Quality Check Stream
The above image depicts a decision making tree when using the semi-automated quality stream.
The output of the "bad scan/slice" detection will be a PDF, a .mat, possibly a text file for each run. The PDF will contain figures and a list of the corrupted scans/slices that were detected. The .mat file will contain all the variables used to generate the PDF. The text file will only be created if there were corrupted scans detected. It will contain the regressor for the GLM.
The unprocessed data check stream is used to detect any scans that are beyond repair. It is also used to detect single slices corrupted by artifacts.
The post-processed data check is intended to check for scans that continue to be abnormal after realignment and normalization. This is to confirm that the abnormality is not due to movement. The variability in these scans is accounted by including a regressor in the first-level GLM. The data check can also detect single slices corrupted by image artifacts not due to movement. The slices can be corrected by methods of interpolation.
Metrics used for bad scan detection:
- z-score of scan mean for difference scans - default threshold TBD
- maybe just used art toolbox for this which also does motion detection
Metrics used for bad slice detection:
- z-score of mean slice values - default threshold 4
- MSE between slice scans - supplementary for z-score metric, threshold > 50
Figures plotted in PDF:
- SNR as defined by Tor Wager histogram (may be useful later)
- z-score of mean slice values
- MSE between slice scans
- Whole scan means
