CN102393963A - Deformable surface model-based magnetic-resonance (MR) image skull stripping method - Google Patents
Deformable surface model-based magnetic-resonance (MR) image skull stripping method Download PDFInfo
- Publication number
- CN102393963A CN102393963A CN2011101863743A CN201110186374A CN102393963A CN 102393963 A CN102393963 A CN 102393963A CN 2011101863743 A CN2011101863743 A CN 2011101863743A CN 201110186374 A CN201110186374 A CN 201110186374A CN 102393963 A CN102393963 A CN 102393963A
- Authority
- CN
- China
- Prior art keywords
- model
- summit
- vector
- surface model
- image
- 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
Images
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention relates to a deformable surface model-based magnetic-resonance (MR) image skull stripping method, which belongs to the image processing field, and is characterized in that: an initial surface model is built by utilizing a triangular mesh, and the initial surface model can approach a brain tissue surface through gradual deformation. The deformation process is required to satisfy two restriction conditions: firstly, the continuity and smoothness of the surface are guaranteed to avoid a broken layer and broken lines; and secondly, the model is promoted to move towards the surface of the image brain tissue. An ellipsoidal surface is innovatively adopted to initialize the surface model, firstly an initial skull margin is obtained through threshold segmentation, initial skull marginal points are used for fitting the ellipsoidal surface, then a spherical mesh model is obtained by segmenting a regular icosahedron, and finally mesh points of the spherical model are mapped to corresponding mesh points of the ellipsoidal surface through the affine transformation. Since a skull of the human being is approximately in an elliptical shape, by adopting the surface model, the convergence velocity of the algorithm can be facilitated to be rapider, the iterations is greatly reduced, and the operand is reduced.
Description
Technical field
The invention belongs to technical field of image processing, relate to the skull stripping means of MR image.
Background technology
Mr imaging technique is non-invasive with it, signal to noise ratio (S/N ratio) is high, scanning angle is flexible and the advantage on soft-tissue imaging and be widely used in the brain pathological research, like brain tumor, encephalatrophy, hydrocephalus, brain multiple sclerosis and senile dementia etc.These are in a bad way and are threatening people's health and lives, and will make diagnosis accurately to these diseases, be unable to do without quantitative measurment and analysis to human brain linked groups.Usually need area-of-interest be split from whole cranium brain image accurately.And a condition precedent of accurately cutting apart is exactly from brain MR image, tissues such as scalp, skull, muscle, blood vessel to be rejected, and only keeps brain tissue.This is because on the one hand, when the nervous system disease such as senile dementia were carried out pathological study, whole brain was exactly the object of being studied; On the other hand, when brain tissue further was divided into cerebrospinal fluid, ectocinerea and white matter of brain, it is complicated that the existence of tissues such as scalp, the skull situation that can make becomes, and gives to cut apart and bring a lot of inconvenience.
For being cut apart, follow-up brain tissue more accurately is prone to row; At first need before processing, tissues such as the scalp in the MRI, skull, muscle, blood vessel be rejected; Realize peeling off of skull, mainly contain three kinds of methods at present and realize peeling off of skull: the method for cutting apart manually, based on morphology methods with based on the method for surface model.
The method that manual work is cut apart as its name suggests, is image is cut apart by the people of specialized training.Generally speaking, the segmentation result accuracy is the highest, is commonly used to the standard as other dividing methods of check.But the artificial method of cutting apart needs the manpower of labor, cuts apart one group of cranium brain MR image and need spend 15 minutes to 2 hours usually.When lots of data, the expense of this time obviously is huge.
Based on morphology methods is to consider a prior imformation of brain tissue, that is, in cranium brain MR image, the brain tissue region is maximum connected region, can search out this connected region through morphology methods, and then realizes peeling off of skull.The final purpose of algorithm is to seek connected region maximum in the MR image.Usually adopt the border detection operator to seek the border at brain area place, handle through morphological method then, the discontinuous border that will be interrupted connects into the curve of sealing, and the inner zone of curve is exactly the brain tissue region.But the result of this algorithm depends on the result of border detection, and for the MR image that comprises eyeball, the result is relatively poor for border detection operator gained, is difficult to be partitioned into result accurately.
