MRI-Seed-Wizard: combining deep learning algorithms with magnetic resonance imaging enables advanced seed phenotyping

Area of Science:

• Agricultural Science
• Biotechnology
• Imaging Technology

Background:

• Seed trait evaluation is crucial for plant breeding and biotechnology.
• Current methods lack non-destructive, 3D assessment of internal seed features.
• There's a need for automated, high-throughput seed phenotyping.

Purpose of the Study:

• Introduce MRI-Seed-Wizard, a novel tool integrating deep learning and MRI for plant seed analysis.
• Enable in vivo, non-destructive quantification of seed morphometry and internal composition.
• Automate the analysis of digital MRI data for seed phenotyping.

Main Methods:

• Developed MRI-Seed-Wizard, combining deep learning with magnetic resonance imaging (MRI).
• Implemented advanced MRI protocols for simultaneous multi-seed evaluation.
• Applied the tool to wheat and barley grains for trait quantification.

Main Results:

• Successfully quantified 23 in vivo grain traits, including internal volumetric parameters.
• Demonstrated capabilities beyond conventional techniques like X-ray computed tomography.
• Showcased automated identification, labeling, and analysis of MRI data.

Conclusions:

• MRI-Seed-Wizard offers a versatile, non-destructive method for comprehensive seed phenotyping.
• AI-powered plant MRI significantly enhances crop improvement research.
• The tool is applicable to a wide range of crop seeds, increasing throughput.