A Comparative Machine Learning Study of Connectivity-Based Biomarkers of Schizophrenia

Preprocessing

BOLD fMRI preprocessing scripts are in SCZ/preprocessing. Post fMRIPrep preprocessing requires this package.

Gradient Dispersion

All script used to compute neighborhood and centroid dispersion are in Dispersion.

Data Preparation

All features need to be vectorized and concatenated in one array. /SCZ/modeling/aggregate_features.py does that provided you have a dictionary of paths with preprocessed data.

Permutation Feature Importance

The pipeline described in Figure 4 of the manuscript is implemented in SCZ/modeling/lr_pipeline.py.

Multi-Classifier Analyses

All scripts related to the assessment of classification performance are in SCZ/modeling.

All helper functions used in the stages mentioned above are in SCZ/modeling_utils.py