Object: We sought to investigate the dependence of intravoxel incoherent motion (IVIM)-related perfusion fraction (f) estimates on the transverse relaxation of brain tissue, blood, and cerebrospinal fluid (CSF), attempting to overcome the influence of CSF on conventional f maps.
Materials and methods: Eighteen patients with gliomas underwent DWI with 14 b-values (0-1,300 s/mm(2)) and two distinct echo times (TEs). Regions of interest representing tumour and normal brain tissue were analysed by calculating the f values for both TEs. A mask for pixels with relevant CSF partial volume was subsequently created. The f values were tested for significant differences.
Results: We found statistically significant differences between the two TEs in the f values for cortical and juxtacortical structures and non-enhancing areas of the tumour /oedema. Normal white matter and gadolinium-enhancing tumour tissue appeared insensitive to TE variation. In all tissue types examined, the masking of voxels with considerable CSF content was able to overcome issues of erroneous f estimation and calculation of f values insensitive to TE changes was feasible.
Conclusion: Due to the complex interaction in the relaxation rates of CSF, blood, and tumour tissue, the estimation of f values is affected by the choice of TE. Only f values in normal white matter and tumour tissue--which largely comprise blood voxels with minor CSF partial volume--may be clinically applicable in the present form of IVIM-based DWI analysis. Going a step further, and after removing voxels with heavily TE-susceptible f values, we were able to obtain accurate and TE-independent f values in contrast-enhancing tumour tissue, white matter, and grey matter, which were essentially consistent with those reported in the literature.