Meeting Banner
Abstract #0444

Reference-Less MR Thermometry Using Iteratively-Reweighted L1 Regression

William Allyn Grissom1, Michael Lustig2, Viola Rieke, Andrew B. Holbrook3, John B. Pauly2, Kim B. Butts-Pauly

1Electrical Engineering and Radiology, Stanford University, Stanford, CA, USA; 2Electrical Engineering, Stanford University, Stanford, CA, USA; 3Bioengineering, Stanford University, Stanford, CA, USA

Conventional reference-less thermometry techniques derive baseline phase images using least-squares polynomial regression performed on image phase during thermotherapy. Because least-squares regression is sensitive to outliers, i.e., the phase within the hot spot, the hot spot must be masked out of this regression, so its location must be known or tracked. We propose a new thermometry method that uses reweighted-L1 polynomial regression to prevent hot spot bias, obviating the need for masking or tracking. The method is therefore more robust to motion than conventional reference-less thermometry, and requires less user interaction.