Accuracy and precision of random walk with barrier model fitting: Simulations and applications in head and neck...

Random Walk with Barrier Model (RWBM) can quantify microstructural parameters (free diffusivity, cell size, and membrane permeability) in head and neck cancers (HNCs) using time-dependent diffusion MRI. However, model fitting remains challenging due to limited number of measurements, low signal-to-noise ratio (SNR) and complex nonlinear biophysical models. In this work, we numerically evaluated RWBM fitting performance under a clinically feasible imaging protocol by comparing fits to the short-time-limit expression (STL-RWBM) and to the general RWBM across a broad range of SNRs. We also examined the model fitting degeneracy and fitting landscape for both fitting methods to elucidate their fitting behaviors. Numerical findings were further compared with model-fitting results from HNC patient data. We observed that under clinically relevant SNRs, fitting the STL-RWBM provided low-variance estimates of free diffusivity and membrane permeability, with biases comparable to those obtained from fitting the general RWBM. However, the general RWBM yielded low-bias estimates for cell size. The flat and tube-like fitting landscapes of the general RWBM led to high variance, frequent convergence to boundary constraints and spurious correlation among fitted parameters in vivo. The more restricted fitting landscape of the STL-RWBM substantially reduced variance. Both fitting methods produced cell size estimates in HNCs that were consistent with prior pathological findings. In conclusion, this work provided a comprehensive analysis of RWBM fitting in clinical settings and may guide optimizations of data acquisition and model fitting methods.