Dynamic manganese-enhanced MRI signal intensity processing based on nonlinear mixed modeling to study changes in neuronal activity

Abstract

We analyze data on the impact of testosterone on the dynamics of Mn2+ accumulation measured by magnetic resonance imaging in three songbird brain areas: the nucleus robustus arcopallii (RA), area X, and the high vocal center (HVC). Birds with and without testosterone were included in the experiment, and repeated measurements were available in both a pre- and post-drug administration period. We formulate a nonlinear modeling strategy, allowing for the incorporation of (1) within-bird correlation, (2) the nonlinearity of the profiles, and (3) the effect of treatment. For two of the outcomes (RA and area X), biological theory suggests a parametric form, but for HVC this is not the case. Because the HVC outcome bears some resemblance with the two-compartment model known from pharmacokinetics, this model was considered a sensible choice. We use a different model, based on fractional polynomials, as a sensitivity analysis for the latter. All methods used provide good fits to the data, confirm results from previous, simple analyses undertaken in the literature, but were able to detect additional effects of treatment that had so far gone undetected. The fractional polynomial and two-compartment models provide similar substantive conclusions; the two together can be seen as a form of sensitivity analysis.