Improvement of perfusion characterisation in two lung tumour subtypes using de-noised CT perfusion maps

Purpose: This study aims at assessing if, and to what extent, removing from blood flow (BF) maps those values based on the error underlying their computation, allows highlighting significant differences in perfusion of squamous cell carcinoma (SCC) and adenocarcinoma (ADK). Methods and Materials: Six SCC and twenty-eight ADK referring to thirty-four patients who underwent a CT perfusion (CTp) at the diagnosis stage, were considered in this study. BF values were computed according to the maximum slope method. Also, BF values arising from time-concentration curves with high fitting errors have been automatically removed, yielding corrected maps. Finally, sixty-eight couples of corrected and uncorrected BF maps were achieved. Five features from first-order statistics were computed on each BF map and the differences between SCC and ADK were assessed through the one-tail Welch's t-test (p-value<0.01). Results: Differences in BF values between SCC and ADK groups are not detected with uncorrected perfusion maps. On the contrary, significant differences were highlighted by kurtosis, using corrected maps. In particular, at baseline, perfusion values of the SCC subtype (mean BF=62.2) shows a kurtosis significantly greater (p=0.003) than that of the ADK (mean BF=89.6). Conclusions: The correction of BF maps emphasizes functional features of these cancer subtypes. In particular, before starting therapy SCC has significantly smaller BF values, pointing out the presence of wide necrotic regions, while ADK BF values are generally higher and more heterogeneous. These findings should be considered for planning optimization of the anti-angiogenic therapies, assessed with CTp, thus encouraging its use in clinic.