Over the last 10–15 years, it has been established that disproportionate placental size relative to fetal size is associated with adverse pregnancy outcomes. However, since this knowledge is relatively new, placental volume measurements are not part of the standard prenatal examinations today. The only established method for such measurements during pregnancy is MRI imaging. However, this is both expensive and time consuming and thus not feasible as a part of the standard examinations. Many researchers have tried to estimate placental volume using ordinary 2D and 3D ultrasound, but none of them have validated these estimates against a gold standard such as MRI before. Our findings suggest that these techniques are not accurate, probably mainly due to the large size and variable geometry of the placenta. The aim of this project is to develop an algorithm based on machine learning for automatic analysis of 2D ultrasound images to accurately measure placental volume and to verify the method on real data. This algorithm can also be applied to other imaging needs using only 2D ultrasound. Development of valid measurements of placental size in ongoing pregnancies will enable the identification of pregnancies with high risk of adverse outcome that we cannot identify today. Such identification is necessary for preventive interventions and can thus potentially improve maternal care and reduce fetal death rates all over the world.
Project leader: Inge Hovd Gangås
Institution: SINTEF TTO AS