J Comput Assist Tomogr. 2016 Jan-Feb;40(1):152-9. doi: 10.1097/RCT.0000000000000331.
Ippolito D1, Casiraghi AS, Talei Franzesi C, Bonaffini PA, Fior D, Sironi S.
OBJECTIVES: To evaluate intrareader and inter-reader agreement in CT perfusion (Perf) measurements in cirrhotic patients with hepatocellular carcinoma (HCC) lesions.
METHODS: Sixteen patients with HCC (9 untreated, 5 recurrence/residual disease after transarterial chemoembolization, and 2 after radiofrequency ablation treatment) underwent standard CT examination and a Perf study; a quantitative map of arterial and portal Perf by means of a color scale was generated. With dedicated Perf software, the following parameters were calculated on untreated and treated HCC lesions and on cirrhotic parenchyma: hepatic Perf, tissue blood volume, hepatic perfusion index, arterial perfusion, and time to peak. Intraobserver and interobserver agreements were assessed for 2 readers with Bland-Altman plots, intraclass correlation coefficients (ICCs), coefficients of variation, and repeatability.
RESULTS: In HCC lesions, agreement analysis demonstrated higher intra-agreement values in comparison with interagreement (range of ICC values, 0.26-0.59 between readers and 0.67-0.94 between readings). Lower interagreement was found for treated HCC lesions in comparison with untreated lesions (range of ICC values, respectively, 0.12-0.54 and 0.31-0.89).
CONCLUSIONS: For all Perf parameters intra-agreement was higher than interagreement, even though a relatively wide range of 95% limits of agreement was found in both cases.