Consistent association between image features of fetal lungs from different ultrasound equipments and fetal lung maturity from amniocentesis

Respiratory disorders in premature are outlined as one of the major causes of newborn morbidity. Therefore, estimation in utero of the risk that a particular newborn will present respiratory distress syndrome is critical in perinatal medicine. Fetal lung maturity assessment by non-invasive methods is an unsolved problem since 80´s. There has been a significant effort reporting that ultrasound images carry information associated with changes of fetal lung microstructure but it required the lung-to-liver measurements to be calibrated, which limits clinical feasibility. In this work, we explored the association between image features only from a fetal lung computed by a previously reported probabilistic model and fetal lung maturity. Two studies validated image features obtained from different US equipments: 957 scans were correlated to gestational age (r>;0.97); and 83 scans were modeled to estimate a normalized lung maturity index in 7 blind samples. Bootstrapping provided median r>;0.82.