Article
Can we eliminate the erroneous fibers? Stepwise threshold increase in diffusion based depiction of the cranial nerves
Search Medline for
Authors
-
Amir Zolal
- Klinik und Poliklinik für Neurochirurgie, Universitätsklinikum Carl Gustav Carus der Technischen Universität Dresden
-
Gabriele Schackert
- Klinik und Poliklinik für Neurochirurgie, Universitätsklinikum Carl Gustav Carus der Technischen Universität Dresden
-
Stephan B. Sobottka
- Klinik und Poliklinik für Neurochirurgie, Universitätsklinikum Carl Gustav Carus der Technischen Universität Dresden
| Published: | June 2, 2015 |
|---|
Outline
Text
Objective: In order to visualize cranial nerves using DTI, some extent of manual fiber selection methods and fractional anisotropy (FA) threshold manipulation has been used according to previous reports in order to eliminate the resulting fibers not corresponding to nerve anatomy. We have hypothesized that a stepwise increase in a threshold value could lead to selection of anatomically correct fibers without resorting to manual manipulation. In order to test this hypothesis, we have evaluated three tracking methods on a set of publicly available normal subject DTI data.
Method: Three fiber tracking methods have been used to track the cranial nerves II, III, V and the VII+VIII bundle using publicly available DWI data with a 2 ROI method to simulate a problem where the position of the nerve is unknown (DTI, QBI and probabilistic tracking with FSL). In total, 40 multi-step analyses were performed with each of the methods. The correct fibers were identified by comparing the results with the coregistered FLAIR maps. The elimination of erroneous fibers was attempted by gradually increasing the respective thresholds (FA, QA, probability index). The success in eliminating the erroneous fibers was graded as:
- 1.
- Successful stepwise selection – incorrect fibers were eliminated.
- 2.
- Unsuccessful stepwise selection – the correct fibers were eliminated before the erroneous fibers (or in the same step).
- 3.
- No depiction possible.
Results: With the use of DTI fiber tracking, anatomically correct fibers could be identified within the fiber tracking result in 36 of the 40 analyses with initial settings (90%). The stepwise selection was unsuccessful in 18 cases (45%), and successful in 18 cases (45%, exact binomial test, p=0.6). Using QBI fiber tracking, anatomically correct fibers were initially present in 39 of the 40 analyses (97.5%). The stepwise selection method based on the QA threshold was unsuccessful in 21 cases (52.5%) and successful in 18 cases (45%, p=0.6). The FSL based probabilistic tracking was possible in 39 cases (97.5%). The stepwise selection based on the probability threshold led to a correct depiction in 36 cases (90%, p<0.01).
Conclusions: Probabilistic tracking with stepwise threshold increase might represent a method usable to identify the course of the cranial nerves using diffusion MRI without manual intervention.
