23: Artificial Intelligence–Driven Monitoring for Early Detection and Management of Harlequin Syndrome During Veno-Arterial ECMO Support

Background: Harlequin syndrome, also referred to as differential hypoxemia or North South syndrome, is a critical yet underrecognized complication of venoarterial extracorporeal membrane oxygenation and hybrid configurations such as ECpella. It occurs when recovering native cardiac output competes with retrograde extracorporeal flow, leading to differential oxygenation between the upper and lower body. Current detection methods are largely operator dependent and often delayed, increasing the risk of cerebral and myocardial hypoxia. Objective: To propose an artificial intelligence integrated multimodal monitoring platform for early detection and proactive management of Harlequin syndrome using continuous physiologic and imaging data analysis. Methods: This conceptual framework integrates six multimodal data streams, including continuous right radial pulse oximetry, continuous left radial pulse oximetry, continuous cerebral near infrared spectroscopy, intermittent arterial blood gas derived partial pressure of oxygen measurements, dynamic respiratory system mechanics, and artificial intelligence assisted chest radiograph analysis. The artificial intelligence model continuously evaluates trends and interrelationships across these inputs to detect early oxygenation mismatch. Intermittent arterial blood gas values serve as calibration references to maintain accuracy over time. Outputs are displayed on an interactive extracorporeal membrane oxygenation dashboard that provides early alerts and actionable management suggestions, including ventilator adjustments, extracorporeal flow optimization, or cannulation strategy modification. Expected Outcomes: The proposed system is expected to detect right radial oxygen desaturation and cerebral near infrared spectroscopy changes earlier than conventional clinician recognition. Multimodal signal validation is anticipated to provide high sensitivity with a low false alert rate. Conclusion: This conceptual model introduces a predictive and data driven approach for early identification and management of Harlequin syndrome. By integrating physiologic monitoring with imaging analytics, the platform has the potential to enhance patient safety, optimize extracorporeal membrane oxygenation performance, and support the development of future intelligent closed loop extracorporeal systems.