Artificial Intelligence for Risk Stratification of Acute Pulmonary Embolism: Perspectives on Clinical Needs, Expanding Toolkit, and Pathways Forward

Abstract

Despite a significant number of innovations for management of acute pulmonary embolism (PE) over the past decade, PE-related mortality has not decreased as expected. Significant heterogeneity in PE presentations and limitations in contemporary risk stratification approaches are hypothesized to be important drivers of poorer than expected outcomes. Recently, artificial intelligence (AI) has gained attention in cardiovascular medicine, notably in wearable technology, electrocardiography, and cardiovascular imaging. The utility of AI has been studied in PE diagnosis and risk stratification, especially in hospitalized patients and has the potential to predict presence of PE based on electrocardiography and clinical risk factors, decrease time to diagnosis, and improve characterization of PE as acute versus chronic. However, AI systems do not appear to have better accuracy in identification of PE compared with radiologists. Additionally, whether utilization of AI in diagnosis and management of PE will improve clinician workflow and patient outcomes remains unknown. In this review, we critically appraise the literature on AI-based strategies to diagnose and refine risk stratification of acute PE and discuss how integration of AI may move the field of PE forward with the universal goal of improving short- and long-term PE-related outcomes.