A deep learning-based pipeline for large-scale echocardiography data curation and measurements.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Echocardiographic data is a valuable but underutilized resource for cardiac research.
• Manual measurements are time-consuming and prone to variability.
• Data curation and quality control are essential for real-world data analysis.

Purpose of the Study:

• To develop an automated pipeline for data curation of a large echocardiogram database.
• To evaluate automated measurements of left ventricular ejection fraction (LVEF) and left atrial volume index (LAVI) using deep learning.

Main Methods:

• Developed an automated pipeline for curating 14,326 echocardiogram exams.
• Applied deep learning for automated measurements of LVEF and LAVI.
• Validated automated measurements against manual measurements using Bland-Altman analysis.

Main Results:

• Automated LVEF measurements showed a bias [SD] of -1.8% (7.6%).
• Automated LAVI measurements showed a bias [SD] of 3.3 mL/m² (8.1 mL/m²).
• The pipeline achieved 79% high-confidence measurements in a large dataset without clinical data.

Conclusions:

• Deep learning-based automated measurements show potential for data mining in large echocardiogram databases.
• The automated pipeline facilitates rapid, consistent, and reliable cardiac measurements.
• This approach can advance cardiac imaging research and diagnostics.