NAMIC Wiki - User contributions [en]

File:GT-SPD-img1.png

2009-09-28T21:53:10Z

Slankton: Sample detected plaque shown on a on a segmented vessel

Sample detected plaque shown on a on a segmented vessel

Algorithm:GATech

2009-09-28T21:52:28Z

Slankton: /* Georgia Tech Projects */

Back to [[Algorithm:Main|NA-MIC Algorithms]]
__NOTOC__
= Overview of Georgia Tech Algorithms (PI: Allen Tannenbaum) =

At Georgia Tech, we are broadly interested in a range of mathematical image analysis algorithms for segmentation, registration, diffusion-weighted MRI analysis, and statistical analysis. For many applications, we cast the problem in an energy minimization framework wherein we define a partial differential equation whose numeric solution corresponds to the desired algorithmic outcome. The following are many examples of PDE techniques applied to medical image analysis.

= Georgia Tech Projects =

{| cellpadding="10" style="text-align:left;"

| style="width:15%" | [[Image:GT-SPD-img1.png|200px|]]
| style="width:85%" |

== [[Projects:SoftPlaqueDetection|Soft Plaque Detection in CTA Imagery]] ==

The ability to detect and measure non-calciﬁed plaques (also known as soft plaques) may improve physicians’ ability to predict cardiac events. This work automatically detects soft plaques in CTA imagery using active contours driven by spatially localized probabilistic models. Plaques are identified by simultaneously segmenting the vessel from the inside-out and the outside-in using carefully chosen localized energies [[Projects:SoftPlaqueDetection|More...]]

'''New: ''' Soft Plaque Detection and Automatic Vessel Segmentation. PMMIA Workshop in MICCAI, Sep. 2009.

|-

| | [[Image:GT-DWI-Reorientation-1.jpg|200px]]
| |

== [[Projects:DWIReorientation|Re-Orientation Approach for Segmentation of DW-MRI]] ==

This work proposes a methodology to segment tubular fiber bundles from diffusion weighted magnetic resonance images (DW-MRI). Segmentation is simplified by locally reorienting diffusion information based on large-scale fiber bundle geometry. [[Projects:DWIReorientation|More...]]

'''New: ''' Near-tubular fiber bundle segmentation for diffusion weighted imaging: segmentation through frame reorientation. Neuroimage, Mar 2009.

|-

| | [[Image:GTTubSurfaceSeg-Img1.png|200px]]
| |

== [[Projects:TubularSurfaceSegmentation|Tubular Surface Segmentation Framework]] ==

We have proposed a new model for tubular surfaces that transforms the problem of detecting a surface in 3D space, to detecting a curve in 4D space. Besides allowing us to impose a "soft" tubular shape prior, this also leads to computational efficiency over conventional surface segmentation approaches. [[Projects:TubularSurfaceSegmentation|More...]]

'''New: ''' V. Mohan, G. Sundaramoorthi and A. Tannenbaum. Tubular Surface Segmentation for identifying anatomical structures from medical imagery. September 2008. Proceedings of the Second Workshop on Mathematical Foundations of Computational Anatomy (MFCA'08), Int Conf Med Image Comput Comput Assist Interv. 2008.

|-

| | [[Image:Circle seg.PNG|200px|]]
| |

== [[Projects:KPCASegmentation|Kernel PCA for Segmentation]] ==

Segmentation performances using active contours can be drastically improved if the possible shapes of the object of interest are learnt. The goal of this work is to use Kernel PCA to learn shape priors. Kernel PCA allows for learning non linear dependencies in data sets, leading to more robust shape priors. [[Projects:KPCASegmentation|More...]]

'''New: ''' S. Dambreville, Y. Rathi, and A. Tannenbaum. A Framework for Image Segmentation using Image Shape Models and Kernel PCA Shape Priors. IEEE Trans Pattern Anal Mach Intell. 2008 Aug;30(8):1385-99

|-
| | [[Image:ZoomedResultWithModel.png|200px]]
| |

== [[Projects:GeodesicTractographySegmentation|Geodesic Tractography Segmentation]] ==

In this work, we provide an energy minimization framework which allows one to find fiber tracts and volumetric fiber bundles in brain diffusion-weighted MRI (DW-MRI). [[Projects:GeodesicTractographySegmentation|More...]]

