Deep-learning model for prenatal congenital heart disease screening generalizes to community setting and outperforms clinical detection.

OBJECTIVES: Despite nearly universal prenatal ultrasound screening programs, congenital heart defects (CHD) are still missed, which may result in severe morbidity or even death. Deep machine learning (DL) can automate image recognition from ultrasound. The main aim of this study was to assess the performance of a previously developed DL model, trained on images from a tertiary center, using fetal ultrasound images obtained during the second-trimester standard anomaly scan in a low-risk population. A secondary aim was to compare initial screening diagnosis, which made use of live imaging at the point-of-care, with diagnosis by clinicians evaluating only stored images. METHODS: All pregnancies with isolated severe CHD in the Northwestern region of The Netherlands between 2015 and 2016 with available stored images were evaluated, as well as a sample of normal fetuses' examinations from the same region and time period. We compared the accuracy of the initial clinical diagnosis (made in real time with access to live imaging) with that of the model (which had only stored imaging available) and with the performance of three blinded human experts who had access only to the stored images (like the model). We analyzed performance according to ultrasound study characteristics, such as duration and quality (scored independently by investigators), number of stored images and availability of screening views. RESULTS: A total of 42 normal fetuses and 66 cases of isolated CHD at birth were analyzed. Of the abnormal cases, 31 were missed and 35 were detected at the time of the clinical anatomy scan (sensitivity, 53%). Model sensitivity and specificity were 91% and 78%, respectively. Blinded human experts (n = 3) achieved mean ± SD sensitivity and specificity of 55 ± 10% (range, 47-67%) and 71 ± 13% (range, 57-83%), respectively. There was a statistically significant difference in model correctness according to expert-graded image quality (P = 0.03). The abnormal cases included 19 lesions that the model had not encountered during its training; the model's performance in these cases (16/19 correct) was not statistically significantly different from that for previously encountered lesions (P = 0.41). CONCLUSIONS: A previously trained DL algorithm had higher sensitivity than initial clinical assessment in detecting CHD in a cohort in which over 50% of CHD cases were initially missed clinically. Notably, the DL algorithm performed well on community-acquired images in a low-risk population, including lesions to which it had not been exposed previously. Furthermore, when both the model and blinded human experts had access to only stored images and not the full range of images available to a clinician during a live scan, the model outperformed the human experts. Together, these findings support the proposition that use of DL models can improve prenatal detection of CHD. © 2023 International Society of Ultrasound in Obstetrics and Gynecology.

Impact of extracardiac pathology on head growth in fetuses with congenital heart defect.

OBJECTIVE: Neurodevelopmental delay is frequently encountered in children with a congenital heart defect (CHD). Fetuses with major CHD have a smaller head circumference (HC), irrespective of altered cerebral flow or brain oxygenation. This cohort study compared head growth in cases with isolated vs those with non-isolated CHD to evaluate the effect of additional pathology on head size in these fetuses. METHOD: All CHD cases diagnosed prenatally in the period January 2002-July 2014 were selected from our regional registry, PRECOR. Cases of multiple pregnancy, and those affected by maternal diabetes, severe fetal structural brain anomalies or functional CHD were excluded. Subjects were divided into groups according to whether the CHD was isolated, and the non-isolated group was subdivided into three groups: cases with genetic anomaly, extracardiac malformation or placental pathology. In both isolated and non-isolated CHD groups, CHDs were also grouped according to their potential effect on aortic flow and oxygen saturation. Mean HC Z-scores at 20 weeks and increase or decrease (Δ) of HC Z-scores over the course of pregnancy were compared between isolated and non-isolated groups, using mixed linear regression models. RESULTS: Included were 916 cases of CHD diagnosed prenatally, of which 378 (41.3%) were non-isolated (37 with placental pathology, 217 with genetic anomaly and 124 with extracardiac malformation). At 20 weeks, non-isolated cases had significantly lower HC Z-scores than did isolated cases (Z-score = -0.70 vs -0.03; P < 0.001) and head growth over the course of pregnancy showed a larger decrease in this group (Δ HC Z-score = -0.03 vs -0.01 per week; P = 0.01). Cases with placental pathology had the lowest HC Z-score at 20 weeks (Z-score = -1.29) and the largest decrease in head growth (Δ HC Z-score = -0.06 per week). In CHD subjects with a genetic diagnosis (Z-score = -0.73; Δ HC Z-score = -0.04 per week) and in those with an extracardiac malformation (Z-score = -0.49; Δ HC Z-score = -0.02 per week), HC Z-scores were also lower compared with those in subjects with isolated CHD. CHDs that result in low oxygenation or flow to the brain were present more frequently in isolated than in non-isolated cases. CONCLUSIONS: Smaller HC in fetuses with CHD appears to be associated strongly with additional pathology. Placental pathology and genetic anomaly in particular seem to be important contributors to restricted head growth. This effect appears to be irrespective of altered hemodynamics caused by the CHD. Previously reported smaller HC in CHD should, in our opinion, be attributed to additional pathology. Neurodevelopment studies in infants with CHD should, therefore, always differentiate between isolated and non-isolated cases. © 2019 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of the International Society of Ultrasound in Obstetrics and Gynecology.

