Rapid DCE-MRI parameter generation using principal component analysis and clustering

Introduction: Dynamic contrast-enhanced MRI (DCE-MRI) has considerably potential to evaluate tumour status at presentation, during treatment and for interval follow-up through its ability to assess the state of the blood-brain barrier and blood vessel patency in gliomas. It is likely to gain in importance with the advent of new vascular-targeted therapies for high-grade lesions. Absolute quantification of the data, however, is hampered, first, by the technical demands of the acquisition – rapid sampling and heavy T1 weighting – which generate low SNR images. Secondly, spatial heterogeneity of the derived parameters necessitates a voxel-based, rather than region-based, estimation of their values and thus increases the computational demand. Here we report a method that simultaneously reduces the impact of both these limitations to rapidly produce parameter maps with increased precision. This method may be of interest to clinicians, and both clinical and basic scientists.