Improved Placental Parameter Estimation Using Data-Driven Bayesian Modelling

摘要

The placenta plays a key contribution to successful pregnancy outcome. New MR imaging techniques are able to reveal intricate details about placental structure and function and measure placental blood flow and feto-placental oxygenation. Placental diffusion-weighted MRI is however challenging due to maternal breathing motion and poor signal-to-noise ratio making motion correction important for subsequent quantitative analysis. In this work, we (ⅰ) introduce an iterative model-based registration technique which incorporates a placenta-specific model into the motion correction process and (ⅱ) describe a new technique making use of a Bayesian shrinkage prior to obtain robust estimates of individual and population trends in parameters. Our results suggest that the proposed registration method improves alignment of placental data and that the Bayesian fitting technique allows the estimation of voxel-level placenta flow parameters and the population trend in each parameter with gestational age (GA). We report gestational age dependent differences in vascular compartments and fetal oxygen saturation values observed across 9 normally grown pregnancies between 25-34 weeks gestational age and show qualitatively improved parameter mapping and more precise longitudinal fitting. Fetal oxygen saturation (FO_2) is observed to decrease at FO_2 = -3.6(GA_(weeks)) + 190.2(%). This technique provides a robust framework for analysing longitudinal changes in both normal and pathological placental function.