Abstract
AbstractBackgroundTransthoracic echocardiography (TTE) is the primary modality for diagnosing aortic valve stenosis (AVS), yet it requires skilled operators and can be resource-intensive. We developed and validated an artificial intelligence (AI)-based system for evaluating AVS that is effective in both resource-limited and advanced settings.MethodsWe created a dual-pathway AI system for AVS evaluation using a nationwide echocardiographic dataset (developmental dataset, n=8,427): 1) a deep learning (DL)-based AVS continuum assessment algorithm using limited 2D TTE videos, and 2) automating conventional AVS evaluation. We performed internal (internal test dataset [ITDS], n=841) and external validation (distinct hospital dataset [DHDS], n=1,696; temporally distinct dataset [TDDS], n=772) for diagnostic value across various stages of AVS and prognostic value for composite endpoints (cardiovascular death, heart failure, and aortic valve replacement)FindingsThe DL index for the AVS continuum (DLi-AVSc, range 0-100) increases with worsening AVS severity and demonstrated excellent discrimination for any AVS (AUC 0.91– 0.99), significant AVS (0.95–0.98), and severe AVS (0.97–0.99). DLi-AVSc was independent predictor for composite endpoint (adjusted hazard ratios 2.19, 1.64, and 1.61 per 10-point increase in ITDS, DHDS, and TDDS, respectively). Automatic measurement of conventional AVS parameters demonstrated excellent correlation with manual measurement, resulting in high accuracy for AVS staging (98.2% for ITDS, 81.0% for DHDS, and 96.8% for TDDS) and comparable prognostic value to manually-derived parameters.InterpretationThe AI-based system provides accurate and prognostically valuable AVS assessment, suitable for various clinical settings. Further validation studies are planned to confirm its effectiveness across diverse environments.FundingThis work was supported by a grant from the Institute of Information & communications Technology Planning & Evaluation (IITP) funded by the Korea government (Ministry of Science and ICT) (No.2022000972, Development of a Flexible Mobile Healthcare Software Platform Using 5G MEC); and the Medical AI Clinic Program through the NIPA funded by the MSIT. (Grant No.: H0904-24-1002). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.Research in ContextEvidence before this studyWe screened all English-based research articles in PubMed up to December 2023 using the keywords “artificial intelligence,” “echocardiography,” and “aortic valve stenosis.” While some studies have used artificial intelligence (AI) to evaluate aortic valve stenosis (AVS) in echocardiography, these efforts were typically focused on either predicting significant AVS or automating conventional measurements, not both. For instance, Holste G. et al. trained a deep learning model on 5,257 studies and validated it using two external datasets (4,226 and 3,072 studies), achieving high accuracy in detecting severe AVS (area under the receiver operating characteristic curve (AUC): 0.942–0.952). However, their model was limited to the parasternal long-axis view and did not provide conventional quantitative analysis. In contrast, Krishna H. et al. automated conventional AVS evaluation, demonstrating that AI could accurately measure AVS parameters like aortic valve maximal velocity, mean pressure gradient, and aortic valve area in 256 patients, comparable to human measurements, but did not perform qualitative assessment of AVS. Furthermore, no studies investigated the prognostic value of AI-based AVS assessment.Added value of this studyIn this study, we developed a comprehensive AI-based system to evaluate AVS through a dual pathway: 1) assessing AVS presence and severity by deriving a DL index for the AVS continuum (DLi-AVSc) from parasternal long and/or short axis videos only, and 2) automatically measuring AVS parameters and providing conventional quantitative AVS evaluation if additional images are available. The system was validated internally and in two independent external datasets, where DLi-AVSc increased with AVS severity and demonstrated excellent discrimination for any AVS (AUC 0.91–0.99), significant AVS (0.95– 0.98), and severe AVS (0.97–0.99). Additionally, DLi-AVSc independently predicted adverse cardiovascular events. The automatic measurement of conventional AVS parameters showed a strong correlation with manual measurement, resulting in high accuracy for AVS staging (98.2% for internal test set, 81.0%, and 96.8% for external test sets) and offered prognostic value comparable to manually-derived parameters.Implications of all the available evidenceAI-enhanced echocardiographic evaluation of AVS allows for accurate diagnosis of significant AVS and prediction of severity using only parasternal long or short axis views, typically obtained in the first step of echocardiographic evaluation. This capability can enhance AVS assessment in resource-limited settings and provide novices with guidance on when quantitative analysis is necessary. If additional views are acquired, the system automatically analyses them, enabling conventional quantitative evaluation, thereby saving time and effort while ensuring accurate assessment. Our study findings support the clinical implementation of AI-enhanced echocardiographic analysis.
Publisher
Cold Spring Harbor Laboratory