Compartment-based models of diffusion MRI signals have become popular for probing tissue microstructure. However, the standard models do not explicitly model compartment-specific T2 relaxation. This has been shown to cause a TE-dependence of DTI-derived measures in white matter, which can confound the interpretation and quantification of results. Here we explored the TE-dependence of the widely used NODDI model and proposed a technique to determine parameter estimates that are TE-independent. This could be useful in investgateing diseases where changes of T2 and signal fraction interact.