Cluster analysis of quantitative parametric maps from DCE-MRI: Application in evaluating heterogeneity of tumor response to antiangiogenic treatment

Purpose: The objective of this study was to compare a clustering approach to conventional analysis methods for assessing changes in pharmacokinetic parameters obtained from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) during antiangiogenic treatment in a breast cancer model. Materials and methods: BALB/c mice bearing established transplantable her2+ tumors were treated with a DNA-based antiangiogenic vaccine or with an empty plasmid (untreated group). DCE-MRI was carried out by administering a dose of 0.05mmol/kg of Gadocoletic acid trisodium salt, a Gd-based blood pool contrast agent (CA) at 1T. Changes in pharmacokinetic estimates (K^trans and v<inf>p</inf>) in a nine-day interval were compared between treated and untreated groups on a voxel-by-voxel analysis. The tumor response to therapy was assessed by a clustering approach and compared with conventional summary statistics, with sub-regions analysis and with histogram analysis. Results: Both the K^trans and v<inf>p</inf> estimates, following blood-pool CA injection, showed marked and spatial heterogeneous changes with antiangiogenic treatment. Averaged values for the whole tumor region, as well as from the rim/core sub-regions analysis were unable to assess the antiangiogenic response. Histogram analysis resulted in significant changes only in the v<inf>p</inf> estimates (p<0.05). The proposed clustering approach depicted marked changes in both the K^trans and v<inf>p</inf> estimates, with significant spatial heterogeneity in v<inf>p</inf> maps in response to treatment (p<0.05), provided that DCE-MRI data are properly clustered in three or four sub-regions. Conclusions: This study demonstrated the value of cluster analysis applied to pharmacokinetic DCE-MRI parametric maps for assessing tumor response to antiangiogenic therapy.