Spectral Analysis of Dynamic PET Studies: A Review of 20 Years of Method Developments and Applications

Figure 4

Sensitivity of SA filter methods to filter setting. The figure shows the performance of SAIF when different passband bounds are used to fit a particular dataset, corresponding to a simulated spectrum with two equilibrating components and one trapping. Both fit performance (indicated by weighted residual sum of squares, WRSS) and parameter estimates bias (tracer trapping, ) can be heavily affected by filter setting. It is important to note that the filter bounds returning the lowest bias do not match those returning the best data fit (black arrows). This indicates that the filter setting cannot be determined only on the basis of data fit quality.

(a) WRSS analysis

(b) Tracer trapping