Automatic differentiation of placental perfusion compartments by time-to-peak analysis in mice

The aim of this study was to develop an automatic differentiation of two perfusion compartments within the mouse placenta based on times of maximal contrast enhancement for a detailed and reproducible perfusion assessment.

Placentas (n = 17) from pregnant BALB/c mice (n = 10) were examined in vivo at 7T on gestation day 16.5. Coronal dual-echo 3D T1-weighted gradient-echo sequences were acquired after application of contrast agent for dynamic MRI. An adapted gamma variate function was fitted to the discrete concentration time curves to evaluate the effect of noise on perfusion and segmentation results. Time-to-peak maps based on fitted and discrete curves of each placenta were used to classify each voxel into the high- or low-blood flow compartment using k-means clustering. Perfusion analysis was performed using the steepest slope model and also applied to fitted and discrete curves. Results were compared to manually defined compartments from two independent observers using the Dice coefficient D.

Manually defined placental areas of high-flow and low-flow were similar to the automatic segmentation for discrete (D = 0.76/0.75; D = 0.76/0.79) and fitted (D = 0.80/0.80; D = 0.81/0.82) concentration time curves. Mean perfusion values of discrete and fitted curves ranged in the high-flow compartment from 134 to 142 ml/min/100 ml (discrete) vs. 138-143 ml/min/100 ml (fitted) and in the low-flow compartment from 91 to 94 ml/min/100 ml (discrete) vs. 74-82 ml/min/100 ml (fitted).

Our novel approach allows the automatic differentiation of perfusion compartments of the mouse placenta. The approach may overcome limitations of placental perfusion analyses caused by tissue heterogeneity and a potentially biased selection of regions of interest.