Statistical methods in neuroimaging with particular application to emission tomography

Abstract

We review statistical methods being applied in four key areas connected with PET and SPECT neuroimaging: (i) image reconstruction (briefly); (ii) tracer-kinetic, or compartmental, modelling; (iii) inference from region-of-interest data; (iv) inference at the pixel or voxel level. Under the last heading, we pay particular attention to the analysis of data from serial scans. We conclude by identifying some topics for future statistical research.