Detection of Mushroom Virus X (MVX) Infection in Asymptomatic Mushrooms Using FTIR Microscopic Imaging
Document Type Conference Paper
Alvarez-Jubete L, Bonnier F, Byrne H, Grogan H, Frias JM (2011) Detection of mushroom Virus X (MVX) infection in asymptomatic mushrooms using FTIR microscopic imaging. Poster Presentation at the 11th International Conference of Engineering and Food (ICEF11), May 2011, Athens, Greece
Abstract
Mushroom Virus X affects important traits associated with mushroom quality including colour and appearance. As a result, the spread of Mushroom Virus X to mushroom crops may potentially result in devastating economical effects for mushroom growers and producers. To prevent cross-contamination from occurring, it is essential than MVX infected crops can be readily identified. At present, the only valid method available to confirm MVX infection is based on PCR technology, by detecting the presence of viral dsRNA. However, this method is time-consuming and requires highly skilled personnel. FT-IR spectroscopy has been used successfully in several studies to measure key parameters associated with mushroom quality. In this study, we investigated the use of FT-IR imaging spectroscopy as a rapid method to effectively detect the presence of Mushroom Virus X in asymptomatic mushroom tissue. The microscopic spectral images corresponding to the different mushroom tissues were analysed by k-means to section the images into the different morphological regions of interest. Discrimination models of MVX infection were built using Random Forest (RF) and Partial Least Squares Discriminant Analysis (PLS-DA) classification techniques. Spectra from the stalk were found to be most efficient for identifying MVX infection, with 98% correct discrimination of the test subset through RF. Regarding the cap tissue, spectra from the surface layer (pileipellis) proved to be the location with the best ability to discriminate between infected and non-infected mushrooms, with a 93% correct classification of the test subset through RF. In general RF performed better than PLS-DA in predicting the test subset.