Generalizability of prior-knowledge basis functions across dynamic CBCT perfusion scans and injection protocols

Determine whether the prior-knowledge basis function set, derived via singular value decomposition from two dynamic CT perfusion liver scans, suffices to reconstruct dynamic cone-beam CT liver perfusion scans for other subjects, particularly when contrast agent injection protocols differ, without loss of perfusion map accuracy.

Background

The paper applies the time separation technique (TST) for dynamic cone-beam CT (CBCT) liver perfusion imaging, modeling time attenuation curves using orthonormal basis functions. One basis set is obtained from prior knowledge by extracting orthonormal temporal components via singular value decomposition (SVD) from dynamic CT perfusion (dCTp) time series of two animals.

Using this prior-knowledge basis set with four components, the authors demonstrate promising reconstruction quality for both simulated and real dynamic CBCT perfusion scans. However, they explicitly note uncertainty regarding whether this limited prior-knowledge basis generalizes to other CBCT perfusion scans, especially under different contrast agent injection protocols.