Habenula fMRI ROIs

Generates subject-level habenula (Hb) regions of interest (ROIs) with high BOLD sensitivity for use in fMRI studies. Hb ROIs are generated at functional resolution (e.g. 2mm isotropic) based on an input Hb segmentation at anatomical resolution (e.g. 0.7mm isotropic). Optionally, output Hb ROIs can be warped to a desired template space (e.g. MNI). Note that the main method (Habenula_fMRI_ROIs.sh) is an updated version of the shape optimization method described in (Ely BA et al. 2019 NeuroImage), modified to allow arbitrary input anatomical/output functional resolutions and greater control over output ROI thresholding/laterality/etc (for original 2019 version see Legacy Method #3).

---

Requirements

• FSL version 5.0.6 or later
• Unix OS (Mac OSX or Linux)

---

Input Files

• Left and Right Hb segmentations at anatomical resolution (can be single Bilateral or separate Left/Right files)
  • We recommend automated probabilistic Hb segmentation if T1w and T2w anatomical data are available (Kim JW et al. 2016 NeuroImage; https://github.com/junqianxulab/habenula_segmentation)
  • Works best with probabilistic Hb segmentation inputs but can also accept binary Hb masks (e.g. manual segmentations a la Lawson RP et al. 2013 NeuroImage; https://github.com/junqianxulab/habenula_segmentation_lawson)
  • Intensity values should range from 0-1 (probabilistic) or equal 1 (binary); inputs exceeding this range will be rescaled
  • These will be treated as "native" subject space
• Example file at the target functional resolution and XYZ dimensions (i.e. NIFTI header parameters match desired output)
• [Optional] FNIRT-compatible relative warpfield files for template-to-native-space transformation
  • If using HCP pipelines, located at: <subject_directory>/MNINonLinear/xfms/standard2acpc_dc.nii.gz'

Input files should be in standard NIFTI format (.nii or .nii.gz) and can be given with or without filetype (i.e. file or file.nii.gz both work).

---

Output Files

• Thresholded (default 0.25) and optionally binarized (bin_) shape-optimized Hb ROIs at target functional resolution
  • This is the recommended ROI for use in fMRI studies (see Ely BA et al. 2019 NeuroImage)
  • Default output is bilateral (_B), optionally also outputs separate left/right (_L/_R)
  • <subject_ID>_Hb_ROI_ShapeOpt_thr_<bin_><B/L/R>.nii.gz
• Full probabilistic (weighted 0.01-1) ROIs of the Hb region at target functional resolution
  • <subject_ID>_Hb_ROI_ShapeOpt_full_<B/L/R>.nii.gz

Intermediate and log files in working directory can be deleted after ROI generation, if desired.

---

How to Run

Options

Habenula_fMRI_ROIs.sh
# Mandatory inputs:
	--sub    = subject ID
	--func   = example functional image file
	--odir   = output directory (can be absolute or relative path, will be created if not found)
	--segB   = bilateral Hb segmentation file
		OR
	--segL   = left Hb segmentation file
	--segR   = right Hb segmentation file
# Optional inputs:
	--warp   = FNIRT  warpfield, if included ROI position will be transformed accordingly
	--wdir   = working directory (created inside --odir by default)
	--thresh = final ROI threshold, default = 0.25 (works well for HCP data)
# Optional flags:
	--bin    = binarize after thresholding
	--quiet  = don't print to stdout (still logs)
	--LRout  = output separate left/right ROIs in addition to bilateral

Example calls from the unix terminal:

Create default bilateral shape-optimized Hb ROIs in "native" space (i.e. same position as segmentation)

Habenula_fMRI_ROIs.sh \
	--sub=sub_001 \
	--segB=bilateral_Hb_segmentation.nii.gz \
	--func=example_functional.nii.gz \
	--odir=output_directory

Create bilateral + separate left/right Hb ROIs in template space

Habenula_fMRI_ROIs.sh \
	--sub=sub_001 \
	--segB=bilateral_Hb_segmentation.nii.gz \
	--func=example_functional.nii.gz \
	--odir=output_directory \
	--warp=warpfield.nii.gz \
	--LRout

---

Templates

A few other potentially useful files:

• 2mm T1w MNI template brain: NIFTI header dimensions match 3T HCP fMRI so can be used as --func target for those datasets

  • MNI152_T1_2mm_brain.nii.gz

• Thalamus ROIs: Bilateral ROIs located anterior to and lateral to the Hb in MNI space. ROIs are non-specific but approximately correspond to dorsomedial (anterior) and centromedian (lateral) thalamic nuclei. Can be used as "control" ROIs/to regress out nearby thalammic signals during analysis. Derived by averaging Volume Optimized Hb ROIs (see Legacy Methods below) from 50 HCP subjects and shifting 6mm (3 voxels) anteriorly/laterally (Ely BA et al. 2016, Human Brain Mapping).

