Non-invasive evaluation of Ascochyta blight disease severity in chickpea using field asymmetric ion mobility spectrometry and hyperspectral imaging techniques

Evaluation of disease severity of breeding materials to a disease is a key step to develop disease-resistant cultivars in plant breeding. Visual evaluation that requires intensive training and experience can be subjective, though it is used as a common practice in plant breeding. Optical sensing techniques have been evaluated for disease monitoring with encouraging results, however, it is limited by factors, such as illumination, architecture of plants, and disease types. In this regard, volatile organic compounds (VOCs)-based sensing techniques can serve as an alternative solution. In this study, a field asymmetric ion mobility spectrometry (FAIMS) system was evaluated as a rapid VOCs-based phenotyping tool for monitoring disease severity in chickpea. Chickpea cultivars with three levels of resistance to Ascochyta blight (Ascochyta rabiei) were cultivated in greenhouse and inoculated with pathogen spores or treated with sterilized water (control). The VOC sensing system with a FAIMS system and a customized sampling chamber was used to acquire VOC profiles of chickpea plants. Meanwhile, visual ratings (1–9 rating scale) and hyperspectral images of chickpea plants were collected and used as reference data. Features were extracted from FAIMS and hyperspectral data, respectively, and used as inputs to develop machine learning models (discriminant analysis, support vector machine, and kernel classification model) to predict the disease severity rating. When using one ion current matrix from all samples for model development, prediction accuracy of classification models with FAIMS data were lower than those based on features from hyperspectral images. When three matrices from all (control and inoculated) samples or data from inoculated samples were utilized, the overall prediction accuracy of classification models with FAIMS data were higher (up to 94%) than those developed with only one matrix. The study demonstrated that FAIMS system can be used as a VOCs-based phenotyping tool to evaluate disease severity rapidly and illustrated comparable performance as hyperspectral imaging without destructive sampling. With further improvement on sampling and instrumentation, the FAIMS system can be a useful tool for monitoring disease and assisting in evaluating new cultivars for desired resistance to the diseases.