VRoot is an open-source software designed for the immersive root system architecture reconstruction from MRI scans. Often times, automated methods still fall short of the relatively high accuracy demand that is required for downstream simulation of root system dynamics. VRoot provides a user-friendly VR environment that allows researchers to manually trace and reconstruct root architectures with high precision, facilitating advanced plant phenotyping and modeling.
The VRoot framework is described in detail in the 2025 Plant Phenomics article, which demonstrates its capabilities as annotation application in a controlled setting using novice users.
VRoot has already proven itself to be effective prior to our systematic analysis of user behavior: in a 2023 study, Selzner et al. used VRoot as manual annotation baseline to compare the impact of 3D segmentation on root reconstruction accuracy.
Key Features
- Coupling possibilities: Use any pre-traced root system architecture from tools such as NMRooting or Toporoot
- Deep Learning Integration: The workflow is based on Python and is extendible for deep learning.
- High-Accuracy processing: VRoot enables highly accurate tracing, including the provision of common toolsets
- Flexible Input Support: Compatible with a range of imaging modalities, including X-ray CT, MRI, and optical imaging.
- Open Source & Extensible: Freely available and designed for integration with other plant modeling and visualization tools.
- Runs on many systems: VRoot runs as direct VR link or as standalone application on devices such as Quest - and my PC is not even that good.
How VRoot Works
VRoot is an immersive visualization tool that allows users to interactively explore and reconstruct root architectures in a virtual reality environment. The software processes 3D imaging data to provide an immersive experience supporting manual and semi-automated tracing of root structures.
The core design process of VRoot was always centered around usability for a broad range of users, critically to accomodate the workflows of annotation experts as well as novice users, both of which are in abundance in most research groups. The learning curve is minimal and users can achieve high accuracy (>0.95 median F1 score) after a very brief introduction.