Partial volume effect compensation for improved reliability of quantitative blood–brain barrier permeability

Abstract

Blood-brain barrier (BBB) plays an important role in the pathophysiology of many central nervous system disorders. In the past, a number of laboratory techniques have been proposed to quantify permeability coefficient, k(i), an important index of barrier function. Recently, MRI has been used to estimate k(i) based on the unidirectional tracer kinetics model in one compartment as proposed by Patlak et al. and has been found to be in good agreement with the gold standard quantitative autoradiography technique. Rapid data acquisition, a prerequisite of this MRI-based technique, causes a compromise in spatial resolution resulting in partial volume (PV) averaging, an effect that is seldom explicitly compensated for in quantitative neuroimaging studies. This may have profound effect on the reliability of estimates obtained using quantitative methods. Existing PV compensation techniques that use complex statistical algorithms perform corrections on stationary images. In this proof-of-principle study, the effect of PV averaging on BBB permeability coefficient has been evaluated using a simulation model, and a postprocessing technique that makes use of dynamic information has been proposed for PV compensation in order to improve the reliability of this quantitative method. A computer simulation model is presented, which evaluates the effect of PV averaging on permeability coefficient estimates. Beginning with a known k(i), a PV compensation technique is proposed, which aims at correcting calculated k(i) to obtain the original estimate. The application of the PV compensation technique is demonstrated in a rat stroke brain model. Magnetic resonance imaging experiments were performed in Wistar rats (n=2) on a 4.7-T scanner. After acquiring localizer, T2-weighted and diffusion-weighted images, a rapid T1 mapping protocol was implemented to acquire one pre-gadolinium-diethylenetriaminepentaacetic acid baseline data set followed by a series of postinjection data sets. The data were postprocessed without and with application of PV compensation technique to obtain a k(i) estimate. The issue of PV averaging as a result of limited spatial resolution is often not addressed in quantitative MRI studies. In this work, simulation experiments have provided useful insight into the PV effects on permeability coefficient estimate. The findings of the simulation experiments agree well with the results obtained from MR experiments. Results from the MR experiments suggest that it may be important to perform PV compensation in order to improve the reliability of permeability coefficient estimates. Future work involves classification of tissue component into gray and white matter and CSF to improve the accuracy of the compensation technique and to investigate repeatability of the technique in a larger group of animals.