Add file collection metadata as array-type fields to JSONs

[tsalo]

Summary

We could either add this functionality to cubids add-nifti-info or create a new command (cubids add-collection-info?). The idea would be to look for file collections (the combinations of entity and suffix to look for will need to be hardcoded), grab all files in a collection, load the relevant metadata fields, and concatenate them across the files to create an array-type metadata field that would be added to all of the JSONs for the files in the collection.

For example, with a multi-echo BOLD scan, we might have the following files:

• sub-01_task-rest_echo-1_bold.nii.gz
• sub-01_task-rest_echo-2_bold.nii.gz
• sub-01_task-rest_echo-3_bold.nii.gz

The relevant metadata field in this scenario is EchoTime. Let's say that the files have the following metadata dictionaries, respectively:

• {"EchoTime": 0.015}
• {"EchoTime": 0.03}
• {"EchoTime": 0.045}

The aggregation step should flag all three files, collect the EchoTime value from each, and concatenate those values in a new array that gets written out to all three metadata files, so they end up with the following:

• {"EchoTime": 0.015, "EchoTimes": [0.015, 0.03, 0.045]}
• {"EchoTime": 0.03, "EchoTimes": [0.015, 0.03, 0.045]}
• {"EchoTime": 0.045, "EchoTimes": [0.015, 0.03, 0.045]}

Then, cubids group will only have to compare echo 1 to echo 1 across subjects, echo 2 to echo 2, etc. but if there is a variant run where echo 3 has a value of 0.05, then all three echoes of that variant run will be flagged as a variant because the array of echo times ([0.015, 0.03, 0.05] vs. the regular [0.015, 0.03, 0.045]) will differ from the dominant group.

In order to make this work, we need to encode the following in the function:

1. The "bold" suffix + the "echo" entity defines a file collection.
2. A file collection defined by the "echo" entity relies on the "EchoTime" metadata field.