Referenceless magnetic resonance temperature imaging using Gaussian process modeling | JoVE Visualize

Area of Science:

Medical Imaging
Biophysics
Therapeutic Technologies

Background:

Water proton resonance frequency shift (PRFS)-based MR temperature imaging is crucial for monitoring tissue temperature during MR-guided thermal therapies.
PRFS techniques are susceptible to artifacts from tissue motion, susceptibility changes, magnetic field drift, and applicator interference.
Existing methods often rely on reference data, which can be compromised by these artifacts.

Purpose of the Study:

To investigate a referenceless Gaussian process modeling (GPM) approach for estimating PRFS to mitigate background field changes.
To develop a method that provides probabilistic predictions and uncertainty estimates for temperature mapping.
To improve the accuracy and reliability of MR thermometry in the presence of common artifacts.

Main Methods:

Gaussian process modeling (GPM) was used to estimate the covariance between spatial position and MR phase measurements.
The GPM model extrapolated background phase values from nonheated voxels to predict phase in heated regions.
The method was validated in ex vivo rabbit liver tissue during focused ultrasound heating and in vivo during laser interstitial thermal therapy in human brain and liver.

Main Results:

Referenceless GPM MR thermometry achieved a root-mean-square (RMS) error of <0.8°C during ex vivo focused ultrasound heating, outperforming artifact-affected reference-based methods.
Nonheated surrounding tissues showed minimal temperature deviation (<0.5°C) from artifact-free measurements.
In vivo laser ablations demonstrated that GPM-estimated temperatures and damaged regions were within 95% confidence intervals.

Conclusions:

A novel referenceless PRFS temperature imaging method using GPM was developed for accurate background phase extrapolation.
The technique provides reliable temperature estimates despite background phase variations and artifacts.
This GPM-based approach shows significant promise for in vivo thermal therapy monitoring in brain and liver applications.