EP01.54: Placenta functional study with BOLD MRI: random biopsies versus whole‐placenta analysis

Functional MRI has been identified as potential tool to study the placental function in-vivo. However, the offline processing of these data is time-consuming, particularly when it comes to manual segmentation. We compared the whole-placenta analysis to random samples with different placental volumes. We performed a BOLD MRI acquisition in normoxia, followed by a period of oxygen supply until placental oxygen saturation in a cohort of normal pregnancies. After a non-rigid registration to correct motion artifacts, a whole-placenta mask was manually segmented. Thereafter, three random samples were automatically generated within the whole-placenta mask, mimicking random biopsies and covering 10%, 30% and 50% of the placental area. The signal change over time was modelled using a sigmoid function, from which the intensity of enhancement () and a temporal variation coefficient (min-1, controlling the slope of the curve) were deduced. The parameters from the random samples were compared to the whole-placenta parameters. The mean differences in between the whole placenta and the random samples were 20.08%, 8.29% and 5.15% for the 10%, 30% and 50%-random masks, respectively. Subsequently, the ICCs were 0.91 [0.8, 0.96], 0.99 [0.97, 1] and 0.99 [0.99, 1] for the 10%, 30% and 50%-random masks, respectively. Similarly, the mean differences in (min-1) between the whole placenta and the random samples were 32.13%, 10.25% and 7.71% for the 10%, 30% and 50%-random masks, respectively. The ICCs were 0.02 [-0.41, 0.45], 0.96 [0.91, 0.99] and 0.95 [0.88, 0.98] for the 10%, 30% and 50%-random masks, respectively. In addition, the fitting failed in 3 cases for the 10%-mask, 2 cases for the 30%-mask and 1 case for the 50% mask, requiring new random samples. Our results suggest that small random placenta samples do not reflect the functional parameters with precision, illustrating the presence of functional variations within the placenta.