Stochastic Data Assimilation Approaches For Magnetic Resonance Temperature Imaging

<p>During magnetic resonance (MR)-guided thermal therapies, proton resonance frequency shift (PRFS) based MR temperature imaging can quantitatively monitor tissue temperature changes. It is widely known that the PRFS technique is easily perturbed by tissue motion, tissue susceptibility changes, magnetic field drift, and modality–dependent applicator induced artifacts. Due to recent advances in computational algorithms and hardware, much more powerful statistical analysis methods are becoming realizable in the real-time processing environment. To this end, my dissertation research focused on the development, validation, and implementation of stochastic data-driven processing techniques to increase the robustness of MR temperature monitoring during thermal therapies. MR temperature imaging was demonstrated to achieve a high degree of accuracy in damage predictions during rapid ablation procedures. In the event of temperature imaging data loss, a Kalman filtered MR temperature imaging algorithm using an uncorrelated, sparse covariance matrix for a Pennes bioheat model was developed to predict temperature in regions of missing or erroneous measurement. Temperature predictions were demonstrated to be accurate, while being less computationally expensive than the dense covariance matrix used in standard Kalman filtering. A second approach developed and investigated was the use of a Gaussian process for MR temperature imaging to allow for an accurate probabilistic extrapolation of the background phase. The technique demonstrated reliable temperature estimates in the presence of unwanted background field changes. The Gaussian process algorithm was also implemented to forecast temperature using a limited number of <em>a priori </em>temperature images. The performance of these proposed approaches was validated in simulations, <em>ex vivo</em>, and <em>in vivo.</em> These techniques allow for a full probabilistic prediction and an estimate of the uncertainty that provide a statistical model for MR temperature imaging. In conclusion, I have developed two novel approaches to MR temperature imaging post-processing and demonstrated the feasibility of application of these stochastic, data-driven models developed to improve the robustness of MR-guidance during thermal therapies and potentially enhance the safety and efficacy of treatment.</p>

Name thesis:degree_name

Doctor of Philosophy (PhD)

Level thesis:degree_level

Dissertation (PhD)

Year dc:date.available

2014

Author dc:creator

• Yung, Joshua P.

Contributors dc:contributor

• John D. Hazle, Ph.D.
• David Fuentes, Ph.D.
• Valen Johnson, Ph.D.

• magnetic resonance temperature imaging
• thermal therapy
• Medicine and Health Sciences

Repository record dc:identifier

https://digitalcommons.library.tmc.edu/utgsbs_dissertations/512

OAI identifier oai:identifier

oai:digitalcommons.library.tmc.edu:utgsbs_dissertations-1551