In its current behavior, MRIQC generates one set of timeseries (timeseries.tsv) and one set of IQMs (bold.json) per echo in a multi-echo dataset.
This behavior is illogical because the multiple echoes are merged into a single preprocessed BOLD image. Consequently, one IQM has multiple values associated with one preprocessed BOLD image.
We thus need to put some effort into computing only one set of IQMs for multi-echo datasets.
However, we need to carefully consider, for each IQM, which echo (or combination of echoes) is the most preferable for computation (e.g some IQMs might be better computed on the second echo, but others might be better computed on a weighted average of the echoes)
oesteban
commented
3 months ago
In its current behavior, MRIQC generates one set of timeseries (
timeseries.tsv) and one set of IQMs (bold.json) per echo in a multi-echo dataset. This behavior is illogical because the multiple echoes are merged into a single preprocessed BOLD image. Consequently, one IQM has multiple values associated with one preprocessed BOLD image.We thus need to put some effort into computing only one set of IQMs for multi-echo datasets.
However, we need to carefully consider, for each IQM, which echo (or combination of echoes) is the most preferable for computation (e.g some IQMs might be better computed on the second echo, but others might be better computed on a weighted average of the echoes)