Meeting Banner
Abstract #2452

A Robust Algorithm Framework for Small DTI Samples

Ivan Maximov1, Farida Grinberg1, N. Jon Shah1, 2

1Institute of Neuroscience and Medicine 4, Forschungszentrum Juelich GmbH, Juelich, Germany; 2RWTH Aachen University, Department of Neurology, Faculty of Medicine, JARA, Aachen, Germany

Low redundancy DTI data sets are a rather complicated problem for diffusion tensor estimation, especially in a clinical human brain imaging. The diffusion signal attenuations are frequently corrupted by a physiological noise such as a cardiac pulsation, bulk head motion, respiratory motion, etc. As a consequence, diffusion tensor estimation becomes unstable and very poor. We have developed fitting algorithms based on the least trimmed squares and the median absolute deviation robust estimators in order to improve the tensor assessment in small DTI samples.

Abstract PDF Poster PDF

Keywords

absolute acquisition allow allows anal angular anisotropy apparent application applied arranged artificially assessment attenuation axes better brain bulk cardiac caused clinical coefficient comp complicated consistency contrast corrupted developed deviation deviations diagnostic diffusion diffusivity disorders efficiency eigenstates eigenvalues eigenvector eigenvectors ellipsoidal enable encoding equal erase especially established estimation estimations estimator estimators evaluated evaluation example exploits faculty form fractional framework function general good gradient gradients great head human improve in vivo institute invariants least length localize median medicine might motion neurology noise noninvasive objective original outcome outliers pathologies physiological play poor potential powerful problem pulsation rather recommended relatively repetitions residuals restore robust role rotational sample samples sensitive series sets shah short significantly simulated simulations squares stat statistical statistics strongly tensor tensors theory tissue trimmed truncation turn typically valuable various