Fiber up‐sampling according to this paper
dependencies:
- numpy
- dipy
- nibabel
- matplotlib
- sklearn
parameters:
- nrClusters: number of clusters (k) used during KMeans clustering in the cluster() function
- nrRand: number of new streamlines which are generated per cluster
- samplePoints: number of points per streamline during resampling
- cutOff: dimensionality after the PCA transformation
todo:
- replace matplotlib PCA with sklearn PCA (matplotlib PCA is depricated!)
- add ability to load fiber_assignment.txt instead of using k-means clustering
- crop or remove streamlines with points outside of the FOV
- remove streamlines with a too large distance to the bundle mean fiber (e.g. max distance of initial streamlines as threshold)
- further regularize location of streamline start and end points
- explore streamline distribution (bundlewise and total) in PCA-space
- try spline representation instead of resampled points