Single-View Echocardiographic Analysis for Left Ventricular Outflow Tract Obstruction Prediction in Hypertrophic...

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Accurate assessment of left ventricular outflow tract obstruction (LVOTO) is critical for managing hypertrophic cardiomyopathy (HCM).
• Traditional LVOTO assessment methods are often complex, resource-intensive, and infeasible in certain clinical settings.
• There is a need for simplified, accurate methods to predict severe LVOTO in HCM patients.

Purpose of the Study:

• To develop and validate a deep learning (DL) model for predicting severe LVOTO in HCM patients.
• To utilize only the parasternal long-axis (PLAX) view from transthoracic echocardiography (TTE) for LVOTO prediction.
• To create a DL index for LVOTO (DLi-LVOTO) as a predictive tool.

Main Methods:

• A DL model was trained on 1,007 PLAX TTE videos, capturing morphological and dynamic features.
• The model generated a DLi-LVOTO score (0-100) to predict severe LVOTO (pressure gradient ≥ 50mmHg).
• Performance was evaluated on an internal test dataset (n=87) and externally validated on two distinct datasets (n=1,334 and n=156).

Main Results:

• The DL model demonstrated high accuracy in predicting severe LVOTO, with AUROCs of 0.97 (internal) and 0.93 (external).
• High specificity and negative predictive values were achieved at specific DLi-LVOTO thresholds in both validation sets.
• DLi-LVOTO significantly decreased post-treatment (surgical myectomy, Mavacamten), correlating with reduced peak pressure gradients (p<0.001).

Conclusions:

• A DL-based approach using only the PLAX TTE view can effectively predict severe LVOTO in HCM.
• This method serves as a valuable complementary tool, especially when Doppler assessment is challenging or unavailable.
• The DLi-LVOTO shows promise for monitoring treatment response in HCM patients with LVOTO.