final segmentation result of right ventricle；
S5, step S1~S4 is repeated, calculates right ventricular volume and different physical signs, choose relevant parameter as objective appraisal
Index is evaluated and is analyzed to experimental result.
2. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 1
Method, which is characterized in that the extraction area-of-interest specific method is：Left ventricle center is determined using hough-circle transform,
The portion for determining right ventricle position by the position relationship in short axis images of left and right ventricles and intercepting while including left and right ventricles
Partial image is as area-of-interest.
3. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 1
Method, which is characterized in that the log-likelihood estimation uses formula：
θ=argmaxθLn f (D, T, C | θ)
Wherein, θ indicates that the corresponding performance level of the coarse segmentation result is estimated, and f (D, T, C | θ) indicate right ventricle partial data pair
The probability mass distribution function for the random vector answered, D are by the picture of the atlas image after the coarse segmentation result and the registration
The matrix that element is constituted, T indicates the unknown true segmentation of the pixel as a result, C indicates the consistency or mixed of each pixel
Unrest.
4. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 3
Method, which is characterized in that the EM algorithm iterative solution uses formula：
Q(θ|θ(0)) ≡ E [ln f (D, T, C | θ) | D, θ(0)]=∑TF (D, T, C | θ) f (D, T, C | θ(0))
Wherein, Q (θ | θ(0)) indicate θ=θ(0)The conditional expectation of Shi Suoshu partial data log-likelihood functions, θ(0)Indicate described thick
The initial value of the corresponding performance level estimation of segmentation result.
5. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 1
Method, which is characterized in that in step S4, the correcting process method is：Easily occur the 1~2 of segmentation errors to right ventricle top
Layer carries out modified result, when middle part image result centre coordinate is in top layer initial segmentation result, carries out shape constraining, effectively anti-
Stopping leak reveals；When point coordinates is not in top layer images initial segmentation result in the image result of middle part, manual segmentation is carried out.
6. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 1
Method, which is characterized in that in step S5, the physical signs includes end-diastolic volume EDV, end-systolic volume ESV, penetrates blood system
Number EF and stroke output SV.
7. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 1
Method, which is characterized in that the relevant parameter includes accuracy, correlation and consistency.
8. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 7
Method, which is characterized in that the accuracy index includes Hausdorff distance HD, using formula：
HD (A, B)=max (maxa∈A(minb∈BD (a, b)), maxb∈B(mina∈AD (a, b)))
Wherein, A indicates expert's manual segmentation as a result, B indicates the final segmentation result, d (a, b) be Euclid away from
The point in the atlas image from, wherein a, b is the point in the target image, HD (A, B) indicate the atlas image and
The asymmetric difference of maximum of the target image, for reflecting final segmentation result described in right ventricle and expert's manual segmentation
As a result distance difference.
9. a kind of right ventricle multichannel chromatogram segmentation side based on cardiac magnetic resonance film short axis images according to claim 7
Method, which is characterized in that the accuracy index includes matrix similarity DM, using formula：
Wherein, AaIndicate the volume size of the final segmentation result, AmIndicate that the volume of expert's manual segmentation result is big
It is small, A∩=Aa∩AmIndicate the volume size of overlapping region, DM (Aa, Am) indicate the final segmentation result and expert's hand
The registration of dynamic segmentation result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810361167.9A CN108564590A (en) | 2018-04-20 | 2018-04-20 | A kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810361167.9A CN108564590A (en) | 2018-04-20 | 2018-04-20 | A kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108564590A true CN108564590A (en) | 2018-09-21 |
