Blood flow assessment of liver in patients with colorectal cancer: results agreement in a Computed Tomography perfusion multicentre study

Aims and objectives: Improving reproducibility of blood flow (BF) values in Computed Tomography perfusion (CTp) studies of patients with colorectal cancer and normal liver, by accurate data processing and accurate development of two independent CTp computing methods. Methods and materials: 30 patients of two different Centres (17 of Centre A and 13 of Centre B) of PIXEL, the widest French CTp multicentre study, underwent axial CTp liver examinations at colorectal-cancer diagnosis. The time concentration curves (TCCs) have been extracted and an early automatic reliability analysis allowed removing voxel undergoing motion or artefacts. A 3D filtering permitted to reduce spurious time oscillations of TCCs, improving their regularity. After that, in-house maximum slope (MS) and deconvolution (DV) methods have been applied to compute BF, assuming a dual input-mono compartmental model. ANOVA have been considered to assess mean BF (mBF) reproducibility (p>0.05). Intra-Class correlation (ICC) is used to highlight mBF repeatability. Results: ANOVA carried out on Centres A and B reported an excellent mBF reproducibility (p=0.21), while ICC computed for Centres A, B and both together, always show values greater than 0.90 (0.90, 0.93, 0.91, respectively). In addition, it is worth noting that considering both Centres A and B, BF values were spread in very similar ranges, with m±σ equals to 107±25[ml/min/100g] and 108±27[ml/min/100g] for DV and MS, respectively. Conclusion: This early multicentre analysis proves that pre-processing original data with accuracy improves TCCs and the quality of the extracted perfusion parameters, accordingly. The excellent reproducibility and correlation attained between colormaps computed using two popular perfusion computation methods favour CTp standardization.