'''New: ''' J. Melonakos, E. Pichon, S. Angenet, and A. Tannenbaum. Finsler Active Contours. IEEE Transactions on Pattern Analysis and Machine Intelligence, March 2008, Vol 30, Num 3.

|-

| | [[Image:P1_small.png|200px|]]
| |

== [[Projects:LabelSpace|Label Space: A Coupled Multi-Shape Representation]] ==

Many techniques for multi-shape representation may often develop inaccuracies stemming from either approximations or inherent variation. Label space is an implicit representation that offers unbiased algebraic manipulation and natural expression of label uncertainty. We demonstrate smoothing and registration on multi-label brain MRI. [[Projects:LabelSpace|More...]]

'''New: ''' J. Malcolm, Y. Rathi, A. Tannenbaum. "Label Space: A Multi-Object Shape Representation." In Combinatorial Image Analysis, 2008.

|-

| | [[Image:BasePair3DModel.JPG|200px|]]
| |

== [[Projects:NonParametricClustering|Non Parametric Clustering for Biomolecular Structural Analysis]] ==

High accuracy imaging and image processing techniques allow for collecting structural information of biomolecules with atomistic accuracy. Direct interpretation of the dynamics and the functionality of these structures with physical models, is yet to be developed. Clustering of molecular conformations into classes seems to be the first stage in recovering the formation and the functionality of these molecules. [[Projects:NonParametricClustering|More...]]

'''New: ''' E. Hershkovits, A. Tannenbaum, and R. Tannenbaum. Adsorption of Block Copolymers from Selective Solvents on Curved Surfaces. Macromolecules. 2008.

|-

| | [[Image:TruckInitialization.png|200px|]]
| |

== [[Projects:PointSetRigidRegistration|Point Set Rigid Registration]] ==

In this work, we propose a particle filtering approach for the problem of registering two point sets that differ by
a rigid body transformation. Experimental results are provided that demonstrate the robustness of the algorithm to initialization, noise, missing structures or differing point densities in each sets, on challenging 2D and 3D registration tasks. [[Projects:PointSetRigidRegistration|More...]]

'''New: ''' R. Sandhu, S. Dambreville, A. Tannenbaum. Particle Filtering for Registration of 2D and 3D Point Sets with Stochastic Dynamics. In CVPR, 2008.

|-

| | [[Image:ShapeBasePstSegSlicer.png|200px|]]
| |

== [[Projects:ProstateSegmentation|Prostate Segmentation]] ==

The 3D prostate MRI images are collected by collaborators at Queen’s University. With a little manual initialization, the algorithm provided the results give to the left. The method mainly uses Random Walk algorithm. [[Projects:ProstateSegmentation|More...]]

|-

| | [[Image:ProstateRegSupineToProneInParaview.png|200px|]]
| |

== [[Projects:pfPtSetImgReg|Particle Filter Registration of Medical Imagery]] ==

3D volumetric image registration is performed. The method is based on registering the images through point sets, which is able to handle long distance between as well as registration between Supine and Prone pose prostate. [[Projects:pfPtSetImgReg|More...]]

|-

| | [[Image:Results brain sag.JPG|200px]]
| |

== [[Projects:OptimalMassTransportRegistration|Optimal Mass Transport Registration]] ==

The goal of this project is to implement a computationaly efficient Elastic/Non-rigid Registration algorithm based on the Monge-Kantorovich theory of optimal mass transport for 3D Medical Imagery. Our technique is based on Multigrid and Multiresolution techniques. This method is particularly useful because it is parameter free and utilizes all of the grayscale data in the image pairs in a symmetric fashion and no landmarks need to be specified for correspondence. [[Projects:OptimalMassTransportRegistration|More...]]

|-

| | [[Image:Gatech caudateBands.PNG|200px]]
| |

== [[Projects:MultiscaleShapeSegmentation|Multiscale Shape Segmentation Techniques]] ==