Based on the method for surface model, but be the surface model that adopts deformation, under the ordering about of external force and internal force, approach a kind of method on brain tissue surface gradually.External force impels model to move to the brain tissue surface, and internal force then is used for keeping the slickness and the continuity of model surface, avoids model surface to rupture.This algorithm need be constructed an initial surface model; The arithmetic speed of the shape of initial surface model and position and algorithm is closely related; General, at a spherical shape initial surface of cranium brain internal structure model, but this initial surface model is different with the shape of brain tissue outside surface; The position is apart from each other also, and the arithmetic speed of algorithm still has much room for improvement.
Summary of the invention
But the object of the invention provides a kind of skull stripping means based on the deformation surface model.As the initial surface model, the deformation through gradually reaches the purpose that skull is peeled off to this method, belongs to a kind of method based on surface model at the grid model that generates a class elliptical area near the brain area of skull; This method is handled to complete 3DMR image, can guarantee the integrality of segmentation result, without manual intervention, has improved computing velocity and robustness.
At first in 3D cranium brain MR figure, with initial surface model of tri patch structure, the initial surface model can approach the surface of brain tissue through deformation gradually in the present invention.Deformation process need satisfy two constraint conditions, and one of which guarantees surperficial continuously smooth, avoids occurring the situation of " tomography " and " broken line "; Its two, impel model suitable position in image to move, i.e. the surface of brain tissue.Describe for ease, these two kinds of constraint conditions are called external force and internal force respectively.External force acts on surface model, makes it to take place deformation, moves to the brain tissue outside surface, and internal force then keeps the continuous and level and smooth of model.See on the microcosmic that the process of deformation is exactly the process that progressively moves on each summit in the triangle grid model, i.e. the process that all under the effect of " power " (motion vector), moves of each summit in the grid.
Technical scheme of the present invention is following:
But a kind of MR image skull stripping means based on the deformation surface model is as shown in Figure 1, may further comprise the steps:
Step 1: set up the initial table surface model.
Step 1-1: organize definite threshold value T between average pixel value and the brain tissue's average pixel value at skull, according to the edge of the tentatively definite skull of threshold value T, the coordinate of record marginal point is organized ellipsoidal surface A of coordinate figure match with this, establishes ellipsoidal surface A equation and does
Estimate ellipsoidal surface sphere centre coordinate (x
0, y
0, z
0), and parameter a, b, c.
Step 1-2: one with step 1-1 in the concentric sphere face B of gained ellipsoidal surface A inner; (regular dodecahedron is the solid that is made up of 20 identical equilateral triangles to construct a regular dodecahedron C; Its summit is positioned on the same sphere face), each triangular surface of regular dodecahedron C is divided into 4 little triangles (as shown in Figure 2), so divide equally 3~4 times after; Find out the coordinate of all little triangular apex, obtain sphere grid model D in sphere face B surface corresponding point; Noting among the sphere grid model D in each tri patch the coordinate on each summit and each tri patch then constitutes; Describe for ease; Use vector
to represent apex coordinate respectively, the formation of vector
expression tri patch.
Step 1-3: affined transformation.
All summits among the sphere grid model D are mapped to corresponding ellipsoidal surface A surface; Obtain ellipsoid grid model E; Note the apex coordinate information of all tri patchs among the ellipsoid grid model E; Be designated as
n=1,2,3...N; N is the number on summit, with ellipsoid grid model E as the deformation surface model that just begins.
Step 2: iteration updating form surface model.
Utilizing iterative formula
that iteration is carried out on all summits in the deformation surface model that just begins upgrades; Wherein, During the k time iteration of
expression, the coordinate information on n summit; During the k time iteration of
expression, the motion vector on n summit;
Step 2-1: calculate associated vector.
(a) unit normal vector
account form of calculating the summit be this summit of traversal related all tri patchs; With superposeing after the normal vector unitization on these tri patchs, the average of obtaining normal vector then gets final product; Wherein the normal vector of tri patch obtains (as shown in Figure 3) through the vector product of calculating any two vectors on this tri patch.
