Differences in perfusion CT parameter values with commercial software upgrades: a preliminary report about algorithm consistency and stability

Background: Computed tomographic perfusion (CTp) imaging is a promising technique that allows functional imaging, as an adjunct to a morphologic CT examination, that can be used as an aid to carefully evaluate the response to therapy in oncologic patients. Considering this statement, it could be desirable that the measurements obtained with the CT perfusion software, and their upgrades, are consistent and reproducible. Purpose: To determine how commercial software upgrades impact on algorithm consistency and stability among the three version upgrades of the same platform in a preliminary study. Material and Methods: Blood volume (BV), blood flow (BF), mean transit time (MTT), and permeability surface area product (PS) were calculated with repeated measurements (n 1/41119) while truncating the time density curve at different time values in six CT perfusion studies using CT perfusion software version 4D (CT Perfusion 4D), then repeated with the previous version (CT Perfusion 3.0 and CT Perfusion 4.0), using a fixed ROI both for arterial input and target lesion. The software upgrades were compared in pairs by applying a Kolmogorov-Smirnov test to all the parameters measured. Stability and reliability of the three versions were verified through the variation of the truncated parameters. Results: The three software versions provided different parent distributions for approximately 80% of the 72 parameters measured. A complete agreement was found only for one patient in version 3.0 vs. 4.0 and 3.0 vs. 4D. Perfusion 4.0 vs. 4D: A complete agreement was found only in two cases. Parameters obtained with Perfusion 4D always showed the lowest standard deviation in all temporal intervals and also for all individual parameters. Conclusion: The three versions of the same platform tested yield different perfusion measurements. Thus, our preliminary results show that Perfusion 4D version uses a stable deconvolution algorithm to provide more reliable measurements.