Editors' ChoiceHeart Transplantation

Predicting donor heart function in a heartbeat

See allHide authors and affiliations

Science Translational Medicine  08 Apr 2020:
Vol. 12, Issue 538, eabb5667
DOI: 10.1126/scitranslmed.abb5667

Abstract

Machine learning methods can assess contractility of donor hearts maintained via ex vivo perfusion to predict posttransplant cardiac function.

Heart transplantation is potentially curative therapy for patients with advanced cardiac disease but is profoundly limited by scarcity of donor organs. At present, only 35% of hearts from donors with brain death are used for transplantation, although many rejected organs are subsequently found to be histologically normal. These findings motivate concerns that current evaluation techniques may be inadequate and result in discarding potentially life-saving organs. Ex situ heart perfusion (ESHP) is an emerging technique that may hold the key to overcoming donor organ shortage. ESHP provides normothermic perfusion of isolated donor hearts, maintaining viability of the hearts for an extended period ex vivo to assess organ function. However, determination of transplant suitability for a heart maintained by ESHP remains an open question. By applying machine learning methods and advanced physiological measurements to predict posttransplant function for ESHP-supported hearts, Xiao and colleagues recently described an approach that may provide an answer.

Left ventricular end-systolic elastance (Ees) is a metric of cardiac contractility classically determined by varying loading conditions on the left ventricle and may be an ideal measure to assess organ suitability for transplant. In the setting of ESHP, however, measuring Ees requires clamping the left atrial inflow cannula, thereby risking injury to the donor organ. To address this concern, Xiao and colleagues developed a strategy that accurately determines Ees from a single, steady left ventricular pressure-volume (PV) loop measurement. Using a support vector machine algorithm applied to measurements obtained from the PV loop, Xiao et al. estimated the Ees volume axis intercept point (V0), which they used with the measured left ventricular pressure and volume at end systole to determine Ees. Deploying a porcine model of ESHP, the authors obtained 144 PV loop measurements from 27 porcine hearts which they divided into training and validation data sets. They validated their estimated V0 values against measured values obtained by the standard cannula occlusion technique. Further validation of the overall approach demonstrated that Ees values obtained during ESHP supported predicted posttransplant cardiac function in six porcine heart transplants.

The innovative approach described by Xiao and colleagues is likely just the beginning for machine learning methods to provide quantitative metrics of cardiac function. Applying advanced methodologies to extract vital cardiac functional parameters may unlock the potential of mechanical circulatory support technologies and reveal how to optimally titrate support for patients in cardiogenic shock. Xiao and colleagues have shown that it only takes a heartbeat to determine cardiac function.

Highlighted Article

View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article