(b) is calculated for each vertex and its two adjacent vertices of any intermediate point difference vector?
and the difference vector?
orthogonal decomposition component along the direction of the vertex normals?
and the tangential component?
(Figure 4) .
Step 2-2: displacement calculating vector.
wherein:
be used for guaranteeing that each summit of model spatially evenly distributes, be equally spaced between the adjacent mesh; Vector
is used for guaranteeing that the model curved surface is level and smooth; Vector
then makes surface model carry out deformation to the surface of brain tissue.
(b) utilize formula
Calculate component
Wherein
F=6/ (1/r
Min k-1/r
Max k), wherein
When being the k time iteration, the radius-of-curvature on n summit, r
Max kWith r
Min kWhen being respectively the k time iteration, the radius-of-curvature of maximum and minimum radius-of-curvature in all net points, tanh is a hyperbolic tangent function.
(c) utilize formula
Calculate component
Wherein,
When representing the k time iteration, n summit is adjacent the mean distance on summit;
When representing the k time iteration, the local minimum gradation value on n summit;
When representing the k time iteration, the local maximum gradation value on n summit, T
MinBe overall minimum gradation value, T
MaxBe overall maximum gradation value.
The effect of displacement component
is; When the summit of model is positioned at the brain tissue zone; When this summit is inwardly searched for; The pixel of process basically all be in homogeneous region;
coefficient is for just so, and the summit of surface model will outwards be moved; When the summit of model is positioned at the brain tissue border; When this summit is inwardly searched for; The bounding edge line is bound to; The local maximum gradation value
that calculated this moment and the difference of local minimum gradation value
become big; Then
coefficient is for negative, and the summit of surface model will move inward.
Step 3: judge convergence, promptly work as iterations less than threshold value T
nThe time, repeating step 2; Otherwise, when iterations more than or equal to threshold value T
nThe time, iteration finishes, and transfers execution in step 4 to;
Step 4: utilize final surface model to reject the non-brain tissue part beyond the surface model in the MR image, the inner brain tissue of retention surface model obtains final MR image skull peel results.
Supplementary notes:
(1) being that segmentation through each tri patch in to regular dodecahedron obtains the sphere grid model in step 1-2, why adopting regular dodecahedron, is because to after the regular dodecahedron segmentation, the gained grid is more even.Vector
is when representing the k time iteration; The formation of m tri patch of i.e.
vector
expression of the volume coordinate on n summit, promptly
just constitutes the index value in vectorial array
on three summits of m tri patch.
(2) in step 1-3, utilize affined transformation, the summit on the gained sphere grid model D among the step 1-2 is mapped to the summit on the corresponding ellipsoidal surface A.Because when ellipsoid and sphere carry out affined transformation; Topological structure does not change, and therefore vector
is constant.
(3) in step 2, the process of surface model deformation just equals the process that each net point in the triangle grid model all moves under the effect of motion vector
.And in the The model deformation process; The topological structure of model is constant, and therefore vector
is also constant in deformation process.
(4) in step 3, algorithm finally levels off to the surface of brain tissue through deformation, and the summit of surface model can be in the vibration of brain tissue edge, but oscillation amplitude is along with the iterations increase diminishes gradually.General, through after 200 iteration, can access satisfied segmentation effect.Therefore, get T
n=200.
The invention has the beneficial effects as follows:
(1) when the initial table surface model, adopted ellipsoidal surface structure grid model, because people's cranium brain is approximately elliptical shape, therefore adopt this surface model can make that algorithm the convergence speed is faster, greatly reduce iterations, reduced operand.
(2) but adopted the surface model of deformation; Utilize displacement component
and
to retrain deformation process; Make that the result of final gained is complete more more level and smooth, do not allow to be subject to the interference of local noise.Whole algorithm need not manual intervention, and degree of accuracy is higher.
Description of drawings
Fig. 1 is a schematic flow sheet of the present invention.
Fig. 2 is intermediate cam dough sheet segmentation synoptic diagram of the present invention.
Fig. 3 calculates synoptic diagram for vertex scheme vector among the present invention.
