Abstract 4365486: Artificial Intelligence Assisted Characterization of HFpEF in Patients with Indeterminate Diastolic Function

Abstract

Background: Echocardiographic evaluation of diastolic dysfunction is often indeterminate and thus poses a challenge in assisting with diagnosing heart failure with preserved ejection fraction (HFpEF). To overcome this, scoring systems such as H2FPEF utilize a combination of clinical history and echocardiographic data to help make the diagnosis. Recently, an artificial intelligence (AI) algorithm has been validated to detect HFpEF using a single apical 4-chamber view. Less is known about how this algorithm performs in patients with indeterminate diastolic function (IDF) adjudicated by ASE guidelines. The purpose of this study was to characterize patients with IDF on echocardiography who had been detected as having HFpEF by the AI algorithm. Methods: We prospectively examined 173 consecutive patients who underwent a clinical echocardiogram and screened positive for HFpEF using the Ultromics EchoGo HFpEF algorithm (Oxford, UK). Of these, 95 had indeterminate diastolic function by ASE guidelines. A detailed review of these patients’ charts was performed to characterize this population further and calculate individual H2FPEF scores. Results: Of the individuals with IDF and a positive HFpEF diagnosis by AI, the most common co-morbidities included hypertension (90%), coronary artery disease (75%), obesity (66%), and atrial fibrillation/flutter (66%). The H2FPEF score was calculated in 69 of these patients, of which 65% had a score of ≥6. Among the cohort of patients with IDF with a positive HFpEF screen per the AI algorithm, 42% had no documented history of HFpEF in their charts. Conclusion: In individuals with indeterminate diastolic function, per conventional echocardiographic guidelines, the EchoGo AI algorithm proves to be an effective tool in helping identify HFpEF. Incorporation of AI in conjunction with echocardiographic reporting may help with earlier identification of individuals with HFpEF, and could potentially lead to earlier interventions.