FL-ready dataset¶

In this step, you will combine the standardized IDPs and the harmonized phenotypic data into one table per prediction task, in the format expected by Fed-BioMed and the machine learning models we will train.

Note

All commands assume that the working directory is the root of the brainflex-node repository. They will need to be modified if that is not the case.

Record start time¶

brainflex record fl_ready_dataset start

Pipeline installation¶

Install the Nipoppy-to-Fed-BioMed conversion pipeline. This pipeline was written for this FL experiment specifically.

nipoppy pipeline install 23753437 --study nipoppy_study

The pipeline should now be listed when you run:

nipoppy pipeline list --study nipoppy_study

Output¶

WARNING  No available bidsification pipelines
INFO     Available processing pipelines and versions
INFO             - freesurfer        (7.3.2)
INFO             - nipoppy2fedbiomed (0.1.0)
INFO     Available extraction pipelines and versions
INFO             - fs_stats (0.3.0)

Dataset extraction¶

Combine the FreeSurfer IDPs and harmonized phenotypic data into a Fed-BioMed dataset:

nipoppy process --pipeline nipoppy2fedbiomed --pipeline-step age --study nipoppy_study
nipoppy process --pipeline nipoppy2fedbiomed --pipeline-step sex --study nipoppy_study

For each of the prediction tasks (brain age regression and biological sex classification), this will write a CSV file (the Fed-BioMed dataset) and a JSON file that can be used to register it on the Fed-BioMed node in the next step.

tree nipoppy_study/derivatives/nipoppy2fedbiomed

Output¶

nipoppy_study/derivatives/nipoppy2fedbiomed
└── 0.1.0
    └── output
        ├── age
        │   ├── dataset.csv
        │   └── dataset_info.json
        └── sex
            ├── dataset.csv
            └── dataset_info.json

Record end time and answer questions¶

brainflex record fl_ready_dataset end