P-402 - Deep learning-based quantification of arbuscular mycorrhizal fungi in plant roots

Arbuscular mycorrhizal (AM) fungi form symbiotic associations with most land plants, improving nutrient uptake and plant health. However, quantifying fungal colonization in roots remains labor-intensive and prone to human error. To address this, we developed Automatic Mycorrhiza Finder (AMFinder), a deep learning-based tool for high-throughput and reproducible quantification of AM fungal colonization (Evangelisti et al., New Phytologist, 2021). AMFinder uses convolutional neural networks to accurately identify colonized root sections and intraradical hyphal structures in multiple plant species (including Medicago truncatula, Lotus japonicus, and Oryza sativa) colonized with different AM fungi (such as Rhizophagus, Claroideoglomus, Rhizoglomus, and Funneliformis). Our method captures altered colonization in plant mutants (e.g., ram1-1, str, smax1) and enables tracking of colonization dynamics over time. Applying deep learning to fungal imaging opens exciting perspectives for plant-microbe interaction studies. AMFinder has already been adopted by multiple research groups investigating endosymbiosis, highlighting its value for the community. In this presentation, I will showcase recent updates, improvements, and future developments of AMFinder, further expanding its applications in the automated imaging of filamentous microbes.