The use of error-category mapping in pharmacokinetic model analysis of dynamic contrast-enhanced MRI data

This study introduces the use of 'error-category mapping' in the interpretation of pharmacokinetic (PK) model parameter results derived from dynamic contrast-enhanced (DCE-) MRI data. Eleven patients with metastatic renal cell carcinoma were enrolled in a multiparametric study of the treatment effects of bevacizumab. For the purposes of the present analysis, DCE-MRI data from two identical pre-treatment examinations were analysed by application of the extended Tofts model (eTM), using in turn a model arterial input function (AIF), an individually-measured AIF and a sample-average AIF. PK model parameter maps were calculated. Errors in the signal-to-gadolinium concentration ([Gd]) conversion process and the model-fitting process itself were assigned to category codes on a voxel-by-voxel basis, thereby forming a colour-coded 'error-category map' for each imaged slice. These maps were found to be repeatable between patient visits and showed that the eTM converged adequately in the majority of voxels in all the tumours studied. However, the maps also clearly indicated sub-regions of low Gd uptake and of non-convergence of the model in nearly all tumours. The non-physical condition ve ≥ 1 was the most frequently indicated error category and appeared sensitive to the form of AIF used. This simple method for visualisation of errors in DCE-MRI could be used as a routine quality-control technique and also has the potential to reveal otherwise hidden patterns of failure in PK model applications.