Computational Model To Assist Surgical Decision-Making In Borderline Left Ventricles
Isao Anzai, MD1, Yurui Chen2, Justin Tran, PhD3, Emile Bacha, MD1, Vijay Vedula, PhD2, David Kalfa, MD, PhD1.
1Columbia University Irving Medical Center, New York Presbyterian Hospital, New York, NY, USA, 2Columbia University, New York, NY, USA, 3California State University, Fullerton, Fullerton, CA, USA.
Borderline left ventricle (BLV) is a congenital heart disease in which neonates are born with an undersized left ventricle. Decision-making between biventricular repair (BiVR) and single ventricle palliation procedure (SVPP) for BLV patients remains difficult and subjective.
We developed a closed-loop lumped parameter network (LPN) model of the circulatory system for BLV patients. An automatic tuning framework personalizes the LPN model parameters to match patient characteristics and available echocardiographic and cath data. The preoperative LPN was then modified to virtually perform BiVR and Norwood (1st stage SVPP) procedures for each patient. Computer simulations were performed on 10 BLV patients for each surgical procedure to compare the hemodynamic outcomes.
In the patient subset who clinically underwent Norwood, performing a virtual BiVR resulted in increased mean pulmonary artery pressure (mPAP, 32±4.7 vs. 16±3.4 mmHg), mean left atrial pressure (mLAP, 18±1.2 vs. 5±1.0 mmHg), and LV end-diastolic pressure (LVEDP, 16±3 vs. 5±2 mmHg). On the other hand, performing a virtual Norwood on the patient subset who underwent BiVR led to nominal changes in pressures but a high pulmonary to systemic flow ratio (Qp/Qs) due to high systemic vascular resistance.
Our personalized predictive computer model enables surgeons to perform virtual surgeries and predict hemodynamic quantities for BLV patients using clinical, echocardiographic, and cath data as input. On a small initial cohort, our model predictions corroborate the validity of the clinically performed procedure. Sensitivity analysis needs to be performed and the results should be validated on a larger prospective cohort.
Back to 2022 Abstracts