In a pilot study, cardiologists agreed with advanced practice providers' interpretations of major cardiac findings in more than 96% of point-of-care ultrasound examinations, and examinations performed with real-time artificial intelligence guidance required fewer cardiologist corrections than those performed without it, according to findings from the AISAP-INOVA study.
Results from this prospective observational study, conducted by Andrew H. Nguyen, DO, of Inova Schar Heart and Vascular in Falls Church, Virginia, and colleagues, were presented as a President's Circle poster at the American Society of Echocardiography (ASE) 2026 Scientific Sessions and published in the Journal of the American Society of Echocardiography.
Adult patients underwent clinically indicated cardiac point-of-care ultrasound with or without real-time artificial intelligence (AI) guidance in an observation unit. A total of 100 examinations performed by advanced practice providers following didactic training were included. Of the examinations, 64 were performed with real-time AI guidance using the AISAP AI platform, and 36 were performed without AI assistance.
Researchers evaluated advanced practice provider interpretations, blinded cardiologist overreads, and associated clinical management decisions. A complete transthoracic echocardiogram was obtained when point-of-care ultrasound identified high-risk findings.
Baseline patient demographics and clinical profiles were similar between the AI-guided and non–AI-guided groups. Cardiologists agreed with advanced practice provider interpretations on major findings in more than 96% of examinations.
AI-guided examinations were associated with improved acquisition completeness and fewer cardiologist corrections. Cardiologists corrected 27% of AI-guided examinations compared with 47% of examinations performed without AI assistance. Researchers reported that the difference was driven by fewer missed or underestimated left ventricular and valvular abnormalities.
In an interview, Dr. Nguyen said the cardiologist overread process distinguished between significant corrections and substantive additions. Significant corrections included changes that crossed clinically meaningful grading thresholds, such as a point-of-care ultrasound estimate of preserved ejection fraction being revised to a substantially lower value or mitral regurgitation being reclassified from mild to severe.
He said right ventricular size and function remained among the most commonly corrected findings in both groups and noted that AI-supported assessment of right-sided function still needed improvement. In contrast, he said mitral regurgitation appeared to be better supported with AI guidance in this small pilot, with corrections for mitral regurgitation occurring more often in the non–AI-guided group.
Scan duration was 13.7 minutes with AI guidance compared with 16.9 minutes without AI guidance.
Point-of-care ultrasound findings were associated with management changes in 57% of patients, most commonly involving adjustments in fluid or diuretic therapy and formal cardiology consultation. Researchers also reported that 56% of patients were discharged without requiring inpatient transthoracic echocardiography, while 16% were discharged with outpatient transthoracic echocardiography referrals.
Dr. Nguyen cautioned that the study was hypothesis-generating and that the discharge findings should be interpreted cautiously and require further validation. He said the investigators did not have longitudinal follow-up data for patients discharged without an inpatient echocardiogram and identified larger patient volume, randomized and blinded study design, and longer-term follow-up as next steps.
According to the researchers, AI-guided cardiac point-of-care ultrasound performed by advanced practice providers was associated with high agreement with cardiologist review and supported clinical management decisions in this pilot study. The findings suggest that real-time AI guidance may help standardize acquisition and support noncardiologist decision-making, although the discharge-without-inpatient-echocardiography findings require validation in larger randomized, blinded studies with longitudinal follow-up.
No disclosure information was provided.