CardiacField

CardiacField reconstructs a 3D cardiac volume from sequential multi-view two-dimensional echocardiography images and automatically estimates left ventricular and right ventricular ejection fraction.

Key Features:

• 3D Volume Reconstruction: Uses an implicit neural representation network to reconstruct 3D cardiac volume from sequential multi-view 2DE images.
• Automated Segmentation: Automatically segments left ventricular and right ventricular areas from the reconstructed volume.
• Ejection Fraction Calculation: Calculates volume sizes and ejection fraction values for both left and right ventricles.
• Apical Ring Scan Protocol: Employs a straightforward apical ring scan acquisition using a standard 2DE probe.
• Rapid Processing: Generates a 3D heart from 2D echocardiograms with under 2 minute processing time.

Scientific Applications:

• Left Ventricular Function Assessment: Estimates left ventricular ejection fraction with a mean absolute error of 2.48% compared to expert readings.
• Right Ventricular Function Assessment: Estimates right ventricular ejection fraction with a mean absolute error of 2.65% compared to expert readings.
• Comparison with Deep Learning Models: Evaluates performance against two-dimensional video-based deep learning models for ejection fraction estimation.
• Cross-Platform Validation: Tests usability across various ultrasound machines by novice users.

Methodology:

CardiacField uses an implicit neural representation network to reconstruct a 3D cardiac volume from sequential multi-view 2DE images, followed by automatic segmentation of LV and RV areas to calculate volume sizes and EF values.