Non-negative matrix factorization for differentiation of brain metastasis and glioblastoma multiforme, and visualization of tumor infiltration

multiforme. Right: NNMF abundances for the normal constituent vector [0.651 0.194 0.156]. White reflects high abundances and black corresponds to low abundances. The abundances visualize tumor infiltration. Figure 4. Left: T2-weighted MR image of a patient with glioblastoma multiforme. Right: NNMF abundances for the normal constituent vector [0.789 0.082 0.129]. White reflects high abundances and black corresponds to low abundances. The abundances visualize tumor infiltration in adjacent tissue in the contralateral hemisphere. Figure 2. Scatter plot of the difference between the ratios element3/element2 and element3/element1 from the first constituent vector and the second constituent vector for each patient. Metastases and glioblastomas are separated. Figure 1. Scatter plot of the difference between the averaged Cho/Cr and Cho/NAA ratio from the enhancing tumor and the non-enhancing/non-T2 hyperintense area for each patient. Metastases and glioblastomas form two separated clusters. and metastases. Figures 3 and 4 visualize the abundances of the normal component that is obtained with NNMF for two cases with glioblastoma multiforme. These abundances indicate tumor infiltration. In conclusion, the use of 2D-TSI enables to visualize metabolic differences between glioblastomas and metastases. Automated processing with NNMF allows differentiation of these tumors and enables to visualize tumor infiltration.