Quantification of heterogeneity in dynamic contrast enhanced MRI data for tumor treatment assessment

Experimental evidence exists that especially the heterogeneity in contrast enhancement as evaluated by dynamic contrast enhanced MRI (DCE-MRI) is a predictive feature for treatment outcome in a variety of tumor types. In this study it is investigated whether texture descriptors derived from DCE-MRI based heuristic feature maps are suitable to quantify the heterogeneity in contrast uptake. An automated analysis method is proposed that for each voxel first partitions the signal intensity curve into different temporal regions indicating different stages of enhancement. Within these regions, heuristic features describing the contrast dynamics are estimated. The corresponding features maps are used as the basis for texture analysis, based on cooccurrence matrices, to assess tumor contrast uptake heterogeneity. The method has been applied in pre- and post treatment DCE-MRI data in ten patients with soft tissue sarcomas who underwent isolated limb perfusion. The correspondence between texture measures and the heterogeneity in contrast uptake as visually assessed by a radiologist has been evaluated, and the ability of the texture measures to discriminate between two treatment outcome classes has been assessed. The preliminary results suggest that some of the proposed texture measures are suitable to quantify the heterogeneity in contrast uptake in tumor tissue