Model predictive filtering for improved temporal resolution in MRI temperature imaging.

Area of Science:

• Medical Imaging
• Biophysics
• Computational Modeling

Background:

• Accurate temperature monitoring is crucial for therapeutic procedures like high-intensity focused ultrasound (HIFU).
• Current MR thermometry methods face challenges with speed and data acquisition limitations.

Purpose of the Study:

• To introduce and evaluate a novel method, model predictive filtering (MPF), for reconstructing MRI temperature maps from undersampled data.
• To assess the accuracy and real-time capabilities of MPF compared to the proton resonance frequency shift (PRFS) method.

Main Methods:

• Model predictive filtering (MPF) was developed, integrating thermal model predictions with undersampled k-space data.
• MPF was implemented and tested using retrospectively and actually undersampled 2D gradient echo (GRE) sequences and actually undersampled 3D GRE sequences.
• The method was validated against the PRFS method in 39 HIFU heating experiments monitored by MRI.

Main Results:

• MPF demonstrated high accuracy in 2D implementations, with average errors below +/-0.8°C and RMSE below 0.35°C.
• The 3D implementation showed slightly larger errors, with a mean error of -1.4°C and RMSE of 0.61°C for the hottest voxels.
• The method successfully generated temperature maps in near real-time from undersampled MRI data.

Conclusions:

• Model predictive filtering offers a promising approach for rapid and accurate MRI-based temperature mapping.
• This method has the potential to improve the safety and efficacy of thermal therapies like HIFU.
• MPF's ability to utilize undersampled data enhances its applicability in time-sensitive clinical scenarios.