Artificial Intelligence (AI)-Facilitated Analysis of Single-Image Tissue Doppler Signal to Characterize Right...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Quantitative assessment of right ventricular (RV) function via transthoracic echocardiogram (TTE) typically uses tricuspid annular plane systolic excursion (TAPSE) and lateral tricuspid annulus peak systolic velocity (S').
• These measures may not capture the full spectrum of RV systolic function information available from complete cardiac cycle data.

Purpose of the Study:

• To automate the estimation of TAPSE and S' from tissue Doppler imaging (TDI) data.
• To develop an integrated model using full-cycle TDI waveforms and standard parameters to estimate RV systolic function and predict RV dysfunction.

Main Methods:

• Developed and validated an automated algorithm to extract TAPSE and S' from raw TDI data.
• Trained two classifier models: a tabular model (RVD_TABULAR) using TAPSE/S', age, and sex, and an integrated model (RVD_INTEGRATED) using TDI waveforms and tabular data.
• Validated models against cardiac magnetic resonance imaging (CMR) in 387 patients and in an external cohort with pulmonary hypertension (PH).

Main Results:

• The automated algorithm accurately estimated S' and TAPSE.
• The integrated model (RVD_INTEGRATED) demonstrated superior performance in predicting RV dysfunction (RVEF <45%) compared to the tabular model (AUROC: 0.768 vs. 0.71).
• In the PH cohort, the integrated model's predictions were significantly associated with event-free survival (p = 0.036).

Conclusions:

• A fully automated pipeline integrating digitized TDI waveforms and diverse features effectively classifies RV dysfunction.
• This approach offers a valuable tool for risk-stratifying patients, particularly those with pulmonary hypertension.