Abstract #0093

Improving IVIM Derived F-Maps of Pancreatic Tumors with Automatic Duct and Vessel Segmentation

Thomas Joseph Re^1,2, Mirium Klauss³, Andreas Lemke⁴, Frederik B. Laun², Dirk Simon⁵, Riccardo Manfredi¹, Roberto Pozzi Mucelli¹, Stefan Delorme², Bram Stieltjes²

¹Radiology, University of Verona, Verona, Italy; ²Radiology, DKFZ, Heidelberg, BW, Germany; ³Radiology, University of Heidelberg, Heidelberg, BW, Germany; ⁴MS Computer Assisted Clinical Medicine, University of Heidelberg; ⁵Software Development for Integrated Diagnostics and Therapy, DKFZ, Heidelberg, BW, Germany

Maps of IVIM model derived perfusion fraction f (f-maps) of the pancreas show potential for the identification of pancreatic adenocarcinoma lesions which appear hypointense in these images. Unfortunately, since bile and pancreatic ducts also appear as hypointense in f-maps, their presence adjacent to tumors can lead to tumor delineation errors. A novel approach which automatically segmented vessels and ducts in the f-maps based on integrated diffusion coefficient D data was tested in 43 patients and proved to be superior to both the ADC or f-map for tumor delineation.