One view to rule them all? Applying Artificial Intelligence to Assess valvular disease and ventricular function using a Single Apical 4-Chamber echocardiographic View

Introduction:

The role of artificial intelligence (AI) in echocardiography is rapidly expanding, with the promise to improve diagnostic accuracy and efficiency. We hypothesized that a Deep Neural Network (DNN) could be trained to identify and assess significant cardiac dysfunctions using only a single apical 4-chamber echocardiographic view rather than a comprehensive multi-view standard transthoracic echocardiogram.

Material and method:

Echocardiographic reports and corresponding images from 121,767 unique patients linked to clinical data (2007-2022) were reviewed by a designated DNN. A validation cohort, containing only apical 4-chamber view clips, was introduced to the DNN and analyzed to assess for the presence of a significant (mild to moderate up to severe) mitral regurgitation (MR), tricuspid regurgitation (TR), right ventricular failure (RVF), and to assess left ventricular ejection fraction.

Results and discussion:

The model achieved an Area under the curve (AUC) of 0.88 for MR, with a sensitivity of 0.74 and specificity of 0.84 (N = 20,313). For TR, it yielded an AUC of 0.91, with a sensitivity of 0.76 and specificity of 0.88 (N = 21,742). In assessing RVF, the DNN reached an AUC of 0.94, with a sensitivity of 0.75 and specificity of 0.93 (N = 6,010). For the assessment of left ventricular ejection fraction (LVEF), the root mean square error (RMSE) was 4.78, and the mean absolute error (MAE) was 3.34 (N = 23,350).

Conclusion:

Applying artificial intelligence to assess a single apical 4-chamber echocardiographic view can provide estimates of the presence of significant valvular diseases, RVF and predict LVEF with high sensitivity and specificity. These results suggest that AI-driven approaches could be applied to more widespread and efficient diagnostic processes.