Neuromodex VNet DBS

Deep‑learning utilities for DBS workflows, including MRI brain tissue segmentation (VNet) and conductivity mapping. This repository provides a Python package and optional 3D Slicer modules to integrate the models into imaging workflows.

Note: This project is still under development.

Features

• VNet‑based multi‑class brain tissue segmentation
• Conductivity mapping utilities
• Pre/post‑processing
• PyTorch inference with automatic device selection (CPU/GPU)
• 3D Slicer plugin scaffolding for GUI‑based use
• The segmentation model was trained using labels generated with ELMA, a semi‑automatic DBS tissue classification/segmentation tool (commonly used to classify tissues such as grey matter, white matter, blood, and CSF for patient‑specific DBS FEM modeling workflows). [1] [2]

Installation

Requirements: Python 3.9+

pip install neuromodex-vnet-dbs

The wheel bundles the neuromodex_vnet_dbs/weights/ directory so the packaged models can load without any extra downloads.

Quick start Segmentation

Note that the input image must be skull stripped and preferably N4 bias corrected. Good choices are HD_BET and SimpleITK's N4BiasFieldCorrection.

import SimpleITK as sitk
from neuromodex_vnet_dbs import SegmentationPipeline

# Load an input image (e.g., NIfTI)
img = sitk.ReadImage("/path/to/volume.nii.gz")
# or
img = "path/to/volume.nii.gz"

# Run the segmentation pipeline
pipe = SegmentationPipeline(img)  # pass either as string or sitk volume
result = pipe.segment_fast(img)  # ~7 seconds

# or this for clearer csf segmentation
result = pipe.segment_gmm_csf(img)  # ~1.5 minutes

# The returned object is the segmented image

You can also use the segment_vnet function to segment a volume with a single command:

from neuromodex_vnet_dbs.easy_segment import segment_vnet

segmented = segment_vnet("path/to/mri_image.nii.gz")

Quick Start Conductivity Mapping

import SimpleITK as sitk
from neuromodex_vnet_dbs import ConductivityProcessingPipeline

mri_img = sitk.ReadImage("path/to/mri_image.nii.gz")
seg_img = sitk.ReadImage("path/to/seg_image.nii.gz")

pipe = ConductivityProcessingPipeline(seg_img, mri_img)
result = pipe.run()

Or use the easy wrapper:

from neuromodex_vnet_dbs.easy_conductivity_mapping import map_conductivities

conductivities = map_conductivities("path/to/mri_image.nii.gz", "path/to/seg_image.nii.gz")

3D Slicer integration

This repo includes helper scripts and example module folders under slicer/. These scripts can also be used as CLI tools.

• To install one or more module folders into your local Slicer profile, run either:

from neuromodex_vnet_dbs.slicer import interactive_installation

interactive_installation()

or

python neuromodex_vnet_dbs/slicer/slicer_install_plugin.py

Follow the prompts to choose the plugin(s) and target Slicer installation. Restart Slicer afterwards.

The plugins can then be found in Segmentation/BrainSegmentation and Electrical Conductivity/ConductivityMapping.

Note: The first execution in slicer will take a few minutes to download all the required data.

License

This project is licensed under the terms of the MIT License. See the LICENSE file for details.

Citation

If you use this project in your research, please cite the appropriate papers for VNet and any downstream methods you apply.