Sonia Pujol1, Carl-Fredrik Westin2, Ross Whitaker3, Guido Gerig3, Tom Fletcher3, Vincent Magnotta4, Sylvain Bouix5, Ron Kikinis1, William M. Wells III1, Randy Gollub6
1Surgical Planning Laboratory, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA; 2Laboratory of Mathematics in Imaging, Brigham and Women's Hospital, Harvard Medical School, Boston, MA; 3Scientific Computing and Imaging Institute, University of Utah, Salt Lake City, UT; 4University of Iowa, Iowa City, IA; 5Psychiatry Neuroimaging Laboratory, Boston, MA; 6Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Boston, MA
We describe preliminary results obtained from a framework for evaluating DT-MRI tractography algorithms that uses STAPLE (Simultaneous Truth and Performance Level Estimation) to estimate the performance of the methods in the absence of ground truth. In tests on a mathematical phantom we show that STAPLE is able to characterize fiber tracking performance by comparing its findings to true performance findings. We apply the framework to four methods operating on four white matter fasciuli in real data, and describe the results of this new approach for validating tractography.