Autoroot: automatic 3D root phenotyping tool for non-destructive analysis of plant roots in soil

Overview

autoROOT is an advanced imaging and analysis platform designed to non-destructively phenotype plant roots in soil, addressing a critical bottleneck in breeding climate-resilient crops. Unlike most existing root phenotyping methods that rely on soilless growth media, autoROOT preserves the natural soil-root interaction by using optimised x-ray microtomography (mCT) scanning paired with automated 3D root segmentation. The solution builds on the team's existing supervised segmentation tool, ROOT1, and advances it with machine learning to deliver unsupervised, accurate, and rapid root extraction. Potential applications include crop breeding programmes, root trait selection for drought and stress tolerance, soil-root interaction studies, and agricultural research across multiple crop species.

Technical specifications

Key features:

• Automated 3D root segmentation from mCT scan images, reducing the need for manual or semi-automatic processing
• Machine and deep learning image segmentation algorithms integrated into the ROOT1 pipeline, developed in ImageJ as autoROOT
• Optimised mCT scanning parameters that reliably measure root angle, diameter, depth, and root-length density while reducing scan time
• Capability to distinguish roots, pores, and soil structures through trained classification models
• Validated across crops with contrasting root architecture, including chickpea, wheat, and maize
• Tested across multiple soil types, including loam, clay, and sandy soils
• Skeletonisation-based root trait measurement within ImageJ for quantitative phenotyping

Technology readiness level

autoROOT is at an early-to-mid development stage. Pilot experiments have demonstrated that coupling machine or deep learning segmentation tools with ROOT1 can automatically extract 3D roots from mCT images. Ongoing work focuses on reducing noise, improving workflow efficiency, and training the algorithm on diverse soil-root samples. Future validation will involve glasshouse-grown crops in different soil types scanned with varied mCT settings, with autoROOT outputs benchmarked against manually and semi-automatically segmented roots. The technology is positioned for further development and refinement through collaborative research before broader commercial deployment.