Personalized Computational Modeling for Predicting Hemodynamics in Borderline Left Ventricles

The decision to perform a biventricular repair (BiVR) or single ventricle palliation (SVP) in patients born with a borderline left ventricle (BLV) remains challenging and subjective. BLV patients are neonates born with an underdeveloped left ventricle but not severe enough to be classified as hypoplastic left heart syndrome (HLHS).

A BiVR is attractive because it restores two functioning ventricles, though it is technically challenging to perform on neonates. In contrast, SVP involves multiple staged open-heart procedures resulting in alternative physiology

with a single pumping chamber. HLHS patients undergoing SVP have high interstage mortality and poor longterm outcomes, with only 70% reaching adulthood. Current clinical guidelines are primarily governed by morphological measurements and status quo hemodynamic data that have mixed correlations with operative

outcomes. Alternative palliative procedures were developed for specific BLV variants that delay the decisionmaking between BiVR and SVP and allow the left heart structures to grow before further interventions, such as the hybrid approach and staged left ventricular rehabilitation (SLVR). However, about two-thirds of these patients

do not respond favorably to these palliative procedures, and it is unclear if a BiVR is feasible in these patients even at a later stage. There is an unmet clinical need to identify the BLV patients favorable for BiVR and perform risk stratification based on predicted postoperative outcomes. We aim to address this critical need by applying

personalized computational modeling to predict and compare the immediate postoperative outcomes of virtual BiVR and Norwood and choose a procedure that yields the most favorable hemodynamics. Our modeling framework relies on a closed-loop lumped parameter network (LPN) that is elegant and enables prediction in a

clinically adoptable timeframe for neonates. We demonstrated the model’s utility in a small patient cohort selected retrospectively and found that our model predictions support the clinical choice of surgery. Here, we aim to enhance the robustness of the modeling pipeline by (a) using the patients’ ultrasound data as input to the

model, (b) automatically estimating the personalized LPN model parameters using a robust statistical framework, (c) performing uncertainty quantification (UQ) to measure confidence intervals on the predicted hemodynamic variables, and (d) demonstrating statistical significance on a larger cohort. We will apply our computational

framework to assess the viability of performing a BiVR after initial hybrid palliation or SLVR. In Aim 1, we will develop a robust automatic tuning framework to estimate the LPN model parameters matching a BLV patient’s preoperative echo data. In Aim 2, we will perform UQ analysis to estimate the confidence in the predicted

postoperative hemodynamics. In Aim 3, we will evaluate the feasibility of performing a BiVR after an initial palliative procedure. Ultimately, this simulation-guided treatment planning will mitigate the risk of morbidity and mortality and reinterventions in this critical population. Project Summary/Abstract