Compression of MR-Thermometry images using reduced-order Karhunen-Loève basis

In this paper, we develop an approach to achieve MR thermal image compression by exploiting the spatial correlations of image voxels during thermal treatment. A set of reduced-order basis was identified using Karhunen-Loève (KL) decomposition of MR thermal images. Each image can be compressed with a relative small number of identified KL basis functions. The proposed approach can be used in real-time compression of imaging data in order to minimize a potentially massive amount of information that must be exchanged between the MR scanner and the treatment controller, and reduce computer storage requirements during thermal treatment. Simulation and MR thermometry experimental results demonstrate that the proposed image compression method is also effective in suppressing high spatial-frequency MR measurement noises.