A predictive signal model for dynamic cardiac magnetic resonance imaging

Area of Science:

• Medical Imaging
• Cardiovascular Imaging
• Biomedical Engineering

Background:

• Dynamic cardiac MRI aims for real-time imaging to reveal temporal disease signatures.
• Compressed sensing (CS) and artificial intelligence (AI) have advanced cardiac MRI acceleration.
• Current real-time cardiac MRI methods face limitations in reconstruction quality and robustness.

Purpose of the Study:

• To develop a novel predictive signal model for adaptive temporal cardiac dynamics prediction.
• To improve reconstruction fidelity in dynamic cardiac MRI, particularly for irregular cardiac rhythms.
• To establish a versatile foundation for robust dynamic cardiac MRI.

Main Methods:

• A novel predictive signal model utilizing a priori statistics to adaptively predict temporal cardiac dynamics.
• Training the model on a small dataset from the same patient for personalized adaptation.
• Evaluating the model with simulated irregular cardiac dynamics and prospectively undersampled clinical data.

Main Results:

• The predictive model demonstrated improved reconstruction quality in both Kalman filter and compressed sensing (CS) frameworks.
• Robust performance was achieved even with simulated and real irregular cardiac rhythms.
• The model proved effective across various undersampling patterns, including Cartesian, radial, and spiral.

Conclusions:

• The proposed predictive signal model significantly enhances reconstruction fidelity in dynamic cardiac MRI.
• This approach ensures robust dynamic cardiac MRI, even in the presence of irregular cardiac rhythms.
• The model serves as a versatile foundation for future advancements in real-time cardiac imaging.