Meeting Banner
Abstract #0781

A Critical Consideration on the Absence of Significance and the Impact of Structure Size When Interpreting DTI and DKI Results

Filip Szczepankiewicz1, Jimmy Ltt2, Ronnie Wirestam1, Alexander Leemans3, Pia Sundgren2, 4, Danielle van Westen2, 4, Freddy Sthlberg1, 4, Markus Nilsson1

1Department of Medical Radiation Physics, Lund University, Lund, Sweden; 2MR Department, Lund University Hospital, Lund, Sweden; 3Image Sciences Institute, University Medical Center Utrecht, Utrecht, Netherlands; 4Department of Diagnostic Radiology, Lund University, Lund, Sweden

DTI and DKI parameters are commonly used to compare brain tissue in patients and controls. Such comparisons are frequently based on statistical tests, but analysis of statistical power and potential confounders are often overlooked. In this study we use the cingulum to demonstrate how heterogeneity of the variability and structure size at various positions along the structure can influence the diffusion parameters with respect to both statistical power and interpretation of the results.

Abstract PDF Presentation Video

Keywords

absence accompanied accomplished according actual allowed among anatomical anisotropy anterior apparent approximately assuming awareness bars bilaterally blue bottom brain bull bundle cause cent central challenged coefficient commonly component confounder contrast corpus correct corrected correction correlate correlated correlation coverage cross decide decrease defined depicted despite detect deviation diagnostic diffusion diffusivity either employing enabling encoding ends equal especially evaluate evaluated except exhibit exists expected extended field fields fractional framework function furthermore generate generated generating global healthy hence highest house hypothetical important improve indicates indicating induced inferred influence integrity interpretation interpreted interpreting jimmy knowledge kurtosis making matched medical often partial parts pathogenesis patterns placed plot post posterior potential power predictor prominent radiology radius rather readily reasons reduce reduced regression relation relationship relevant render required respectively retaining sciences section segmentation side significance significantly simplify situations solely statistical strong structure subject subjects surrounding terms thereby tract uncorrected variance variation varies varying volume volunteers white whole yield