Quantifying root colonization by a symbiotic fungus using automated image segmentation and machine learning approaches.

Abstract

Arbuscular mycorrhizas (AM) are one of the most widespread symbiosis on earth. This plant-fungus interaction involves around 72% of plant species, including most crops. AM symbiosis improves plant nutrition and tolerance to biotic and abiotic stresses. The fungus, in turn, receives carbon compounds derived from the plant photosynthetic process, such as sugars and lipids. Most studies investigating AM and their applications in agriculture requires a precise quantification of the intensity of plant colonization. At present, the majority of researchers in the field base AM quantification analyses on manual visual methods, prone to operator errors and limited reproducibility. Here we propose a novel semi-automated approach to quantify AM fungal root colonization based on digital image analysis comparing three methods: (i) manual quantification (ii) image thresholding, (iii) machine learning. We recognize machine learning as a very promising tool for accelerating, simplifying and standardizing critical steps in analysing AM quantification, answering to an urgent need by the scientific community studying this symbiosis.

Plant phenotyping relevance

植物根の菌根菌コロニー形成という植物状態を、画像閾値処理と機械学習で定量する半自動手法を開発・比較しており、表現型取得・抽出法が中心である。

Code and data availability

The paper's Data Availability statement deposits the analysed root image datasets (mycorrhizal and non-mycorrhizal) on Figshare, directly reproducing the paper's phenotyping inputs for thresholding and machine learning segmentation. The Zeiss GitHub link is a generic third-party algorithm documentation page, not a code