Detection of oilseed rape clubroot based on low-field nuclear magnetic resonance imaging

Abstract

Plant root diseases threat plant growth and eventually cause plant death without proper treatment. It is difficult to diagnose root diseases without digging the roots from the soil, and it is late when the above-ground parts show symptoms under the stress of root diseases. This study used magnetic resonance imaging (MRI) for non-invasive root phenotyping to detect oilseed rape clubroot. MRI images of healthy oilseed rape roots and roots infected by clubroot were obtained. After image preprocessing, average sample grayscale histograms (Avg-SGH) were extracted to build classification models for disease identification using logistic regression (LR), support vector machine (SVM) and random forest (RF). Reconstruction of three-dimensional (3D) root architectures was also conducted. Root architecture parameters were extracted from the reconstructed roots. Analysis of variance (ANOVA) showed that the root architecture parameters differed significantly between healthy and infected roots. RF model using root architecture parameters showed good performances, and the feature importance for clubroot identification was also explored. The overall results showed that MRI could effectively detect clubroot diseases in a non-invasive manner, indicating significant potential for plant root phenotyping.