Evaluation of glioblastomas and lymphomas with whole-brain CT perfusion: Comparison between a delay-invariant singular-value decomposition algorithm and a Patlak plot

SummaryObjectiveCorrection of contrast leakage is recommended when enhancing lesions during perfusion analysis. The purpose of this study was to assess the diagnostic performance of computed tomography perfusion (CTP) with a delay-invariant singular-value decomposition algorithm (SVD+) and a Patlak plot in differentiating glioblastomas from lymphomas.Materials and methodsThis prospective study included 17 adult patients (12 men and 5 women) with pathologically proven glioblastomas (n = 10) and lymphomas (n = 7). CTP data were analyzed using SVD+ and a Patlak plot. The relative tumor blood volume and flow compared to contralateral normal-appearing gray matter (rCBV and rCBF derived from SVD+, and rBV and rFlow derived from the Patlak plot) were used to differentiate between glioblastomas and lymphomas. The Mann-Whitney U test and receiver operating characteristic (ROC) analyses were used for statistical analysis.ResultsGlioblastomas showed significantly higher rFlow (3.05 ± 0.49, mean ± standard deviation) than lymphomas (1.56 ± 0.53; P < 0.05). There were no statistically significant differences between glioblastomas and lymphomas in rBV (2.52 ± 1.57 vs. 1.03 ± 0.51; P > 0.05), rCBF (1.38 ± 0.41 vs. 1.29 ± 0.47; P > 0.05), or rCBV (1.78 ± 0.47 vs. 1.87 ± 0.66; P > 0.05). ROC analysis showed the best diagnostic performance with rFlow (Az = 0.871), followed by rBV (Az = 0.771), rCBF (Az = 0.614), and rCBV (Az = 0.529).ConclusionCTP analysis with a Patlak plot was helpful in differentiating between glioblastomas and lymphomas, but CTP analysis with SVD+ was not.