Tumor tissue analysis by self organizing maps from combined DCE-/DSC-MRI data

Self organizing maps are utilized to analyze tumor tissue from simultaneously acquired dynamic T1 and T2^* magnetic resonance imaging measurements of five tumor-bearing mice. The method allowed for tumor segmention obtaining regions characterized by distinct perfusion patterns (i.e. T1 and T2^* perfusion time curves). Compared to histopathological analysis, these regions comprise typical tumor areas like pre-necrotic or vascularized tissue. Furthermore, the detected regions showed differences in physiological parameters (K^trans, v_e) extracted by a pharmacokinetic model. In summary, tumor tissue analysis by self organizing maps is feasible and seems to be a valuable tool in model-free assessment of tumor physiology.