To represent multiscale variations in a shape population in order to drive the segmentation of deep brain structures, such as the caudate nucleus or the hippocampus. [[Projects:MultiscaleShapeSegmentation|More...]]

|-

| | [[Image:Caudate Nucleus Denoising.JPG|200px|]]
| |

== [[Projects:WaveletShrinkage|Wavelet Shrinkage for Shape Analysis]] ==

Shape analysis has become a topic of interest in medical imaging since local variations of a shape could carry relevant information about a disease that may affect only a portion of an organ. We developed a novel wavelet-based denoising and compression statistical model for 3D shapes. [[Projects:WaveletShrinkage|More...]]

|-

| | [[Image:Basis membership.png|200px]]
| |

== [[Projects:MultiscaleShapeAnalysis|Multiscale Shape Analysis]] ==

We present a novel method of statistical surface-based morphometry based on the use of non-parametric permutation tests and a spherical wavelet (SWC) shape representation. [[Projects:MultiscaleShapeAnalysis|More...]]

|-

| | [[Image:Dlpfc1.jpg|200px|]]
| |

== [[Projects:RuleBasedDLPFCSegmentation|Rule-Based DLPFC Segmentation]] ==

In this work, we provide software to semi-automate the implementation of segmentation procedures based on expert neuroanatomist rules for the dorsolateral prefrontal cortex. [[Projects:RuleBasedDLPFCSegmentation|More...]]

|-

| | [[Image:Striatum1.png|200px|]]
| |

== [[Projects:RuleBasedStriatumSegmentation|Rule-Based Striatum Segmentation]] ==

In this work, we provide software to semi-automate the implementation of segmentation procedures based on expert neuroanatomist rules for the striatum. [[Projects:RuleBasedStriatumSegmentation|More...]]

|-

| | [[Image:Brain-flat.PNG|200px]]
| |

== [[Projects:ConformalFlatteningRegistration|Conformal Flattening]] ==

The goal of this project is for better visualizing and computation of neural activity from fMRI brain imagery. Also, with this technique, shapes can be mapped to shperes for shape analysis, registration or other purposes. Our technique is based on conformal mappings which map genus-zero surface: in fmri case cortical or other surfaces, onto a sphere in an angle preserving manner. [[Projects:ConformalFlatteningRegistration|More...]]

|-

| | [[Image:Fig1yan.PNG|200px|]]
| |

== [[Projects:BloodVesselSegmentation|Blood Vessel Segmentation]] ==

The goal of this work is to develop blood vessel segmentation techniques for 3D MRI and CT data. The methods have been applied to coronary arteries and portal veins, with promising results. [[Projects:BloodVesselSegmentation|More...]]

|-

| | [[Image:Fig67.png|200px|]]
| |

== [[Projects:KnowledgeBasedBayesianSegmentation|Knowledge-Based Bayesian Segmentation]] ==

This ITK filter is a segmentation algorithm that utilizes Bayes's Rule along with an affine-invariant anisotropic smoothing filter. [[Projects:KnowledgeBasedBayesianSegmentation|More...]]

|-

| | [[Image:Stochastic-snake.png|200px|]]
| |

== [[Projects:StochasticMethodsSegmentation|Stochastic Methods for Segmentation]] ==

New stochastic methods for implementing curvature driven flows for various medical tasks such as segmentation. [[Projects:StochasticMethodsSegmentation|More...]]

|-

| | [[Image:GT-SulciOutlining1.jpg|200px]]
| |

== [[Projects:SulciOutlining|Automatic Outlining of sulci on the brain surface]] ==

We present a method to automatically extract certain key features on a surface. We apply this technique to outline sulci on the cortical surface of a brain. [[Projects:SulciOutlining|More...]]

|-

| | [[Image:Table1.png|200px|]]
| |

== [[Projects:KPCA_LLE_KLLE_ShapeAnalysis|KPCA, LLE, KLLE Shape Analysis]] ==

The goal of this work is to study and compare shape learning techniques. The techniques considered are Linear Principal Components Analysis (PCA), Kernel PCA, Locally Linear Embedding (LLE) and Kernel LLE. [[Projects:KPCA_LLE_KLLE_ShapeAnalysis|More...]]

