Application of a Low-Cost Electronic Nose for Differentiation between Pathogenic Oomycetes Pythium intermedium and Phytophthora plurivora.

Piotr Borowik,Tomasz Oszako,Przemysław Wacławik,Leszek Adamowicz,Rafał Tarakowski,Sławomir Ślusarski,Miłosz Tkaczyk

doi:10.3390/s21041326

Abstract

Compared with traditional gas chromatography–mass spectrometry techniques, electronic noses are non-invasive and can be a rapid, cost-effective option for several applications. This paper presents comparative studies of differentiation between odors emitted by two forest pathogens: Pythium and Phytophthora, measured by a low-cost electronic nose. The electronic nose applies six non-specific Figaro Inc. metal oxide sensors. Various features describing shapes of the measurement curves of sensors’ response to the odors’ exposure were extracted and used for building the classification models. As a machine learning algorithm for classification, we use the Support Vector Machine (SVM) method and various measures to assess classification models’ performance. Differentiation between Phytophthora and Pythium species has an important practical aspect allowing forest practitioners to take appropriate plant protection. We demonstrate the possibility to recognize and differentiate between the two mentioned species with acceptable accuracy by our low-cost electronic nose.

Highlights

The results obtained by the electronic nose measurements are response curves of the sensors array when they are exposed to the studied odor and when they recover to the baseline after exposure to the clear air
We used data collected from all six sensors available in the electronic nose in the first model that we trained
What can be noticed is that the accuracy and other measures calculated by the data collected in the first phase of the experiment, by the cross-validation method, exhibit the high performance of classification of the studied oomycetes, for all considered measures at least 90%

Summary

Introduction

Due to the high labor and costs related to hiring specialized personnel (e.g., gas chromatography, mass spectrometry), they are not widely used in forestry or horticulture applications. For some such methods, the equipment is not portable (like nuclear magnetic resonance or spectrophotometry), and analysis is time-consuming. There is a need for innovative low-cost instruments allowing fast detection of organisms based on sensors detecting volatile organic compounds (VOCs). They should be suitable for on-site monitoring (meaning that the required time of the measurements is short), and their production costs considerably low. Since introducing the e-nose concept [1,2,3], various sensing methods, like optical [4], gravimetric [5], and electrochemical [6], have been developed

Methods

Results

Conclusion