Cerebrospinal fluid (CSF) diffusion into the periventricular white matter (WM) is one of the pathophysiological features of hydrocephalus of any kind, including normal pressure hydrocephalus (NPH) (1,2). The increased volume fraction of CSF in the periventricular WM leads to typical periventricular hyperintensities (PVH) in T 2 -weighted images (3,4). But also other white matter changes such as degeneration or ischemia can lead to PVH and the prognostic implications are still unsettled. Tullberg et al. (5) reported that especially an irregular type of PVH around the frontal horns diminishes after shunt surgery in contrast to smooth PVH. In most cases hyperintensities are only analyzed for the presence, size, and confluence (6) or by the total volume fraction of periventricular hyperintensities compared to the volume fraction of the brain parenchyma (7).Quantification of CSF and tissue volume fractions in a volume-of-interest (VOI) is an established method in quantitative MR spectroscopy (MRS) (8). In this method timedomain data (FIDs) of a T 2 sequence, such as a stimulatedecho acquisition mode (STEAM) sequence (9), is fitted with a biexponential tissue-CSF model. The main application area is the correct estimation of brain metabolite concentrations from localized MR spectra, as neglected CSF volume fractions are a considerable source of error (10,11).Recently the possibility of quantitative blood oxygenation level-dependent (qBOLD) measurements was proposed by He and Yablonskiy (12). The proposed signal model was used to quantitative evaluate hemodynamic parameters such as oxygen extraction fraction (OEF), blood volume, and deoxyhemoglobin concentration. This model can as well be used to evaluate interstitial fluid (ISF) and CSF volume fractions for the whole brain instead of just a selected VOI. This could potentially be used as a method to quantify hydrocephalus noninvasively, or to detect early stages of NPH that as yet show no PVH.The aim of this study was to test if an MR signal model of brain tissue similar to the one proposed by He and Yablonskiy (12), which includes contributions from gray matter (GM), WM, and ISF/CSF, permits quantitative measurements of ISF/CSF volume fractions on a voxel-byvoxel basis. The proposed model was utilized to estimate quantitative ISF/CSF volume fractions in young healthy human subjects based on measurements with long echo times up to 301 ms.
MATERIALS AND METHODS
Signal ModelsIn general a single component model with a monoexponential FID is used in describing signal behavior. The signal in this case can be described in the following way: In this study signal information was gathered with a gradient recalled echo / echo planar imaging (GRE-EPI) sequence. The lowest echo time (TE) that could be used in our setting was 21 ms. Therefore, the fast component of the brain is practically invisible in the measured data and we adopted the basic two-component model for the brain tissue suggested by He and Yablonskiy (12). In this model we have a major intracellular signal component ...