Testing Fidelity of DTI Tractography Using Simulated Fiber: Comparison of FACT and FMT

1. Fiber Tensor Testing Fidelity of DTI Tractography Using Simulated Fiber: Comparison of FACT and FMT Swati Rane1,2, Xiangyang Ma2, Timothy Duong1,2 1 Georgia Institute of Technology, 2 Yerkes Imaging Center, Emory University INTRODUCTION FMT • Randomly oriented tensors were allotted in pixels outside the fiber (i.e. in GM). • Fiber tracking was evaluated using FACT and FMT. • Root-mean-squared distance measure was used to determine accuracy. • The fiber tract truncated at the fork due to averaging of tensors. • DTI tractography is becoming increasingly popular for mapping brain connectivity in vivo. A few fiber tracking algorithms have been proposed. However, the accuracy of the fiber tracts obtained by DTI remains to be rigorously verified. It is also unclear whether fiber crossing and merging can be reliably determined. • The goal of this study is to generate simulated fibers for DTI tractography and to test the reliability of two common approaches in DTI tractography algorithms. • Line propagation technique - using local tensor orientation, e.g. Fiber Assignment by Continuous Tracking (FACT). • Energy minimization technique - using a maximum likelihood of connectivity map, e.g. Fast Marching Technique (FMT). Fig 5: FMT tracing from the base of the fiber • The fiber from one branch diverged at the fork but recovered and tracked correctly. Fig 2: Tensor orientations in WM and GM Fig 6: Front propagated from the base of the fiber and tracked by steepest descent from the branch CONCLUSIONS Simulated fiber RESULTS Reconstructed fiber METHOD FACT Seed • Simulated fibers were generated with gradual curvature and a branching point using Matlab. • Tensor directions of the simulated fibers were estimated using 8 neighboring pixels. • Diffusion tensors along the other 2 principal Eigen-vectors were not considered. • Fibers were generated with fractional anisotropy (FA) values of Gray Matter (GM) FA = 0.1 - 0.2 • White Matter (WM) FA = 0.5 - 0.65 • FACT failed at the fork point due to averaging of tensors. * Robustness depends on techniques used for front marching • Simulated fiber offers a reliable means to study DTI tractography and to compare different tracking algorithms. • FACT can deviate from the main fiber and track an adjacent fiber because it does not utilize any information from neighboring pixels. • FMT uses tensor information in the surrounding pixels and hence is more robust. Although the tracts deviates slightly at the fork, the algorithm recovers quickly and completes the tract. The disadvantage is that FMT is computationally more intensive. • FUTURE WORK • Extend to3D model and experimental data. • Test new methods of tract tracing. Fig 3: FACT tracing from the base of the fiber • Tracking also failed from the branching point. Fig 1: Construction of tensor from the fiber Fig 4: FACT propagation from the branched ends