Multi-parametric response map

A generalized approach towards multi-parametric response mapping using principal component analysis

Parametric response mapping (PRM) has emerged as a powerful image-analysis technique which can be used for early prediction of cancer treatment-response (Galban et al. Nat Med. 2009;15(5):572-6). PRM analysis typically involves comparison of longitudinally acquired and spatially-aligned pairs of single-parameter functional images. For example, MRI-derived relative cerebral blood volume maps acquired before and during radiotherapy has been analyzed to predict treatment response for patients with glioblastoma (Galban et al. Nat Med. 2009;15(5):572-6).
The multi-parametric response map (MPRM) provides an analogous treatment response prediction technique designed for analysis of multi-parametric data. Multi-parametric variability within a reference image region (e.g. non-pathologic tissue) is summarized by PCA and then the multi-parametric variability within a target region of interest (e.g. tumour) is classified in terms of the reference PCA.
mprm_create_reference.m is used to perform the reference PCA
mprm_classify_target.m is used to classify target data according to output of mprm_create_reference.m and outputs an array containing MPRMs built from each of the reference principal components.