|-

| | [[Image:Gatech SlicerModel2.jpg|200px]]
| |

== [[Projects:StatisticalSegmentationSlicer2|Statistical/PDE Methods using Fast Marching for Segmentation]] ==

This Fast Marching based flow was added to Slicer 2. [[Projects:StatisticalSegmentationSlicer2|More...]]

|}

Projects:SoftPlaqueDetection

2009-09-28T21:51:31Z

Slankton: /* Ongoing Work */

== Soft Plaque Detection ==

The ability to detect and measure non-calcified plaques (also known as soft plaques) may improve physicians’ ability to predict cardiac events. This is a particularly challenging problem in computed tomography angiography (CTA) imagery because plaques may have similar appearance to nearby blood and muscle tissue. This paper presents an effective technique for automatically detecting soft plaques in CTA imagery using active contours driven by spatially localized probabilistic models. The proposed method identifies plaques that exist within the vessel wall by simultaneously segmenting the vessel from the inside-out and the outside-in using carefully chosen localized energies that allow the complex appearances of plaques and vessels to be modeled with simple statistics. This method is shown to be an effective way to detect the minute variations that distinguish plaques from healthy tissue. Experiments demonstrating the effectiveness of the algorithm are performed on eight datasets, and results are compared with detections provided by an expert cardiologist.

== Detection Algorithm ==

This technique allows region-based segmentation energies to be spatially localized such that statistical models of the foreground and background adapt to image information as it changes over the domain of the image. This allows for improved modeling accuracy with simplified statistical models. Furthermore, it is particularly powerful for segmenting vessels, which often exhibit changing image intensities over their length, and for the identification of non-calcified plaques, which typically have only slight intensity differences from surrounding structures.

First, vessels are segmented using a localized energy that is ideal for vessel segmentation, because it allows the surface to expand into areas of similar local intensity as long as a larger difference exists between local interiors and exteriors. This allows rapid segmentation of vessels despite changing intensities along the length of the vessel.

Next, two initial surfaces are created that lie just inside and just outside the initial segmented surface. A separate localized energy is used to move the two surfaces toward each other. This energy is chosen to emphasize local differences so that expansion into nearby regions that have slightly different intensities is discouraged, even if the local means are similar. This more stringent constraint is quite valuable when attempting to differentiate between vascular plaques and surrounding tissue.

Areas where the two surfaces do not find the same boundary are identified as plaques.

== Ongoing Work ==

Future work on this method will include coupling the evolution of the interior and exterior surfaces so that information about local intensities and geometries can be shared in order to detect plaques more robustly. Furthermore, a larger study is planned in which a larger number of datasets will be analyzed, a quantitative analysis will be performed, and the method will be compared with intravascular ultrasound imagery to conﬁrm the presence and composition of detected plaques. We believe this work has the potential of being an important step forward in automatically detecting non-calcified plaques, which have been clearly linked with the occurrence of heart attacks and stroke.

== Publications ==

''In Print''

== Key Investigators ==

Georgia Tech: Shawn Lankton, Jacob Huang, Vandana Mohan, and Allen Tannenbaum

2007-08-16T23:58:17Z

Slankton:

Name: Shawn Lankton

School: Georgia Institute of Technology

Interests: Active contour segmentation, Brain atlas registration, Shape analysis

[http://www.bme.gatech.edu/groups/bil/people/Shawn.Lankton/Shawn.Lankton.html Lab Homepage]

[http://www.shawnlankton.com/ Personal Homepage]

User:Lankton

2007-08-16T23:56:09Z

Slankton:

Name: Shawn Lankton

School: Georgia Institute of Technology

Interests: Registration, Segmentation, Algorithms

[http://www.bme.gatech.edu/groups/bil/people/Shawn.Lankton/Shawn.Lankton.html Lab Homepage]

[http://www.shawnlankton.com/ Personal Homepage]