Abstract 15026: The Use of Artificial Intelligence Guidance for Rheumatic Heart Disease Screening by Novices

Introduction: A novel technology utilizing artificial intelligence (AI) to provide real-time image-acquisition guidance, enabling novices to obtain diagnostic echocardiographic (echo) images holds promise to expand the reach of echo screening for rheumatic heart disease (RHD). We evaluated the ability of non-experts to obtain diagnostic quality images in patients with RHD using AI guidance (Caption Health) with color Doppler. Methods: This study was performed in Kampala, Uganda at the Uganda Heart Institute. Nurses and nursing students with no prior echo experience underwent one-day training in the use of AI guidance, which included a 7-view screening protocol with 2D and color Doppler images. Following training, all participants scanned 8-10 patients using AI guidance, half RHD and half normal. The same patients were scanned by two expert sonographers without the use of AI guidance. Each echo was evaluated by a panel of 4 expert echocardiographers blinded to the identity of the scanner (novice/exert) and diagnosis of the patient to provide quality assessment including (1) diagnostic quality to determine presence/absence of RHD, (2) more detailed valvular assessment, and (3) ACEP score 1-5 for each view. Results: Thirty-six novice scanners scanned a total of 50 patients, 25 with RHD and 25 controls. A total of 462 echocardiogram studies were obtained, 351 obtained by non-experts using AI guidance and 111 obtained by expert sonographers without AI guidance. Preliminary analysis of 394 interpretations showed that 95% of non-expert studies were of diagnostic quality to assess for RHD and mitral valve disease (vs 100% by experts, p=0.03), but only 56% of non-expert studies were of diagnostic quality to assess aortic valve disease (vs 97% by experts, p<0.0001). The ACEP scores of non-expert images were highest in the parasternal long axis images (mean 3.62, 84% ≥ 3) compared to lower scores for apical 4 (mean 3.28, 73% ≥3), and apical 5 images (mean 2.52, 46% ≥ 3). Conclusions: AI guidance with color Doppler is feasible to enable rheumatic heart disease screening by non-experts, performing significantly better for assessment of the mitral than aortic valve. Further refinement is needed to optimize acquisition of color Doppler apical views.