Self-supervised deep learning model for ploidy detection in wheat
Developing a self-supervised deep learning model using leaf surface imaging to distinguish ploidy levels in wheat. This approach leverages morphological features and is validated by flow cytometry and fertility tests, aiming for high-throughput applications in breeding programs.
Overview
This innovative solution focuses on a self-supervised deep learning model designed to differentiate haploid from diploid wheat seedlings through leaf surface imaging. By utilizing high-resolution photographs of seedling leaves, the model analyzes morphological features such as stomatal dimensions and leaf venation, which change with nuclear ploidy variations. The model is trained using flow cytometry-validated samples, with predictions refined by empirical ploidy feedback. This technology can be integrated into automated and high-throughput wheat breeding applications, improving efficiency and reducing manual screening.
Technical specifications
- Utilizes high-resolution imaging for detailed analysis of leaf surface morphology
- Morphological features include stomatal and trichome dimensions, and leaf venation
- Employs a combination of feature-based random forest classifiers and convolutional neural networks (CNN) for classification
- Training involves a dataset validated with flow cytometry and fertility tests
- Compatible with standard growth facilities and imaging platforms for seamless integration into existing systems
Technology readiness level
The technology is at TRL 4, indicating that it has been validated in a lab environment. Future validation will involve testing on anonymous seed batches, with plans to grow and type at least 1000 plants. The model's specificity will be assessed on individuals not used in the training dataset, with potential testing across variable genetic backgrounds.
About Wageningen University and Research
Wageningen University & Research is a life sciences–focused public university combined with mission-driven research institutes, integrating fundamental, applied, and field-based R&D at scale. On Wageningen Campus, companies co-locate with WUR teams and use shared pilot plants, advanced analytical labs, and controlled‑environment facilities for rapid prototyping. Being in the Food Valley cluster provides tight links to global corporates, SMEs, and startups through living labs and a strong regional talent pipeline. Research is backed by competitive funding from the Dutch Research Council, European Union programs, and national ministries, alongside significant contract research for industry. A dedicated technology transfer office and campus incubators streamline IP, licensing, and spin‑out formation for corporate partners and entrepreneurs.
delivered to your inbox.