Frost is one of the most economically damaging abiotic stresses affecting Australian wheat production, causing significant yield losses through pollen sterility, floret abortion, and poor grain set during reproductive development. Understanding the spatial distribution of frost damage within wheat spikes is critical for identifying tolerant germplasm; however, current phenotyping methods rely on destructive, labour-intensive manual assessments that limit throughput and consistency.
At The Plant Accelerator (Australian Plant Phenomics Network, Adelaide University), we have developed an automated X-ray computed tomography (CT) platform for non-destructive, high-throughput phenotyping of wheat grain morphology. This project extends the platform to reconstruct complete wheat spike architecture in 3D and automatically quantify reproductive organs that are difficult to assess manually, including rachis nodes, spikelets, and sterile florets.
The workflow comprises CT data preprocessing, manual collection of ground-truth phenotypic traits, 3D annotation of spike organs, and development of an AI-based segmentation pipeline for automated identification of grains, rachis nodes, spikelets, and sterile florets. Deep learning models were trained using manually annotated datasets and validated against ground-truth measurements to evaluate segmentation accuracy and trait extraction performance.
The resulting pipeline enables automated quantification of wheat spike architecture, including rachis node number, spikelet number, grain distribution, and the spatial occurrence of sterile florets. Integrating these measurements with existing CT-derived grain traits provides a comprehensive, non-destructive assessment of frost-induced reproductive damage at high throughput.
This work demonstrates how X-ray CT, advanced image analysis, and artificial intelligence can be integrated to deliver next-generation plant phenotyping. Beyond enabling objective and scalable measurements of frost damage, the developed methodology establishes a framework for automated structural analysis of complex plant organs, supporting the development of frost-tolerant wheat germplasm and advancing climate-resilient crop breeding.