Development of an integrated pipeline for correction of gradient nonlinearity, B0, motion and eddy currents artefacts in the CONNECTOM MRI.(Alfred, Michael, Fakhereh, Cornelius)
Using Machine Learning to classify(and correct) fat artefacts in MRI data.(Nico)
Porting more functionality from the CBS-Tools code base to Nighres.(Kornelius)
Creating a Docker image for CBS-Tools(Daniel)
Preprocessing of anatomical and functional MRI data / time series and manually delineating the lesion site(Seyma)
Using Machine Learning to classify(and correct) fat artefacts in MRI data.(Nico)
start with classification artifact(yes/no)
/data/pt_life_dti/
random sample from 15-40 axial slice
orientation of samples different across dataset
after certain date, protocol changed
exclude samples with other artifacts?
existing annotation:
no, mild, strong artifacts
< 3 slices, 3-5 slices, in > 5 slices
on a subsample of slices with distant of 5
maybe we should ignore the samples with the Minnesota sequence at the first run?
we also have annotation of the original data(not only residuals)
we should look at correlation between those annotations(residuals vs. actual image data)
order subject IDs by scanning date
find moment when angle was changed(decided not to do it)
project ideas