Evaluating the Accuracy and Impact of Registration in Dynamic Contrast-Enhanced Breast MRI

• This article presents a quantitative evaluation framework-incorporating a novel heterogeneous biomechanical model-for objectively comparing the accuracy of registration algorithms in dynamic contrast-enhanced (DCE) MRI of the breast; an evaluation of several algorithms using this framework; and several clinical examples where accurate registration significantly changes the shape of the enhancement curves associated with suspicious regions of interest (ROIs) identified by a radiologist. The experimental results demonstrate: (i) the efficacy of the evaluation framework; (ii) that a good registration algorithm can accurately recover the shapes of voxel enhancement curves from breast DCE-MRI data containing motion; (iii) that motion of as little as I mm can significantly change the shape of the mean enhancement curve for an ROI; and (iv) that accurate registration can significantly change the shape of enhancement curves estimated for small and large ROIs even when there is little or no visible evidence of motion. This suggests that all DCE-MRI breast data should be spatially aligned using a nonrigid registration algorithm before the analysis of contrast enhancement. (C) 2009 Wiley Periodicals, Inc. Concepts Magn Reson Part B (Magn Reson Engineering) 3513: 106-120, 2009

Subject headings

TEKNIK OCH TEKNOLOGIER  -- Elektroteknik och elektronik -- Signalbehandling (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Electrical Engineering, Electronic Engineering, Information Engineering -- Signal Processing (hsv//eng)

NATURVETENSKAP  -- Data- och informationsvetenskap -- Datorseende och robotik (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Computer Vision and Robotics (hsv//eng)

Keyword

registration

lesions

cancer

breast mri

biomechanical modeling

models

performance

image registration

dynamic contrast-enhanced mri

diagnosis

preserving nonrigid registration

validation

enhancement curve

Publication and Content Type

art (subject category)

ref (subject category)