Fig. 3

Schematic illustration of how the motion model is applied to motion correct a single gate of PET data, \(I_g\). The input of a trial value for the internal variable, \(x_i\), into the motion model, \(\phi\), generates a trial motion field, \(M\). The gate to be corrected is transformed using the motion field and quantitatively compared to the reference PET gate using a similarity measure. This process is repeated over a range of trial values for x. The value, \(\hat{x}\), which produces the maximum similarity is then used with the motion model to motion correct \(I_g\)