MycorrhizaFinder: an efficient machine learning tool to quantify endomycorrhizal colonisation of real-world roots

O_LIRoot colonisation by endomycorrhizal fungi can indicate habitat condition. However, due to the significant time required to assess colonisation using traditional microscope techniques, studies of colonisation at large scales are impractical. AI-powered approaches may increase output and facilitate ecosystem assessments. C_LIO_LIWe trained our AI-powered tool MycorrhizaFinder (MFKew) on field roots from diverse ecosystems. It was trained to recognise a range of arbuscular and ericoid mycorrhizal fungal structures, and to differentiate dark septate endophytes common in field-sourced roots. C_LIO_LIHere we describe the semi-automated workflow from root processing and microscope slide scanning to model training and performance evaluation, proposing Macro F1 as the appropriate metric to be optimised. Without human supervision, Macro F1 currently stands at 66% for arbuscular and at 57% for ericoid mycorrhizal colonisation assessment. C_LIO_LIMFKew is user friendly, requires no programming skills and offers flexibility for advanced users who wish to further train the tool using their own labelled mycorrhizal root datasets, including images acquired from different devices or staining protocols. This adaptability allows users to customize the model for specific needs, making it optimal for ecologists and agronomists. Additionally, MFKew supports large-scale, repeatable, medium-throughput monitoring across ecosystems, enabling the assessment of mycorrhizal status and tracking changes over time. C_LI