Project: compare_surf_tools

An extension of BrainHack Project (https://github.com/companat/compare-surf-tools) to compare thickness outputs from different pipelines run on ABIDE-I

Objectives

• Compare output of preprocessing pipelines for structural MR imaging
  • pipelines: Freesurfer (v5.1, v5.3, v6.0, ANTs, CIVET2.1)
  • feature comparisons: ROI-wise (surface parcellations: DKT40, Destrieux, Glasser)
  • analytic comparisons: classifier performance (individual predictions), statistical inference (biological group differences)
• Outlier detection
  • identify outliers at differen scales
  • identify outliers for different tasks

Data

Consolidated from the analysis results provided at http://preprocessed-connectomes-project.org/abide/Description, we provide the following unified data tables in the 'data' directory:

• ABIDE_Phenotype.csv : phenotypic data for the subjects
• ABIDE_ants_thickness_data.csv : thickness data from ANTS analysis
• ABIDE_fs5.3_LandRvolumes.csv : volume data from FreeSurfer 5.3 analysis
• ABIDE_fs5.3_thickness.csv : thickness data from FreeSurfer 5.3 analysis
• abide_fs5.1_landrvolumes.csv : volume data from FreeSurfer 5.1 analysis
• cortical_fs5.1_measuresenigma_thickavg.csv : thickness data from FreeSurfer 5.1 analysis
• subject_check.csv : summary table of the data available per subject

New dataset addition(s)

• Civet: data/ABIDE_civet2.1_thickness.csv (DKT)
• FS: data/fs60_group_stats/* (DKT, Destrieux, Glasser)

Code

Current:

• notebooks: driver code to run analysis
• lib: helper functions for parsing and analysis
• scripts: code to read pipeline output (civet2.1, fs6.0)

.
├──  notebooks           
│   ├── run_atlas_comparisons.ipynb
│   ├── run_pipeline_comparisons.ipynb
│   ├── import_QC_data.ipynb
│   ├── generate_plots.ipynb
│   └── learn_pipeline_transforms.ipynb
└── lib
│   ├── data_handling.py
│   ├── deeplearning.py
│   ├── data_stats.py
│   └── plot_utils.py
└── scripts
    ├── get_vertex_data_fs.py
    ├── get_dkt_data_civet.py
    └── check_vertex_data.py

Legacy:

• see https://github.com/companat/compare-surf-tools for details

Steps (see ./compute_workflow.png)

Prereq: Processed output from a given pipeline (tool): e.g. FreeSurfer

A. Data parsing

1. run scripts/get_vertex_data_fs.py on a FS subject dir to get vertext-wise summay CSV for all subjects.

python get_vertex_data_fs.py -s ../data/subjects/ -k '.fwhm20.fsaverage.mgh' -o ../data/sample_output/fs_fsaverage_vout

2. run scripts/get_roi_data_fs.py on a FS subject dir to get ROI-wise summay CSV for all subjects. Uses aparcstats2table command.

python get_roi_data_fs.py -s ../data/subjects -l ../data/subject_list.txt -m thickness -p a2009s -o ../data/sample_output/

B. Data standardization

C. Comparative analysis

D. Outlier detection