Is liver perfusion CT reproducible? A study on intra- and interobserver agreement of normal hepatic haemodynamic parameters obtained with two different software packages.

ABSTRACT

OBJECTIVE: To evaluate the agreement between the measurements of perfusion CT parameters in normal livers by using two different software packages. METHODS: This retrospective study was based on 78 liver perfusion CT examinations acquired for detecting suspected liver metastasis. Patients with any morphological or functional hepatic abnormalities were excluded. The final analysis included 37 patients (59.7 ± 14.9 y). Two readers (1 and 2) independently measured perfusion parameters using different software packages from two major manufacturers (A and B). Arterial perfusion (AP) and portal perfusion (PP) were determined using the dual-input vascular one-compartmental model. Inter-reader agreement for each package and intrareader agreement between both packages were assessed with intraclass correlation coefficients (ICC) and Bland-Altman statistics. RESULTS: Inter-reader agreement was substantial for AP using software A (ICC = 0.82) and B (ICC = 0.85-0.86), fair for PP using software A (ICC = 0.44) and fair to moderate for PP using software B (ICC = 0.56-0.77). Intrareader agreement between software A and B ranged from slight to moderate (ICC = 0.32-0.62) for readers 1 and 2 considering the AP parameters, and from fair to moderate (ICC = 0.40-0.69) for readers 1 and 2 considering the PP parameters. CONCLUSION: At best there was only moderate agreement between both software packages, resulting in some uncertainty and suboptimal reproducibility. Advances in knowledge Software-dependent factors may contribute to variance in perfusion measurements, demanding further technical improvements. AP measurements seem to be the most reproducible parameter to be adopted when evaluating liver perfusion CT.