From manual scoring to machine learning: recent developments in image-based arbuscular mycorrhizal fungi root quantification.

Selecting an appropriate method to quantify arbuscular mycorrhizal fungi (AMF) colonization in plants can be challenging, particularly for non-experts, due to the wide diversity of available techniques, some of which date back to the 1980s. Despite rapid methodological advances, comprehensive reviews focusing specifically on recent image analysis approaches for AMF root quantification are still scarce in the scientific literature. Given the growing potential of image-based methods to improve the accuracy, reproducibility, and high throughput of AMF root colonization assessment, this review synthesizes conceptual structure and thematic map of studies focusing on AMF colonization from 2001 to 2026. In addition, we examine widely used traditional approaches, such as the gridline intersect method, and emerging tools, including Visual Basic for Application-Arbuscular Mycorrhizal Fungi (VBA-AMF), MycoPatt, WinRHIZO, ImageJ-based workflows, Zeiss Intellesis, and deep learning approaches such as Automatic Mycorrhizal Finder (AMFinder), "Tool for Analysing root images to calculate the Infection rate of arbuscular Mycorrhizal fungi" (TAIM) and Mask Region-based Convolutional Neural Network (Mask R-CNN). By critically comparing their principles, capabilities, and limitations, this review aims to guide researchers in selecting the most appropriate image analysis methods for AMF quantification according to their experimental objectives and technical constraints.