A hyperspectral phenotyping platform for studying plant responses to Xylella fastidiosa infection

Abstract

The plant pathogenic bacterium Xylella fastidiosa is able to infect a wide range of hosts. Its long incubation period and frequent asymptomatic infections hinder early visual detection, allowing infected plants to remain undetected facilitating disease spread. This paper describes a hyperspectral imaging-based phenotyping platform for early, non-destructive assessment of woody plant responses to X. fastidiosa infection under controlled conditions. The platform includes a vertical-scanning structure that integrates controlled illumination with a push-broom VNIR hyperspectral camera, and is supported by an automated hyperspectral image processing pipeline. This workflow includes calibration, spectral smoothing, supervised machine-learning-based image segmentation, and extraction of pure-vegetation spectral signatures, ensuring quantitative and reproducible spectral data. The platform performance was evaluated using asymptomatic olive cultivars inoculated with two X. fastidiosa strains. The resulting spectral signatures showed characteristic vegetation reflectance patterns, along with subtle but consistent differences from non-inoculated control plants, and for qPCR-positive and qPCR-negative X. fastidiosa inoculated plants. These differences were supported by univariate and multivariate statistical analyses, which confirmed a significant treatment-related spectral effect. The developed system can provide quantitative evidence that hyperspectral imaging can capture pre-symptomatic physiological alterations associated with X. fastidiosa inoculation status. This platform is a robust and versatile tool for high-throughput phenotyping, disease monitoring, and resistance screening in olive, and is readily transferable for use on other woody crop hosts. It has strong potential to support future research on vascular plant diseases and pathogen resistance plant breeding programmes.