VP14.02: Machine learning‐based phenogrouping of third trimester echocardiography can predict the neonatal clinical course in fetuses with tetralogy of Fallot

Abstract

deformity and hydrocephalus may require cephalocentesis and assisted delivery. A 30 years old, G2 P1, was referred to Fetal Medicine Unit at 22+6 weeks because of multiples fetal abnormalities on a routine anomaly scan. No significant medical, family and drug history. Ultrasound findings showed a cover-leaf shape, moderate bilateral ventriculomegaly, and holoprosencephaly. Cerebellum and corpus callosum absent with small occiput and encephalocele. Thoracic dystrophy, bell-shaped chest and short ribs. Cardiac views appear normal. Amniocentesis was performed and mutation in fibroblast growth factor receptor gene 3 (FGFR3 gene mutation) confirmed the diagnosis of thanataphoric dysplasia. Patient and partner were informed of the lethality of the condition and counselled throughout the whole pregnancy. She had a breech vaginal delivery at 37 weeks. To conclude, thanatophoric dysplasia is a rare lethal condition. Early diagnosis is essential for appropriate counselling and managment of pregnancy. Objectives: The aim of this study was to investigate the existence of different patterns of fetal cardiac hemodynamic remodelling, and their association with the neonatal presentation in tetralogy of Fallot (TOF), through the integration of echocardiographic data by means of interpretable unsupervised machine learning. Methods: Third trimester fetal echocardiographic data from 36 healthy and 30 TOF fetuses was used. We defined a composite outcome score indicating the occurrence of any of four adverse neonatal events. We used multiple kernel learning and k-means clustering to reduce data dimensionality, position patients based on similarities and find homogeneous groups of patients with similar hemodynamics. Results: Clustering of the low-dimensional space resulted in 4 clusters (Cl) (figure a), with significant different proportion of TOF patients and composite adverse outcome score. Cl4 has the highest proportion of SGA fetuses (figure c) and compromised fetuses with highest pulmonary and tricuspid A wave peak velocities and reduced and delayed aortic peak velocity (figure b). In contrast, Cl1 includes fetuses with bigger pulmonary valves and lowest pulmonary and tricuspid A wave peak velocities (figure b). Conclusions: Our results serve as a proof-of-concept that inter-pretable unsupervised machine learning can be useful to explore and understand different fetal patterns of cardiac remodelling in CHD and relate these to neonatal clinical course. Supporting information can be found in the online version of this abstract VP14.03 Abstract withdrawn VP14.04 Utility of vascular corrosion casting technique in the diagnosis of fetal ductus arteriosus abnormalities Objectives: The anatomy of ductus arteriosus (DA) can be varied in different congenital heart defects (CHDs), and it is difficult to fully discover the DA and other associated cardiac anomalies by prenatal ultrasound. This study was aimed to investigate the utility of cardiovascular casting for the demonstration of fetal DA abnormalities. Methods: This retrospective study enrolled fourteen fetuses (23 to 28 +2 gestational weeks) with severe CHDs diagnosed by prenatal echocardiography and casting technique from January 2017 to October 2018. Cardiovascular cast specimens were made by a modified vascular corrosion casting technique. The architectures of DAs and the whole cardiovascular system were observed and analysed, and then the cast findings were compared with prenatal ultrasonic results. Results: 18 ductal abnormities were indicated by prenatal echocar-diography in 14 case. Cast findings consisted with ultrasound in 10 ductal abnormalities, revealed additional 15 ductal abnormalities misdiagnosed by sonography and corrected six abnormalities misdiagnosed prenatally. Conclusions: Comparing with ultrasound, casting technique has its own superiority in exhibiting ductus abnormalities, especially in certain types such as course, origin and absence abnormalities of ductus. Supporting information can be found in the online version of this abstract VP14.05 Evaluation of ventricular function in the fetal transposition of great arteries by velocity vector imaging Objectives: The aim of this study was to test the hypotheses that: 1. fetuses with D-TGA and TBA has decreased the left ventricular (LV) and right ventricular (RV) mechanical function compared with fetuses with normal cardiac anatomy, and 2. the decreased myocardial mechanical function in fetal D-TGA and TBA was associated with the degree of transposition of the great arteries, pulmonary stenosis and ventricular septal defect. Methods: We retrospectively identified 49 cases of fetal D-TGA and 29 cases of fetal TBA encountered from 2015 to 2019 and compared findings to 49 controls of comparable gestational age. Postprocessing analysis of the left ventricle and right ventricle were done using syngo Velocity Vector Imaging version 2.0.gGlobal and regional longitudinal systolic peak velocity, longitudinal systolic strain and longitudinal systolic strain rate were generated. Both the LV and RV were analysed in fetuses with normal cardiac anatomy for comparison.