A physics-based coordinate transformation for 3-D image matching

Davis, Mark H. A.; Khotanzad, A.; Flamig, Duane P.; Harms, Steven E.

doi:10.1109/42.585766

Cited by 237 publications

(148 citation statements)

References 39 publications

Supporting

Mentioning

138

Contrasting

Unclassified

Order By: Relevance

“…The TPS method is one of the most commonly used methods for landmark registration. It is suitable because it does not need a determination of parameters by the user, as in EBS, and the computation is performed faster in comparison to EBS and VS (Davis et al, 1997). A TPS interpolant, f(x, y, z), minimizes its "bending energy" and is defined for any coordinate of the image by f x; y; z ð Þ= a 0 + a x x + a y y + a z z + X n i = 1…”

Section: Local Landmark-based Registrationmentioning

confidence: 99%

“…Thereafter the displacement field around the landmarks must be interpolated from the displacement of the landmarks. Splines are widely recognized as one of the most successful methods to interpolate the coordinate transformation on intermediate locations, minimizing bending energy (Davis et al, 1997). There are several types of splines: TPS, elastic body splines (EBS), and volume splines (VS).…”

Section: Local Landmark-based Registrationmentioning

confidence: 99%

See 1 more Smart Citation

Local landmark-based registration for fMRI group studies of nonprimary auditory cortex

et al. 2009

View full text Add to dashboard Cite

Interindividual functional and structural brain variability is a major problem in group studies, in which very focal activations are expected. Architectonic studies have shown that the human primary auditory area, which is located with a great constancy on Heschl's gyrus, is surrounded by several nonprimary auditory areas with surface areas of 40-310 mm 2 . The small size of the latter makes them only partially accessible to fMRI group studies, because of imprecision in realignment when using currently available registration procedures. We describe here a new method for sulcal realignment using a non-rigid local landmark-based registration and show its application to the registration of fMRI acquisitions on the supratemporal plane. After an affine global voxel-based registration, which transforms all brains into the same standard space, we propose a non-rigid local landmark-based registration method based on thin-plate splines for matching the two sulci delimiting Heschl's gyrus of a given brain to the corresponding sulci of a reference brain. We show here that, in comparison with global affine and non-rigid approaches, our method leads in group studies to i) a much more precise alignment of Heschl's gyrus; and ii) a putatively optimal superposition of functionally corresponding areas on and around Heschl's gyrus.

show abstract

Section: Local Landmark-based Registrationmentioning

confidence: 99%

Section: Local Landmark-based Registrationmentioning

confidence: 99%

Local landmark-based registration for fMRI group studies of nonprimary auditory cortex

et al. 2009

View full text Add to dashboard Cite

show abstract

“…To interpolate it to something defined on the whole image grid an elastic body spline interpolation is used. The elastic body spline is a 3-D spline that is based on a physical model (the Navier equations) of an elastic material [4]. This interpolation is driven by a physical model, making it a natural choice for regions where no landmarks are found.…”

Section: Methods: Learning Image Deformationsmentioning

confidence: 99%

SVF-Net: Learning Deformable Image Registration Using Shape Matching

Rohé

Datar

Heimann

et al. 2017

Lecture Notes in Computer Science

256

191

View full text Add to dashboard Cite

Abstract. In this paper, we propose an innovative approach for registration based on the deterministic prediction of the parameters from both images instead of the optimization of a energy criteria. The method relies on a fully convolutional network whose architecture consists of contracting layers to detect relevant features and a symmetric expanding path that matches them together and outputs the transformation parametrization. Whereas convolutional networks have seen a widespread expansion and have been already applied to many medical imaging problems such as segmentation and classification, its application to registration has so far faced the challenge of defining ground truth data on which to train the algorithm. Here, we present a novel training strategy to build reference deformations which relies on the registration of segmented regions of interest. We apply this methodology to the problem of inter-patient heart registration and show an important improvement over a state of the art optimization based algorithm. Not only our method is more accurate but it is also faster -registration of two 3D-images taking less than 30ms second on a GPU -and more robust to outliers.

show abstract

“…Transform-model estimation involves the identification of a number of control points on both the image to be registered and the reference image that can be used to warp one to fit the other. Global approaches such as shape-preserving mapping and local approaches such as thin-plate splines (Davis et al 1997) are commonly used transform models.…”

Section: Literaturementioning

confidence: 99%

An optimization technique for addressing DEM misregistration in hilly terrain

Goodchild

2015

Annals of GIS

View full text Add to dashboard Cite

Digital elevation models (DEMs) represent the Earth's topography and support a variety of applications, ranging from extracting watershed drainage structure to measuring glacier volume. Applications that require conflation of multiple DEMs are problematic, however, because of misregistration. We explore a spatial optimization technique to quantify the misregistration on the pixel level between NASA's Shuttle Radar Topography Mission (SRTM) elevation data set and the USGS's National Elevation Dataset (NED). The misregistration, estimated within blocks by horizontal offset and direction, is modelled spatially in terms of typical topographical parameters: slope, aspect and elevation. A Nelder-Mead algorithm was implemented to compute the local shift at two study sites in the San Gabriel Mountains (SGM) and Santa Monica Mountains (SMM) in Los Angeles County, California. The magnitude of misregistration is generally less than the distance between DEM postings, and varies by terrain type, being larger in steeper terrains; nevertheless, misregistration is systematic within these continuous terrains that have similar topographical features. ARTICLE HISTORY

show abstract

A physics-based coordinate transformation for 3-D image matching

Cited by 237 publications

References 39 publications

Local landmark-based registration for fMRI group studies of nonprimary auditory cortex

Local landmark-based registration for fMRI group studies of nonprimary auditory cortex

SVF-Net: Learning Deformable Image Registration Using Shape Matching

An optimization technique for addressing DEM misregistration in hilly terrain

Contact Info

Product

Resources

About