Meeting Banner
Abstract #3989

Unsupervised Brain Tumor Tissue Differentiation Based on MRSI with Correction for the Chemical Shift Displacement Artifact

Diana M. Sima1, 2, Sofie Van Cauter3, Yuqian Li4, Anca R. Croitor-Sava1, 2, Uwe Himmelreich5, Sabine Van Huffel1, 2

1Electrical Engineering / ESAT-SCD, KU Leuven, Heverlee, Leuven, Belgium; 2Future Health Department, iMinds, Heverlee, Leuven, Belgium; 3Department of Radiology, University Hospitals Leuven, Leuven, Belgium; 4School of Electronic Engineering, University of Electronic Science and Technology of China, Chengdu, China; 5Department of Imaging & Pathology, Biomedical MRI Unit, KU Leuven, Leuven, Belgium

We propose and validate a new computational method for unsupervised tissue differentiation for brain tumor patients, based on short echo time 2D magnetic resonance spectroscopic imaging data, which takes into account possible spectral distortions due to significant Chemical Shift Displacement artifact.

Keywords

account acquisition actual affected ahead already although another applied approximate artifact assigned audience bands beam biomedical blue border borders bottom brain called carrier central chemical china circular classification clinical coefficients columns components computed concentration considered correction cropped developed differentiation directly disappears displacement distorted distribution domain eddy electrical electronic encode encoding engineering error estimation evaluating even excitation extend extract factorization filtering frequency future grade green grid health hierarchical highest hinder ignored illustration implies indicate inner linear lipid localization localized location maps matrices matrix metabolites mixed mixing necrosis negative offset outer outside part patient patients posterior press previously prior problem product profile proportional protocol pulses radiology rapid real recovery reflect reliability represent rows samples saturation scanner school science selected semi sense shimming side significantly simulated since skull slice solved source sources space spatial spectra spectral spectroscopic spectroscopy still suffer suppressed system taken target tissue tumor tumors type typically unit unsupervised volume water wrongly