Fig. 4 calculates and the quadrature decomposing schematic representation for difference vector among the present invention.
Embodiment
Detailed technology scheme such as summary of the invention part are said, repeat no more once more.
Technical scheme of the present invention is at first used Matlab language simulated program when realizing; Use MRI medical image sequence data to carry out parameter setting and program optimization processing then; Use C Plus Plus rewriting program code and interactive interface framework at last, to improve program feature.
Claims (3)
1. but MR image skull stripping means based on the deformation surface model may further comprise the steps:
Step 1: set up the initial table surface model;
Step 1-1: organize definite threshold value T between average pixel value and the brain tissue's average pixel value at skull, according to the edge of the tentatively definite skull of threshold value T, the coordinate of record marginal point is organized ellipsoidal surface A of coordinate figure match with this, establishes ellipsoidal surface A equation and does
Estimate ellipsoidal surface sphere centre coordinate (x
0, y
0, z
0), and parameter a, b, c;
Step 1-2: one with step 1-1 in the concentric sphere face B of gained ellipsoidal surface A inner; Construct a regular dodecahedron C; Each triangular surface of regular dodecahedron C is divided into 4 little triangles; After so dividing equally 3~4 times, find out the coordinate of all little triangular apex, obtain sphere grid model D in sphere face B surface corresponding point; Noting among the sphere grid model D in each tri patch the coordinate on each summit and each tri patch then constitutes; Describe for ease; Use vector
to represent apex coordinate respectively, the formation of vector
expression tri patch;
Step 1-3: affined transformation;
All summits among the sphere grid model D are mapped to corresponding ellipsoidal surface A surface; Obtain ellipsoid grid model E; Note the apex coordinate information of all tri patchs among the ellipsoid grid model E; Be designated as
n=1; 2,3...N, N are the number on summit; With ellipsoid grid model E as the deformation surface model that just begins;
Step 2: iteration updating form surface model;
Utilizing iterative formula
that iteration is carried out on all summits in the deformation surface model that just begins upgrades; Wherein, During the k time iteration of
expression, the coordinate information on n summit; During the k time iteration of
expression, the motion vector on n summit;
Step 2-1: calculate associated vector;
(a) unit normal vector
account form of calculating the summit be this summit of traversal related all tri patchs; With superposeing after the normal vector unitization on these tri patchs, the average of obtaining normal vector then gets final product; Wherein the normal vector of tri patch obtains through the vector product of calculating any two vectors on this tri patch;
(b) is calculated for each vertex and its two adjacent vertices of any intermediate point difference vector
and the difference vector
orthogonal decomposition component along the direction of the vertex normals
and tangential component
wherein:
be used for guaranteeing that each summit of model spatially evenly distributes, be equally spaced between the adjacent mesh; Vector
is used for guaranteeing that the model curved surface is level and smooth; Vector
then makes surface model carry out deformation to the surface of brain tissue;
(b) utilize formula
Calculate component
Wherein
Wherein
When being the k time iteration, the radius-of-curvature on n summit, r
Max kWith r
Min kWhen being respectively the k time iteration, the radius-of-curvature of maximum and minimum radius-of-curvature in all net points, tanh is a hyperbolic tangent function;
(c) utilize formula
Calculate component
Wherein,
When representing the k time iteration, n summit is adjacent the mean distance on summit;
When representing the k time iteration, the local minimum gradation value on n summit;
When representing the k time iteration, the local maximum gradation value on n summit, T
MinBe overall minimum gradation value, T
MaxBe overall maximum gradation value;
Step 3: judge convergence, promptly work as iterations less than threshold value T
nThe time, repeating step 2; Otherwise, when iterations more than or equal to threshold value T
nThe time, iteration finishes, and transfers execution in step 4 to;
Step 4: utilize final surface model to reject the non-brain tissue part beyond the surface model in the MR image, the inner brain tissue of retention surface model obtains final MR image skull peel results.
2. but the MR image skull stripping means based on the deformation surface model according to claim 1 is characterized in that the T of threshold value described in the step 3
nEqual 200.
