Motion compensation using principal component analysis and projection onto dipole fields for abdominal magnetic resonance thermometry.

Abstract

High intensity focused ultrasound (HIFU) has the potential to locally and non-invasively treat cancer with fewer side effects than alternative therapies. However, motion and tissue heterogeneity in the abdomen can compromise the HIFU focus and confound current thermometry methods. The proposed thermometry method combines principal component analysis (PCA), as a multi-baseline technique, and projection onto dipole fields (PDF), as a near-referenceless method. PCA forgoes tracking tools by projecting incoming images onto a subspace spanning the motion history. PDF is subsequently used to synthesize the naturally feasible components of the residual phase using a magnetic dipole model. This leaves only the phase shifts that are induced by HIFU. With in vivo measurements, in porcine and human kidneys, the mean pixel-wise temperature SD was 0.86 ± 0.41°C in selected regions of interest (ROIs) across all data sets, without any user-interaction or supplementary tracking tools. This is an improvement over a benchmark hybrid method, which scored 1.36 ± 1.20°C on the same data. Uncorrected subtraction of the data yielded a score of 3.02 ± 2.87°C. The PCA-PDF hybrid method achieves superior artifact correction by exploiting the motion history and intrinsic magnetic susceptibility of the underlying tissue.