Fully Automatic AI-Based Quantification of LV Mass in Echocardiography: A Multimodality Validation

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate left ventricular (LV) myocardial mass assessment is crucial for treatment decisions.
• Echocardiography measurements are limited by operator variability.
• Artificial intelligence (AI) offers potential for improved precision and reproducibility in LV mass assessment.

Purpose of the Study:

• To compare the accuracy of AI-based versus expert manual echocardiography for LV mass measurements.
• To validate AI echocardiography against computed tomography (CT) as a reference standard.
• To evaluate the real-world test-retest reliability of AI echocardiography.

Main Methods:

• LV mass, interventricular septal diameter (IVSD), posterior wall diameter (PWD), and end-diastolic volume were analyzed in 218 patients.
• LV mass was assessed using linear and hybrid methods.
• AI and expert echocardiographic measurements were compared to CT using intraclass correlation coefficient (ICC) and mean absolute percentage error (MAPE).
• Test-retest reliability was assessed using coefficient of variation.

Main Results:

• AI echocardiography demonstrated strongest agreement with CT for LV mass using the hybrid method (ICC: 0.76).
• AI reduced measurement error by approximately 20% compared to expert evaluation.
• AI echocardiography showed higher diagnostic accuracy for increased LV mass (AUC: 0.84 for hybrid method) and superior test-retest reliability (CV: 7.8%).

Conclusions:

• AI echocardiography offers superior accuracy, diagnostic performance, and reproducibility for LV mass assessment compared to expert evaluation.
• AI-based methods enhance standardization in cardiac imaging.
• Clinical integration of AI echocardiography is supported to improve LV mass assessment.