#
# first set up dependencies then reorganize data into 
# dataset/subjectID/timepoint/modality/subjectID_timepoint_modality.nii.gz
#
# subjectID should only have numeric or alphabetic characters or underscore
# and should never include characters such as @ ! # space etc
#
# in this example case based on kirby data , we do this for you in the step00 script
#
# ./pipelines/step_01_organize_data.sh ./pipelines/dependencies.sh
#
# now use the ants framework to build single subject and group templates 
framework=ants
# 
#  Should check subject leverage i.e. not select bad subjects! 
#
#  Define a csv file for each subject of size N-Time-Points by regressors
# 
#  FrameIsGood,GrandMean,CompCorrEvec1...N,SNRperFrame,ContrastToNoiseperFrame,TimeValueInSec,MotionParametersFrameToFrame,MotionParametersFrameToReference 
#   ENH: update 11/20 have GrandMean,CompCorrEvec1...N from output of ImageMath CompCorrAuto 
#   ENH: update 11/20 have spatial and temporal smoothing in sccan 
#
# TimeValueInSec - NA unless otherwise specified 
#
# ADD MOTION CORRECTION BASED QC (annotate bad frames) AND OPTIONAL BRAIN EXTRACTION???
#
# get subject specific templates 
# ./pipelines/${framework}/step_02_create_4D_templates.sh ./pipelines/dependencies.sh 
# build group templates
# ./pipelines/${framework}/step_03_create_group_templates.sh ./pipelines/dependencies.sh 
# map 4D data to template space 
# ./pipelines/${framework}/step_04_apply_transform_to_4D.sh ./pipelines/dependencies.sh
# here, preprocessing is done
# now enter the statistics part of the script
# 
# get the gray matter mask , the ROI of interest and then compute CompCorr , then evaluate test-retest and visualize results
# 
# 5.1 - segment cortex --- done 11/20 in ImageMath
# 5.2 - label ROI in cortex  --- done 11/20
# 5.3 - CompCorr (physio noise)  --- done 11/20
#     5.3.1 - compute the time series variance at each voxel in the brain 
#     5.3.2 - build a histogram of the 5.3.1 variance over the brain   
#     5.3.3 - take the high variance (>95%) voxels and put them in the nuisance matrix
#     5.3.4 - do PCA on the nuisance matrix & factor out the top N eigenvectors from the original time series data 
#     Note : we also subtract the grand mean in this implementation --- maybe the grand mean (and compcorr output) should be written out as a csv file that could be used as covariates   
#     Note2 : temporal smoothing is implemented here for now but should be elsewhere 
# 5.4 - Preprocessing e.g. spatial and temporal smoothing (define parameters in dependencies.sh)  --- done 11/20 in sccan
# 5.5 - Level1 stats e.g. resting state correlations (via R) --- done 11/20 in statistics directory for rsf network see Rscript ~/data/kirby/statistics/antsr_resting_state_corr.R  ~/code/sccan/bin/sccan ${ID}cortmask.nii.gz $ID

# 5.6 - Level2 stats e.g. group consistency of resting state correlations (via R) --- done 11/20 via t-test 
# 5.7 - evaluation 
# 
# call this to get the compcorr outputs , the segmentations and brain masks .... you'd need roimasks to get the rsf networks  

./pipelines/${framework}/step_05_brainmask_and_compcorr.sh ./pipelines/dependencies.sh