Search

Background: The objective of this IRB-approved retrospective monocentric study was to identify risk factors for mortality after surgery for congenital heart defects (CHDs) in pediatric patients using machine learning (ML). CHD belongs to the most common congenital malformations, and remains the leading mortality cause from birth defects. Methods: The most recent available hospital encounter for each patient with an age n = 73 events]). All models were then applied to predict the outcome in an independent holdout test dataset (40% of the cohort). Results: RSF and XGB achieved average C-indices of 0.85 (±0.01) and 0.79 (±0.03), respectively. Feature importance was assessed with ‘SHapley Additive exPlanations’ (SHAP) and ‘Time-dependent explanations of machine learning survival models’ (SurvSHAP(t)), both of which revealed high importance of the maximum values of serum creatinine observed within 72 h post-surgery for both ML methods. Conclusions: ML methods, along with model explainability tools, can reveal interesting insights into mortality risk after surgery for CHD. The proposed analytical workflow can serve as a blueprint for translating the analysis into a federated setting that builds upon the infrastructure of the German Medical Informatics Initiative.

(1) Purpose: to compare right ventricular (RV) functional parameters in children with surgically repaired congenital heart disease (CHD) using single/double breath hold (BH) and free-breathing (FB) real-time compressed sensing (CS) cine cardiac magnetic resonance (cMRI) with standard retrospective segmented multi breath hold (RMB) cine cMRI. (2) Methods: Twenty patients with CHD underwent BH and FB, as well as RMB cine cMRI, at 3T to obtain a stack of continuous axial images of the RV. Two radiologists independently performed qualitative analysis of the image quality (rated on a 5-point scale; 1 = non-diagnostic to 5 = excellent) and quantitative analysis of the RV volume measurements. (3) Results: The best image quality was provided by RMB (4.5; range 2–5) compared to BH (3.9; range 3–5; p = 0.04) and FB (3.6; range 3–5; p < 0.01). The RV functional parameters were comparable among BH, FB, and RMB with a difference of less than 5%. The scan times for BH (44 ± 38 s, p < 0.01) and FB (24 ± 7 s, p < 0.01) were significantly reduced compared to for RMB (261 ± 68 s). (4) Conclusions: CS-FB and CS-BH real-time cine cMRI in children with CHD provides diagnostic image quality with excellent accuracy for measuring RV function with a significantly reduced scan time compared to RMB.

For therapeutic decisions regarding uni- or biventricular surgical repair in congenital heart disease (CHD), left ventricular mass (LVM) is an important factor. The aim of this retrospective study was to determine the LVM of infants with CHD in thoracic computed tomography angiographies (CTAs) and to evaluate its usefulness as a prognostic parameter, with special attention paid to hypoplastic left heart (HLH) patients. Manual segmentation of the left ventricular endo- and epicardial volumes was performed in CTAs of 132 infants. LVMs were determined from these volumes and normalized to body surface area. LVMs of patients with different types of CHD were compared to each other using analyses of variances (ANOVA). An LVM cutoff for discrimination between uni- and biventricular repair was determined using receiver operating characteristics. Survival rates were calculated using Kaplan–Meier statistics. Patients with a clinical diagnosis of an HLH had significantly lower mean LVM (21.88 g/m2) compared to patients without applicable disease (50.22 g/m2; p < 0.0001) and compared to other CHDs, including persistent truncus arteriosus, left ventricular outflow tract obstruction, transposition of the great arteries, pulmonary artery stenosis or atresia, and double-outlet right ventricle (all, p < 0.05). The LVM cutoff for uni- vs. biventricular surgery was 33.9 g/m2 (sensitivity: 82.3%; specificity: 73.7%; PPV: 94.9%). In a subanalysis of HLH patients, a sensitivity of 50.0%, specificity of 100%, PPV of 100%, and NPV of 83.3% was determined. Patient survival was not significantly different between the surgical approaches or between patients with LVM above or below the cutoff. LVM can be measured in chest CTA of newborns with CHD and can be used as a prognostic factor.