Meeting Banner
Abstract #2029

Influence of the Non-Linear Fitting Approach on Intravoxel Incoherent Motion (IVIM) MRI: A Model Selection Study

Moritz Schneider1, Michael Ingrisch1, Matthias Moser2, Heidrun Hirner, Clemens Cyran, Konstantin Nikolaou, Maximilian Reiser, Olaf Dietrich1

1Josef Lissner Laboratory for Biomedical Imaging, Department of Clinical Radiology, LMU Ludwig Maximilian University of Munich, Munich, Germany; 2Department of Clinical Radiology , LMU Ludwig Maximilian University of Munich, Munich, Germany

Intravoxel incoherent motion (IVIM) MRI is performed by fitting a biexponential model function to the measured diffusion-weighted signal attenuation. While the conventional biexponential model has 4 free parameters including the initial signal at b-value 0, several studies use only the signal attenuation relative to this initial value, discriminating, thus, artificially the measurement at b=0 and at higher b-values. In this study, the Akaike Information Criterion is used to determine whether the 3- or 4-parameter model is more appropriate for typical IVIM data. The comparison results in a very strong preference of the 3-parameter approach (relative attenuation).

Keywords

acquisitions animal animals apparent appearing applied approaches appropriate appropriateness artificially assess assessed attenuation attenuations biomedical breast carcinoma certain channel clear clinical coefficient coil confirmed criterion decide dedicated defined describe deviations diffusion discriminating either evaluated evaluation examined example excluding exhibit field fits fitted fitting fraction free fully function highest homogeneous implanted improved inclusion incoherent indicates influence initial justified laboratory linear logarithm marginally materials matrix measured model models motion necrosis normalization nude perfusion phys preference probability procedure properties proposed pseudo quality radiology ranging rapid rats reflected selection several shot slice slices slightly squares step strategies strong studies sufficient summarized surface system table terms tissue tumor tumors typical typically view whether