PhenoTrack3D: an automatic high-throughput phenotyping pipeline to track maize organs over time

Area of Science:

• Plant Science
• Computational Biology
• Genetics

Background:

• High-throughput phenotyping platforms enable studying genotype-by-environment (GxE) interactions.
• Automated image analysis pipelines are crucial for processing large datasets from controlled and field conditions.
• Accurate capture of 3D plant architecture and organ development over time is essential for understanding plant growth.

Purpose of the Study:

• To develop and validate PhenoTrack3D, a novel pipeline for 3D+time reconstruction and organ-level tracking in maize.
• To enable high-throughput phenotyping of maize plant architecture and development throughout the entire growth cycle.
• To facilitate GxE analyses by providing detailed, automated organ development data.

Main Methods:

• Utilized existing 3D segmentation methods (e.g., Phenomenal) to prepare time-series plant data.
• Developed a deep-learning-based stem detection method for precise leaf separation.
• Implemented a novel multiple sequence alignment algorithm for temporal tracking of ligulated leaves.
• Employed a distance-based approach for tracking growing leaves.

Main Results:

• PhenoTrack3D successfully reconstructed maize plant architecture in 3D+time.
• Achieved precise stem tip detection (RMSE < 2.1 cm) and high accuracy in tracking leaf ranks (97.7% ligulated, 85.3% growing).
• Extracted organ-level development and architecture traits with good correlation to manual observations.

Conclusions:

• PhenoTrack3D offers a novel, automated, high-throughput phenotyping method for maize.
• The pipeline effectively characterizes maize architecture development at the organ level.
• Validated on extensive datasets, it is applicable for large-scale GxE studies in maize.