Meeting Banner
Abstract #3559

Local Regularization of the Diffusion Tensor by Means of Independent Component Ananlysis and Total Variation - Application to High Resolution DTI

Gernot Reishofer1, Karl Koschutnig2, Christian Langkammer3, Stefan Ropele3, Stephen Keeling4, Robert Merwa5, Franz Ebner

1Radiology, Medical University of Graz, Graz, Styria, Austria; 2Psychology, University of Graz; 3Neurology, Medical University of Graz; 4Mathematics and Scientific Computing, University of Graz; 5Medical Engeneering, Upper Austria University of Applied Sciences

Advanced scan techniques in diffusion tensor imaging (DTI), such as readout segmented EPI enable high resolution DTI, but suffer from low signal to noise ratio. In this work we present a novel method for the spatially dependent regularization of the diffusion tensor based on independent component analysis and total variation regularization. We demonstrate that the diffusion tensor evaluated from noisy DWI data is successfully denoised, while fine structural details are preserved. This allows for the application of high resolution DTI in a clinically acceptable time.

Abstract PDF E-Poster Video

Keywords

acceptable according acquisition advanced anatomical anisotropy application applied basis become bound brain calculation carried channel clinical clinically coil color comparing complex component components computing containing dashed dependent derived details diffusion ding distribution domain done eigenvalues element elements enables entropy evaluated excellent fact fiber field fine focus fractional functional giving head healthy homogeneous improve independent input insufficient inverting keeling linear local long magnified maps mark marked mathematics medical micro might minimization minimize minimized minimizes moving neurology noise noisy novel optimization original orthogonal post preserved prevent problem procedure processed processing projection propose proposed psychology quantities radiology readout rectangle reduction regularization regularized relation representing resolution restored sciences scientific segmented separate separating served setup several shape shorter shots slice software spatial spatially specially specifically steps structural structures subspace successfully suppressed suppression system systems tanner tensor tensors term tissue tracking trans transformation transformed trio uncertainties uncertainty uniform upper variance variation view vision visualization volunteer white whole