Artificial intelligence implementation in automated heart chambers quantification during pharmacological stress...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Stress echocardiography (SE) is crucial for diagnosing coronary artery disease.
• Manual volumetric analysis in SE is time-consuming and prone to variability.
• Automated methods offer potential to enhance SE efficiency and accuracy.

Purpose of the Study:

• To evaluate the performance of a machine learning (ML) algorithm for automated volumetric chamber analysis during pharmacological SE.
• To compare ML-derived measurements with expert manual tracings for left and right heart chambers.
• To assess the impact of image quality on ML algorithm performance.

Main Methods:

• A multi-center study involving 240 participants undergoing pharmacological SE.
• Acquisition of SE imaging data during rest and stress phases.
• Comparison of ML-derived volumetric measurements against manual tracings by expert cardiologists.
• Analysis of image quality's influence on ML performance.

Main Results:

• The ML algorithm showed strong agreement with expert measurements for left ventricular (LV) and left atrial (LA) volumes (ICCs 0.84-0.93).
• Moderate agreement was observed for right atrial (RA) and right ventricular (RV) areas, with better results at rest.
• Poor image quality significantly reduced the diagnostic yield of the ML algorithm.

Conclusions:

• AI-driven volumetric analysis is a reliable tool for quantifying left-sided heart chambers in pharmacological SE.
• High-quality imaging and standardized protocols are essential for accurate right-sided chamber analysis.
• Integrating AI into SE workflows can improve efficiency and clinical decision-making.