3. but the MR image skull stripping means based on the deformation surface model according to claim 1; It is characterized in that the concrete grammar of rejecting the non-brain tissue part beyond the surface model in the image in the step 4 is the pixel value zero setting with the pixel beyond the surface model in the MR image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011101863743A CN102393963A (en) | 2011-07-05 | 2011-07-05 | Deformable surface model-based magnetic-resonance (MR) image skull stripping method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011101863743A CN102393963A (en) | 2011-07-05 | 2011-07-05 | Deformable surface model-based magnetic-resonance (MR) image skull stripping method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102393963A true CN102393963A (en) | 2012-03-28 |
Family
ID=45861273
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2011101863743A Pending CN102393963A (en) | 2011-07-05 | 2011-07-05 | Deformable surface model-based magnetic-resonance (MR) image skull stripping method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102393963A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102999754A (en) * | 2012-10-26 | 2013-03-27 | 湖南爱威科技股份有限公司 | Method for extracting multi-fractal texture features of blood cells |
CN105869218A (en) * | 2016-03-28 | 2016-08-17 | 中国科学院深圳先进技术研究院 | Method and apparatus for editing tumor lesion of blood vessel digital model |
CN107437251A (en) * | 2017-07-26 | 2017-12-05 | 广州慧扬健康科技有限公司 | Head mri image skull strip module |
US9977107B2 (en) | 2013-04-03 | 2018-05-22 | Siemens Healthcare Gmbh | Atlas-free brain tissue segmentation method using a single T1-weighted MRI acquisition |
CN111260547A (en) * | 2020-04-29 | 2020-06-09 | 北京智拓视界科技有限责任公司 | Method, apparatus and computer-readable storage medium for presenting brain image |
CN111292348A (en) * | 2020-01-21 | 2020-06-16 | 滨州医学院 | MRA skull stripping method based on autonomous probe and three-dimensional labeling replacement and application thereof |
CN112822981A (en) * | 2018-10-09 | 2021-05-18 | 皇家飞利浦有限公司 | Automatic EEG sensor registration |
CN114066922A (en) * | 2021-11-19 | 2022-02-18 | 数坤(北京)网络科技股份有限公司 | Medical image segmentation method and device, terminal equipment and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050004446A1 (en) * | 2003-06-25 | 2005-01-06 | Brett Cowan | Model assisted planning of medical imaging |
CN101238987A (en) * | 2007-09-06 | 2008-08-13 | 深圳先进技术研究院 | Processing method of CT cerebral hemorrhage image |
-
2011
- 2011-07-05 CN CN2011101863743A patent/CN102393963A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050004446A1 (en) * | 2003-06-25 | 2005-01-06 | Brett Cowan | Model assisted planning of medical imaging |
CN101238987A (en) * | 2007-09-06 | 2008-08-13 | 深圳先进技术研究院 | Processing method of CT cerebral hemorrhage image |
Non-Patent Citations (2)
Title |
---|
STEPHEN M. SMITH: "Fast Robust Automated Brain Extraction", 《HUMAN BRAIN MAPPING》, vol. 17, no. 3, 30 November 2002 (2002-11-30), pages 143 - 155 * |
章志勇,杨柏林: "视觉图像相似性在三维模型相似性比较中的应用", 《计算机辅助设计与图形学学报》, vol. 18, no. 7, 31 July 2006 (2006-07-31), pages 1049 - 1053 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102999754B (en) * | 2012-10-26 | 2015-07-01 | 爱威科技股份有限公司 | Method for extracting multi-fractal texture features of blood cells |
CN102999754A (en) * | 2012-10-26 | 2013-03-27 | 湖南爱威科技股份有限公司 | Method for extracting multi-fractal texture features of blood cells |
US9977107B2 (en) | 2013-04-03 | 2018-05-22 | Siemens Healthcare Gmbh | Atlas-free brain tissue segmentation method using a single T1-weighted MRI acquisition |
CN105869218A (en) * | 2016-03-28 | 2016-08-17 | 中国科学院深圳先进技术研究院 | Method and apparatus for editing tumor lesion of blood vessel digital model |
CN105869218B (en) * | 2016-03-28 | 2018-08-14 | 中国科学院深圳先进技术研究院 | The neoplastic lesion edit methods and device of blood vessel mathematical model |
