Meeting Banner
Abstract #2173

A Novel Efficient Denoising Method for ASL Data: Assessment Using Voxel-Wise Network Analysis

Xiaoyun Liang1, Alan Connelly1, 2, Fernando Calamante1, 2

1Brain Research Institute, Florey Institute of Neuroscience and Mental Health, Heidelberg, VIC, Australia; 2Department of Medicine, Austin Health and Northern Health, University of Melbourne, Melbourne, VIC, Australia

The SNR of fMRI images is critical for functional connectivity studies. Although ASL has some advantages over BOLD fMRI, the reliability for detecting networks may be compromised due to its intrinsic low SNR. In this study, we proposed a denoising method combining block-wise non-local means and dual-tree complex wavelet transform to enhance the SNR of ASL images. Simulations show that the proposed method was superior to discrete wavelet transform. The validity of the proposed method has been further confirmed by the more robust detection of functional connectivity from in vivo data. Overall, the proposed method can enhance the SNR of ASL data significantly and thus enable more reliable network detection.

Abstract PDF Poster PDF

Keywords

accepted achieved activity added addition although analyses anatomical applied applying band bands better biological biomedical block blue bold bottom brain calibration characterized coefficients coherent combining common compensated complex component components compromised computational conducted confirmed connections connectivity consistent consistently control correlation decompose defined degree degrees delay details detect detecting detection determine determines develop diagnostic disease distribution domain dual efficient efficiently employed enhance epilepsy evaluate exponentially extracted filtering frequencies frequency functional general generated gray greatly health highest in vivo inputs institute intensity intrinsic inversion journal labeling local mapping matrix maxi measure measuring medicine mental metrics middle model network networks neural noise noises noisy normalization northern note novel optimal original overall peers perfusion plotted power processed properties proposed quality quantitative random reliability remove resolution retaining royal scanned search sensitivity significantly simulated simulation simulations since smoothing society software square stroke subject subjects superior table threshold topologies transform tree true truncated validity volume watts wavelet wise world yields