Installation

Background. HABIT is a Python package that splits tumours into habitats (sub-regions whose voxels behave alike across images) and quantifies them.

Purpose. At the end you have a conda environment named habit in which habit --version prints the installed version.

Key terms.

  • conda environment – a separate folder with its own Python and packages; conda activate habit switches the terminal to it, so HABIT’s dependencies do not clash with other software.

This page is for absolute beginners on Windows (macOS / Linux notes where they differ). HABIT must run in a conda-activated terminal. On Windows use Anaconda Prompt or Anaconda PowerShell Prompt (Miniconda: Miniconda Prompt). System Python without conda activate often fails with habit: command not found or uses the wrong interpreter.

Python 3.10–3.14 are supported; these docs use 3.10.

1. Install Miniconda

If you already have Anaconda or Miniconda, skip to step 2.

  1. Open the official downloads page: Miniconda.

  2. Check Settings → System → About → System type. Most PCs are 64-bit; download the Windows installer (.exe) that matches that type.

  3. Run the installer and accept the defaults (Next → Next → Install).

You do not need Visual Studio Code for HABIT.

2. Open the conda terminal

Windows (Anaconda / Miniconda installed):

  1. Click the Start button (Windows logo), or press the Windows key. On Windows 10 the Start icon is usually at the bottom-left; on Windows 11 it is often at the bottom-center of the taskbar.

  2. Open the Anaconda3 (or Miniconda) folder in the Start menu, then click Anaconda Prompt or Anaconda PowerShell Prompt (Miniconda: Miniconda Prompt).

  3. Or: with Start open, type Anaconda Prompt / Miniconda Prompt and open that app.

  4. Optional: Anaconda Navigator → Environments → Open Terminal on the habit env after you create it.

Windows Start menu: open Anaconda Prompt

Start → Anaconda3 → Anaconda Prompt / Anaconda PowerShell Prompt (example on Windows 10; on Windows 11 the Start icon may sit at the bottom-center). Your folder name may differ (e.g. Miniconda).

After you open the Prompt you often see (base). That is normal.

macOS / Linux — open Terminal.app / your shell. If conda is not found, run conda init once for your shell and reopen the terminal.

3. Create and activate the habit env

In the conda terminal:

conda create -n habit python=3.10 -y
conda activate habit

The env name in these docs is ``habit`` (any name is fine if you activate it). The prompt must show (habit) before any pip or habit command. Every new terminal session: run conda activate habit again.

4. Install HABIT from official PyPI

With (habit) showing in the prompt:

pip install -U pip
pip install habitat-analysis -i https://pypi.org/simple
habit --version

5. Optional: PyTorch GPU

Needed only for GPU texture / TorchRadiomics. CPU habitat analysis does not require it.

Open https://pytorch.org/ → Get Started (or Start Locally). Select Your OS, Pip, Python, and the Compute Platform that matches your machine, then run the command the page prints.

PyTorch Start Locally installer: pick OS, Pip, Python, and CUDA

Example on pytorch.org: Windows + Pip + Python + CUDA 12.6. Change the red cells to match your GPU or choose CPU. Copy Run this Command into the (habit) prompt.

6. Optional: PyRadiomics

Not installed by default. Needed for voxel_radiomics / supervoxel_radiomics, habit radiomics, and texture habitat features. Habitat maps that do not use those extractors do not need it.

Windows — use a prebuilt wheel from Release v1.0.2 (do not use bare pip install pyradiomics):

pip install https://github.com/lichao312214129/HABIT/releases/download/v1.0.2/pyradiomics-3.1.0-cp310-cp310-win_amd64.whl

Change cp310 in the filename for Python 3.11–3.14.

macOS / Linux:

pip install "pyradiomics>=3.0.1,<3.2"

Optional packages

These do not change habitat computation. Install only what you use.

pyarrow — write a direct pooling voxel table with on the order of millions of rows as parquet. Typical two-step / one-step tables are small; CSV needs nothing extra. Saving the table is optional:

pip install pyarrow

napari — interactive habit view:

pip install "napari[pyqt5]" "npe2>=0.8.2" "pydantic!=2.11.*,>=2.8,<3" -i https://pypi.org/simple

GPU texture speedup numbers: Voxel texture and GPU.

From source (developers)

git clone https://github.com/lichao312214129/HABIT.git
cd HABIT
conda create -n habit python=3.10 -y
conda activate habit
pip install -U pip
pip install -e .
habit --version

Problems?

See FAQ. After install: Quickstart: run the demo (YAML + CLI) or Quickstart: Python API.