  • Thalamus_control_ROI_anterior_to_Hb_from_Ely_HBM_2016.nii.gz
  • Thalamus_control_ROI_lateral_to_Hb_from_Ely_HBM_2016.nii.gz

---

Legacy Methods

Original six methods for creating Hb fMRI ROIs evaluated in (Ely BA et al. 2019 NeuroImage). Method #3 (Hb_shape_optimized_ROIs_2mm.sh) had the highest Hb BOLD sensitivity (see above). Other methods are provided for completeness only; not recommended. Note that these scripts use input Hb segmentations created in MNI template space rather than native subject space (also not recommended) and are configured to optionally run in parallel in a LSF-based cluster computing environment.

1: Unoptimized Hb ROIs

Basic nearest-neighbor downsampling of binary Hb segmentation to functional resolution.

Hb_unoptimized_ROIs.sh

2: Volume Optimized Hb ROIs

Conservatively defined to match the estimated volume of the segmented Hb, as described in (Ely BA et al. 2016 Human Brain Mapping).

Hb_volume_optimized_ROIs_2mm.sh

3: Shape Optimized Hb ROIs

Liberally defined to approximate the outer boundaries/overall shape of the segmented Hb.

Hb_shape_optimized_ROIs_2mm.sh

4: Mean Timeseries Optimized Hb ROIs

Includes voxels most strongly correlated with the mean timeseries of the bilateral Shape Optimized Hb ROI. Volume matched to the segmented Hb, as in method 2. Requires Matlab + NifTI_tools package.

Step 1: extract voxelwise and mean timeseries values from denoised rs-fMRI data within bilateral Hb Region ROI

extract_voxelTS.sh

Step 2: assign weight to each voxel based on correlation with the mean timeseries

Hb_mean_timeseries_optimized_ROIs.m

Step 3: copy correct header information from an example functional file (minor bugfix for matlab)

correct_matlab_headers.sh meanTSopt

Step 4: match final ROI volume to estimated volume of the segmented Hb

threshold_timeseries_optimized_ROIs_2mm.sh meanTSopt

5: Pairwise Timeseries Optimized Hb ROIs

Includes voxels with strongest pairwise correlations within the Shape Optimized Hb ROI. Volume matched to the segmented Hb, as in method 2. Requires Matlab + NifTI_tools package. Step 1: extract voxelwise and mean timeseries values from denoised rs-fMRI data within bilateral Hb Region ROI

extract_voxelTS.sh

Step 2: calculate timeseries correlation of each voxel with each other voxel and assign weight based on strongest correlation

Hb_pairwise_timeseries_optimized_ROIs_2mm.m

Step 3: copy correct header information from an example functional file (minor bugfix for matlab)

correct_matlab_headers.sh pairTSopt

Step 4: match final ROI volume to estimated volume of the segmented Hb

threshold_timeseries_optimized_ROIs_2mm.sh pairTSopt

6: Hb Template ROI

Generic ROI created by averaging Hb segmentations from 68 HCP subjects then downsampling to functional resolution, with volume matched to the averaged segmented Hb volume, similar to method 2.

Hb_template_ROI_2mm_MNI.nii.gz

---

References:

[1] Ely BA et al. 'Detailed mapping of human habenula resting-state functional connectivity', NeuroImage 2019 200:621-634, https://pubmed.ncbi.nlm.nih.gov/31252057

[2] Ely BA et al. 'Resting-state functional connectivity of the human habenula in healthy individuals: associations with subclinical depression', Human Brain Mapping 2016 37:2369-84, https://www.ncbi.nlm.nih.gov/pubmed/26991474

[3] Kim JW et al. 'Human habenula segmentation using myelin content', NeuroImage 2016 130:145-56, http://www.ncbi.nlm.nih.gov/pubmed/26826517

[4] Kim JW et al. 'Reproducibility of myelin content-based human habenula segmentation at 3 Tesla', Human Brain Mapping 2018 39:3058-71, https://pubmed.ncbi.nlm.nih.gov/29582505

[5] Lawson RP et al. 'Defining the habenula in human neuroimaging studies', NeuroImage 2013 64:722-7, https://www.ncbi.nlm.nih.gov/pubmed/22986224

If using these scripts, please cite [1]