CN107437251B (en) * | 2017-07-26 | 2020-11-13 | 广州慧扬健康科技有限公司 | Skull stripping module for head MRI (magnetic resonance imaging) image |
CN107437251A (en) * | 2017-07-26 | 2017-12-05 | 广州慧扬健康科技有限公司 | Head mri image skull strip module |
CN112822981A (en) * | 2018-10-09 | 2021-05-18 | 皇家飞利浦有限公司 | Automatic EEG sensor registration |
CN111292348A (en) * | 2020-01-21 | 2020-06-16 | 滨州医学院 | MRA skull stripping method based on autonomous probe and three-dimensional labeling replacement and application thereof |
CN111292348B (en) * | 2020-01-21 | 2023-07-28 | 滨州医学院 | MRA skull peeling method based on autonomous probe and three-dimensional labeling displacement |
CN111260547A (en) * | 2020-04-29 | 2020-06-09 | 北京智拓视界科技有限责任公司 | Method, apparatus and computer-readable storage medium for presenting brain image |
CN114066922A (en) * | 2021-11-19 | 2022-02-18 | 数坤(北京)网络科技股份有限公司 | Medical image segmentation method and device, terminal equipment and storage medium |
CN114066922B (en) * | 2021-11-19 | 2022-06-03 | 数坤(北京)网络科技股份有限公司 | Medical image segmentation method and device, terminal equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102393963A (en) | Deformable surface model-based magnetic-resonance (MR) image skull stripping method | |
CN106204587B (en) | Multiple organ dividing method based on depth convolutional neural networks and region-competitive model | |
CN102800089B (en) | Main carotid artery blood vessel extraction and thickness measuring method based on neck ultrasound images | |
EP3375364A1 (en) | System and method for analyzing blood flow state | |
CN106780518B (en) | A kind of MR image three-dimensional interactive segmentation method of the movable contour model cut based on random walk and figure | |
CN106203432A (en) | A kind of localization method of area-of-interest based on convolutional Neural net significance collection of illustrative plates | |
CN102243759B (en) | Three-dimensional lung vessel image segmentation method based on geometric deformation model | |
CN107909585A (en) | Inner membrance dividing method in a kind of blood vessel of intravascular ultrasound image | |
CN107067393B (en) | A kind of three-dimensional medical image segmentation method based on user's interaction and shape prior knowledge | |
CN104318553B (en) | CT image liver segmentation methods based on adaptive surface deformation model | |
CN102682449A (en) | Automatic segmentation method based on self-adaptive external force level set for magnetic resonance images (MRIs) of soft tissue and realization method thereof | |
CN106408576B (en) | Automatic region of interest segmentation method and system based on three-dimensional ultrasonic image | |
CN103229210A (en) | Image registration apparatus | |
CN106846330A (en) | Human liver's feature modeling and vascular pattern space normalizing method | |
CN108109151B (en) | Method and device for segmenting ventricle of echocardiogram based on deep learning and deformation model | |
CN108492300A (en) | Tubular structure enhances the pulmonary vascular tree dividing method combined with energy function | |
CN107680110A (en) | Inner ear three-dimensional level-set segmentation methods based on statistical shape model | |
CN107507189A (en) | Mouse CT image kidney dividing methods based on random forest and statistical model | |
CN109215040A (en) | A kind of tumor of breast dividing method based on multiple dimensioned weighting study | |
WO2022001026A1 (en) | Systems and methods for determining blood vessel parameters | |
CN114972362A (en) | Medical image automatic segmentation method and system based on RMAU-Net network | |
CN113947662B (en) | Ultrasonic simulation method and system based on medical tomography | |
CN110689080B (en) | Planar atlas construction method of blood vessel structure image | |
CN112950611A (en) | Liver blood vessel segmentation method based on CT image | |
Gamal et al. | Brain tumor segmentation using 3D U-Net with hyperparameter optimization |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120328 |