Machine and Deep Learning for Congenital Diaphragmatic Hernia (CLANNISH)

Congenital Diaphragmatic Hernia

congenital diaphragmatic hernia, pulmonary hypertension, big data, machine learning

Congenital Diaphragmatic Hernia (CDH) is characterized by an incomplete diaphragm formation, resulting in poor lung development (pulmonary hypoplasia), associated with altered vascularization of the lung (pulmonary hypertension), with respiratory and cardiovascular insufficiency at birth. Mortality and morbidity are extremely variable. Several efforts have been done to identify possible prenatal and postnatal indicators which could accurately predict patients' prognosis and to promote an individualized management. However, to date the accuracy of these factors with respect to the prediction of survival and disease severity still has limits. In the last years, there has been an impressive development of new research methodologies based on the artificial intelligence, also in the neonatal field. The Machine Learning (ML) method explores the possibility of building algorithms starting from the acquisition of relevant clinical data, and using them to make predictions or take decisions. Nevertheless, the ML method has never been applied to predict patient's outcome in newborns with CDH so far. Moreover, with the available tools, a reliable prediction on patient's risk of developing severe postnatal PH is not feasible. Our hypothesis is that the use of ML approach, based on multivariate analysis of different clinical pre- and postnatal variables, could allow the development of algorithms able to accurately predict patient's outcome.

The investigators will collect clinical and instrumental data regarding prenatal history as well as the medical and surgical postnatal course. In particular, the investigators will record data from a prenatal ultrasound performed between 25+0 and 30+6 weeks of gestation (before Fetoscopic Endotracheal Occlusion (FETO) procedure, in case of prenatal treatment): estimated fetal weight (EFW), amniotic fluid, Doppler velocimetry of umbilical artery, defect side, herniated organs, observed/expected lung-to-head ratio tracing (O/E LHR%), grading of hernia severity, Doppler velocimetry of contralateral pulmonary artery. Gestational age at diagnosis, details about FETO procedure, and the course of pregnancy will be also recorded. On fetal MRI, the investigators will calculate: observed/expected total fetal lung volume (O/E TFLV%), percentage of liver herniation (%LH), signal intensity of lung and liver on T2 sequences, mediastinal shift angle, apparent diffusion coefficient (ADC) on diffusion-weighted sequences (DWI). The radiographic pulmonary area will be calculated on digital chest x-ray performed within 24 hours after birth, by tracing the perimeter of the lung outlined by the rib cage and the diaphragm, excluding the mediastinal structures and the herniated organs. Regarding the neonatal course, the investigators will focus on pulmonary hypertensive status, need for ECMO, and deaths. In particular, pulmonary hypertension will be evaluated based on clinical parameters (such as systemic pressure, heart rate, oxygen saturation, and oxygen supplementation, inotropic drugs, vasopressors, pulmonary vasodilators) as well as echocardiographic parameters (systolic pulmonary artery pressure (PAPs) from tricuspid valve regurgitation, mean pulmonary artery pressure from pulmonary valve regurgitation, pulmonary artery flow, characteristics of the interventricular sept, shunts, cardiac anomalies). Echocardiograms in our NICU are performed bedside throughout the hospital stay. The investigators will consider one exam per day from birth to 48 hours after surgery, one exam per week in the following 4 weeks, one exam per months until discharge. Other relevant data, like neurologic complications, metabolic disorders or infections, will be recorded as well. Finally, the investigators will record data regarding the surgical course: day of intervention, type of surgical repair, use of patch, intra- or post-operative complications.