Gas chromatography-ion mobility spectrometric discrimination of trunk borer infested Platycladus orientalis using a novel topographic segmentation strategy

Abstract

Gas chromatography coupled with ion mobility spectrometry (GC-IMS), a volatile analysis technique, has been widely used in the agricultural field over the past decade. However, complex three-dimensional (3D) fingerprint, including millions of data elements, causes significant challenges for its data process. This study proposed a novel topographic segmentation-based strategy, which could automatically extract the maxima and volume features from individual peaks. Compared with the manual marker selection approach, its efficiency was identified using different classification and prediction models based on trunk borer infested Platycladus orientalis samples. The grid search-support vector machine (GS-SVM) classifier combined with a topographic segmentation strategy could correctly classify at least 93.67% of P. orientalis samples into their corresponding infestation stages. The volume feature performed best, and its cross-validation classification accuracy could reach 97.05%. The PLSR prediction model based on the volume feature got the most satisfactory result, whose Rc2 = 0.9624 and RMSEC = 6.328 in calibration set and Rp2 = 0.9056 and RMSEP = 9.926 in validation set. In a word, our proposed topographic segmentation strategy had enough capability to extract local maxima and volume features of peaks from GC-IMS fingerprints. The GC-IMS-based approach combined with chemometric methods had the potential to discriminate the infestation stages of trunk borer damaged P. orientalis plants.