Automatic Tumor Delineation by Multiparametric MR Analysis Based on Endogenous Contrast

Stefanie J.C.G. Hectors¹, Igor Jacobs¹, Gustav J. Strijkers¹, Klaas Nicolay¹

HIFU tumor treatment planning is generally performed with contrast-enhanced T₁-weighted imaging. In view of possible toxicity issues of the Gd-chelate during HIFU, often a wait time is applied before start of the treatment. It was assessed whether accurate tumor delineation is also possible by multiparametric MR analysis solely based on endogenous contrast. k-means clustering was applied to T₁, T₂ and ADC maps of tumor-bearing mice with all feature vectors and variable numbers of clusters k. Maximum sensitivity (0.87) and specificity (0.90) were found for feature vector {T₂,ADC} with k=6, indicating feasibility of tumor delineation based on endogenous MR parameters.