UC San Diego

Diffusion-weighted imaging (DWI) is routinely used to aid in the detection and characterization of prostate cancer. Given imaging time constraints in a clinical setting, it is important to maximize the statistical efficiency of a DWI examination of the prostate. The objective is to maximize the accuracy with which microstructural information about the prostate can be obtained while minimizing diffusion scan time. In this study, we apply estimation theory to evaluate the statistical efficiency of different DWI acquisitions and methods. Specifically, we show that the variance of DWI parameters estimated using nonlinear multiexponential signal models is considerably higher than the variance observed using linear signal models. We then derive a simple analytical expression for the efficiency of a linear estimator and use it to optimize b -value sampling for DWI of the prostate.