CT perfusion maps of liver computed with different methods match

Purpose: To achieve accurate information regarding the reproducibility of Computed Tomography perfusion (CTp), by comparing perfusion parameters computed using different independent CTp algorithms. Methods and Materials: 49 patients with colorectal cancer and normal liver underwent axial CTp examinations at diagnosis. Blood flow (BF) values were computed on selected Regions of Interest (ROI) assuming a dual-input one-compartment model, using three different methods: maximum slope, deconvolution and the Materne model. After removing unreliable BF values, the perfusion maps were compared visually and correlation analyses between maps were performed (high correlations only are considered, when R2≥0.95). Results: In all the examinations, most of CT perfusion maps show a perfect match, for both high- and low-perfused regions, preserving the details. For some examinations, the Materne model tends to make the transitions between differently perfused regions sharper. All correlations had R2≥0.90, four of them with R2<0.95, and 45 examinations were finally considered. 23 of them showed R2=0.99 and five only had R2<0.97. It is worth remarking that a marked visual correspondence also exists for maps with R2<0.95. Conclusion: The excellent match between the perfusion maps achieved proves that the different algorithms can yield comparable results, useful for visual evaluations of local perfusion. Moreover, the very good correlations between BF values obtained with computational methods that are independent from each other improve the reliability of voxel-based perfusion values, this taking a step forward to the reproducibility of the CTp technique.