Family
ID=63535999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810361167.9A Pending CN108564590A (en) | 2018-04-20 | 2018-04-20 | A kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108564590A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110120051A (en) * | 2019-05-10 | 2019-08-13 | 上海理工大学 | A kind of right ventricle automatic division method based on deep learning |
| CN110288581A (en) * | 2019-06-26 | 2019-09-27 | 电子科技大学 | A kind of dividing method based on holding shape convexity Level Set Models |
| CN110363772A (en) * | 2019-08-22 | 2019-10-22 | 西南大学 | Cardiac MRI dividing method and system based on confrontation network |
| CN110503655A (en) * | 2019-08-21 | 2019-11-26 | 山东师范大学 | A kind of MRI glioma dividing method and system based on sparse Bayesian model and multichannel chromatogram fusion |
| CN111553895A (en) * | 2020-04-24 | 2020-08-18 | 中国人民解放军陆军军医大学第二附属医院 | Magnetic resonance left atrium segmentation method based on multi-scale fine granularity |
| CN111815597A (en) * | 2020-07-01 | 2020-10-23 | 杭州脉流科技有限公司 | Left ventricle long and short axis tangent plane extraction method and device based on CT image, computer equipment and storage medium |
| CN111882543A (en) * | 2020-07-29 | 2020-11-03 | 南通大学 | Cigarette filter stick counting method based on AA R2Unet and HMM |
| CN113379760A (en) * | 2021-05-20 | 2021-09-10 | 电子科技大学 | Right ventricle image segmentation method |
| CN114119645A (en) * | 2021-11-25 | 2022-03-01 | 推想医疗科技股份有限公司 | Method, system, device and medium for determining image segmentation quality |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102646268A (en) * | 2012-02-22 | 2012-08-22 | 中国科学院自动化研究所 | Magnetic resonance image brain structure automatic dividing method based on statistics multi-map registration optimization |
| CN103700104A (en) * | 2013-11-12 | 2014-04-02 | 中国科学院自动化研究所 | Method for automatically segmenting human lateral geniculate nucleus (LGN) by using prior knowledge |
| WO2015109331A1 (en) * | 2014-01-17 | 2015-07-23 | The Johns Hopkins University | Automated anatomical labeling by multi-contrast diffeomorphic probability fusion |
| CN105009171A (en) * | 2013-03-01 | 2015-10-28 | 因派克医药系统有限公司 | Method and apparatus for learning-enhanced altas-based auto-segmentation |
-
2018
- 2018-04-20 CN CN201810361167.9A patent/CN108564590A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102646268A (en) * | 2012-02-22 | 2012-08-22 | 中国科学院自动化研究所 | Magnetic resonance image brain structure automatic dividing method based on statistics multi-map registration optimization |
| CN105009171A (en) * | 2013-03-01 | 2015-10-28 | 因派克医药系统有限公司 | Method and apparatus for learning-enhanced altas-based auto-segmentation |
| CN103700104A (en) * | 2013-11-12 | 2014-04-02 | 中国科学院自动化研究所 | Method for automatically segmenting human lateral geniculate nucleus (LGN) by using prior knowledge |
| WO2015109331A1 (en) * | 2014-01-17 | 2015-07-23 | The Johns Hopkins University | Automated anatomical labeling by multi-contrast diffeomorphic probability fusion |
Non-Patent Citations (3)
| Title |
|---|
| CAROLINE PETITJEAN,ET AL: "Right Ventricle Segmentation From Cardiac MRI: A Collation Study", 《MEDICAL IMAGE ANALYSIS》 * |
| 王丽嘉: "磁共振图像全自动分割量化方法研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 陈雯艳: "基于ROI多图谱配准的海马磁共振图像分割", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110120051A (en) * | 2019-05-10 | 2019-08-13 | 上海理工大学 | A kind of right ventricle automatic division method based on deep learning |
| CN110288581A (en) * | 2019-06-26 | 2019-09-27 | 电子科技大学 | A kind of dividing method based on holding shape convexity Level Set Models |
| CN110288581B (en) * | 2019-06-26 | 2022-11-04 | 电子科技大学 | Segmentation method based on model for keeping shape convexity level set |
| CN110503655B (en) * | 2019-08-21 | 2022-04-01 | 山东师范大学 | MRI glioma segmentation method and system based on sparse Bayesian model and multi-map fusion |
| CN110503655A (en) * | 2019-08-21 | 2019-11-26 | 山东师范大学 | A kind of MRI glioma dividing method and system based on sparse Bayesian model and multichannel chromatogram fusion |
