Unsupervised multiparametric classification of dynamic susceptibility contrast imaging: study of the healthy brain

Characterization and quantification of magnetic resonance perfusion images is important for clinical interpretation, though this calls for a reproducible and accurate method of analysis and a robust healthy reference. The few studies which have examined the perfusion of the healthy brain using dynamic susceptibility contrast (DSC) imaging were largely limited to manual definition of the regions of interest (ROI) and results were dependent on the location of the ROI. The current study aimed to develop a methodology for DSC data analysis and to obtain reference values of healthy subjects. Twenty three healthy volunteers underwent DSC. An unsupervised multiparametric clustering method was applied to four perfusion parameters. Three clusters were defined and identified as: dura-blood-vessels, gray matter and white matter and their vascular characteristics were obtained. Additionally, regional perfusion differences were studied and revealed a prolonged mean transient time and a trend for higher vascularity in the posterior compared with the anterior and middle cerebral vascular territories. While additional studies are required to confirm our findings, this result may have important clinical implications. The proposed unsupervised multiparametric method enabled accurate tissue differentiation, is easy replicable and has a wide range of applications in both pathological and healthy brains.