A user-friendly machine-learning program to quantify stomatal features from fluorescence images

Abstract

In nearly all plants, pores on the leaf surface called stomata are essential for photosynthesis and gas exchange. The shape and distribution of stomata on the leaf varies widely between plants and is directly connected to photosynthetic efficiency. However, our understanding of the factors, both genetic and environmental, that exert subtle but significant effects on stomatal morphology is limited by the time required to manually annotate stomata in large imaging datasets. Here, we present a lightweight and efficient tool, QuickSpotter, for semi-automated stomatal annotation from fluorescence images. First, we establish QuickSpotters ability to automatically and accurately annotate mature stomata across developmental time. We also introduce an optional, speedy proofreading utility, StomEdit, that allows the researcher to quickly validate and correct machine-generated annotations. We use QuickSpotter and StomEdit to quantify how stomatal morphology evolves at the population level during cotyledon development and demonstrate how the programs can be used to extract subtle differences in stomatal development following pharmacological treatments. Finally, we describe PairCaller, a pair-calling classifier that accompanies QuickSpotter and can be used to identify stomatal clusters, a physiologically relevant and widely studied developmental phenotype. Taken together, our suite of programs facilitates quantitative analyses of stomatal development at scale, enabling high-throughput analyses of leaf phenotypes under varied conditions.

Plant phenotyping relevance

蛍光画像から気孔形態・分布を半自動抽出するソフトウェア群を開発し、精度検証と植物表現型への適用を行っており、フェノタイピング手法が研究の中心である。

Code and data availability

The supplied blocks describe QuickSpotter, StomEdit, and PairCaller, plus training/validation image datasets, but contain no explicit code/data availability statement, deposit, or authors' public URL for any program, model, dataset, or images. Only the bioRxiv DOI itself is present, which is the article, not a paper-­‐

No evidence-backed public reproduction asset is currently recorded.