Why are congenital heart defects being missed?

OBJECTIVE: Congenital heart defects (CHD) are still missed frequently in prenatal screening programs, which can result in severe morbidity or even death. The aim of this study was to evaluate the quality of fetal heart images, obtained during the second-trimester standard anomaly scan (SAS) in cases of CHD, to explore factors associated with a missed prenatal diagnosis. METHODS: In this case-control study, all cases of a fetus born with isolated severe CHD in the Northwestern region of The Netherlands, between 2015 and 2016, were extracted from the PRECOR registry. Severe CHD was defined as need for surgical repair in the first year postpartum. Each cardiac view (four-chamber view (4CV), three-vessel (3V) view and left and right ventricular outflow tract (LVOT, RVOT) views) obtained during the SAS was scored for technical correctness on a scale of 0 to 5 by two fetal echocardiography experts, blinded to the diagnosis of CHD and whether it was detected prenatally. Quality parameters of the cardiac examination were compared between cases in which CHD was detected and those in which it was missed on the SAS. Regression analysis was used to assess the association of sonographer experience and of screening-center experience with the cardiac examination quality score. RESULTS: A total of 114 cases of isolated severe CHD at birth were analyzed, of which 58 (50.9%) were missed and 56 (49.1%) were detected on the SAS. The defects comprised transposition of the great arteries (17%), aortic coarctation (16%), tetralogy of Fallot (10%), atrioventricular septal defect (6%), aortic valve stenosis (5%), ventricular septal defect (18%) and other defects (28%). No differences were found in fetal position, obstetric history, maternal age or body mass index (BMI) or gestational age at examination between missed and detected cases. Ninety-two cases had available cardiac images from the SAS. Compared with the detected group, the missed group had significantly lower cardiac examination quality scores (adequate score (≥ 12) in 32% vs 64%; P = 0.002), rate of proper use of magnification (58% vs 84%; P = 0.01) and quality scores for each individual cardiac plane (4CV (2.7 vs 3.9; P < 0.001), 3V view (3.0 vs 3.8; P = 0.02), LVOT view (1.9 vs 3.3; P < 0.001) and RVOT view (1.9 vs 3.3; P < 0.001)). In 49% of missed cases, the lack of detection was due to poor adaptational skills resulting in inadequate images in which the CHD was not clearly visible; in 31%, the images showed an abnormality (mainly septal defects and aortic arch anomalies) which had not been recognized at the time of the scan; and, in 20%, the cardiac planes had been obtained properly but showed normal anatomy. Multivariate regression analysis showed that the volume of SAS performed per year by each sonographer was associated significantly with quality score of the cardiac examination. CONCLUSIONS: A lack of adaptational skills when performing the SAS, as opposed to circumstantial factors such as BMI or fetal position, appears to play an important role in failure to detect CHD prenatally. The quality of the cardiac views was inadequate significantly more often in undetected compared with detected cases. Despite adequate quality of the images, CHD was not recognized in 31% of cases. A high volume of SAS performed by each sonographer in a large ultrasound center contributes significantly to prenatal detection. In 20% of undetected cases, CHD was not visible even though the quality of the images was good. © 2019 Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Postnatal outcome of fetal isolated ventricular size disproportion in the absence of aortic coarctation.

OBJECTIVE: Cardiac ventricular size disproportion is a marker for aortic coarctation (CoA) in fetal life, but approximately 50% of fetuses do not have CoA after birth. The aim of this study was to evaluate the postnatal outcome of cases with fetal ventricular size disproportion in the absence of CoA after birth. METHODS: All cases with fetal isolated ventricular size disproportion diagnosed between 2002 and 2015 were extracted from a prenatal congenital heart defects regional registry. Cases were stratified according to presence or absence (non-CoA) of aortic arch anomalies after birth. Postnatal outcome of non-CoA cases was evaluated by assessing the presence of cardiac and other congenital malformations, genetic syndromes and other morbidity after birth. Non-CoA cases were further classified according to whether they had cardiovascular pathology requiring medication or intervention. RESULTS: Seventy-seven cases with fetal ventricular size disproportion were identified, of which 46 (60%) did not have CoA after birth. Of these, 35 did not require cardiovascular intervention or medication, whereas 11 did. Of the 46 non-CoA cases, six presented with clinical pulmonary hypertension requiring treatment after birth, cardiac defects were present in 24 cases and syndromic features were seen in four. Overall, 43% of all non-CoA children were still under surveillance at the end of the study period. CONCLUSIONS: The postnatal course of cases with fetal ventricular size disproportion is complicated by prenatally undetected congenital defects (46%) and pulmonary or transition problems (35%) in a significant number of cases that do not develop CoA. Proper monitoring of these cases is therefore warranted and it is advisable to incorporate the risks for additional morbidity and neonatal complications in prenatal counseling. Copyright © 2017 ISUOG. Published by John Wiley & Sons Ltd.