Fast temperature estimation from undersampled k-space with fully-sampled center for MR guided microwave ablation

Area of Science:

• Medical Physics
• Magnetic Resonance Imaging
• Interventional Radiology

Background:

• Real-time temperature monitoring is crucial for effective thermal ablation.
• Proton Resonance Frequency (PRF) shift thermometry is a key MRI technique.
• Accelerating PRF-shift imaging is essential for clinical translation.

Purpose of the Study:

• To accelerate MR temperature imaging using PRF shift for real-time monitoring.
• To develop an algorithm for estimating temperature from undersampled k-space data.
• To validate the method's accuracy and speed in various experimental conditions.

Main Methods:

• A hybrid multi-baseline, referenceless treatment image model was employed.
• Temperature was estimated from undersampled k-space with a fully sampled center.
• Low-resolution images from k-space centers informed hybrid model parameters.
• Registration corrected for subject displacement.
• Phantom and in-vivo experiments were conducted, comparing with SPIRiT and conventional methods.

Main Results:

• The proposed method demonstrated lower Root Mean Square Errors (RMSEs) compared to conventional methods in phantom studies.
• Motion simulations confirmed the method's robustness to subject displacement.
• Accurate temperature maps (RMSE < 0.4°C) were reconstructed with reduced computation time (3.5x acceleration).
• In-vivo experiments showed feasibility with 4.3x acceleration in the presence of motion and complex anatomy, reducing computation time up to 10-fold.

Conclusions:

• The developed method significantly accelerates PRF-shift MR thermometry.
• It yields more accurate temperature maps in the presence of motion.
• The reduced computation time makes real-time MR-guided microwave ablation feasible.