| CN110363772A (en) * | 2019-08-22 | 2019-10-22 | 西南大学 | Cardiac MRI dividing method and system based on confrontation network |
| CN110363772B (en) * | 2019-08-22 | 2021-08-13 | 西南大学 | Heart MRI segmentation method and system based on antagonistic network |
| CN111553895A (en) * | 2020-04-24 | 2020-08-18 | 中国人民解放军陆军军医大学第二附属医院 | Magnetic resonance left atrium segmentation method based on multi-scale fine granularity |
| CN111815597A (en) * | 2020-07-01 | 2020-10-23 | 杭州脉流科技有限公司 | Left ventricle long and short axis tangent plane extraction method and device based on CT image, computer equipment and storage medium |
| CN111815597B (en) * | 2020-07-01 | 2024-04-12 | 杭州脉流科技有限公司 | Left ventricle long-short axis section extraction method and device based on CT image, computer equipment and storage medium |
| CN111882543A (en) * | 2020-07-29 | 2020-11-03 | 南通大学 | Cigarette filter stick counting method based on AA R2Unet and HMM |
| CN111882543B (en) * | 2020-07-29 | 2023-12-26 | 南通大学 | Cigarette filter stick counting method based on AA R2Unet and HMM |
| CN113379760A (en) * | 2021-05-20 | 2021-09-10 | 电子科技大学 | Right ventricle image segmentation method |
| CN114119645B (en) * | 2021-11-25 | 2022-10-21 | 推想医疗科技股份有限公司 | Method, system, device and medium for determining image segmentation quality |
| CN114119645A (en) * | 2021-11-25 | 2022-03-01 | 推想医疗科技股份有限公司 | Method, system, device and medium for determining image segmentation quality |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108564590A (en) | A kind of right ventricle multichannel chromatogram dividing method based on cardiac magnetic resonance film short axis images | |
| Ngo et al. | Combining deep learning and level set for the automated segmentation of the left ventricle of the heart from cardiac cine magnetic resonance | |
| Peng et al. | A review of heart chamber segmentation for structural and functional analysis using cardiac magnetic resonance imaging | |
| US10521902B2 (en) | Automated segmentation of organ chambers using deep learning methods from medical imaging | |
| US6961454B2 (en) | System and method for segmenting the left ventricle in a cardiac MR image | |
| US7400757B2 (en) | System and method for segmenting the left ventricle in a cardiac image | |
| Feng et al. | Image segmentation and bias correction using local inhomogeneous iNtensity clustering (LINC): A region-based level set method | |
| Ravishankar et al. | Hybrid approach for automatic segmentation of fetal abdomen from ultrasound images using deep learning | |
| Mahapatra | Cardiac image segmentation from cine cardiac MRI using graph cuts and shape priors | |
| Hu et al. | Automatic segmentation of left ventricle from cardiac MRI via deep learning and region constrained dynamic programming | |
| Zhou et al. | Shape regression machine | |
| US20140247977A1 (en) | Method and Apparatus for Learning-Enhanced Atlas-Based Auto-Segmentation | |
| Stern et al. | Fully automatic bone age estimation from left hand MR images | |
| Khalvati et al. | Automated segmentation of breast in 3-D MR images using a robust atlas | |
| Zhang et al. | A multi-level convolutional LSTM model for the segmentation of left ventricle myocardium in infarcted porcine cine MR images | |
| Cao et al. | Segmenting images by combining selected atlases on manifold | |
| Lu et al. | Group sparse reconstruction for image segmentation | |
| CN106910194B (en) | Method for segmenting endocardium and epicardium in cardiac functional magnetic resonance image | |
| US20220301224A1 (en) | Systems and methods for image segmentation | |
| Savaashe et al. | A review on cardiac image segmentation | |
| Luo et al. | Cardiac left ventricular volumes prediction method based on atlas location and deep learning | |
| Jiang et al. | One shot PACS: Patient specific Anatomic Context and Shape prior aware recurrent registration-segmentation of longitudinal thoracic cone beam CTs | |
| WO2018201437A1 (en) | Image segmentation method and system | |
| US20130279764A1 (en) | Unique Cardiovascular Measurements for Human Identification | |
| Raina et al. | Exploiting bilateral symmetry in brain lesion segmentation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180921 |
|
| RJ01 | Rejection of invention patent application after publication |