Detection of Citrus Tristeza Virus in Mandarin Orange Using a Custom-Developed Electronic Nose System
Autor: | Utpal Sarma, Babak Montazer, Rajdeep Choudhury, Manash Protim Goswami, Sudipta Hazarika, Subhash Medhi |
---|---|
Rok vydání: | 2020 |
Předmět: |
biology
Electronic nose Computer science business.industry 020208 electrical & electronic engineering Citrus tristeza virus Pattern recognition 02 engineering and technology Olfaction biology.organism_classification Ensemble learning Mandarin Chinese language.human_language 0202 electrical engineering electronic engineering information engineering language Artificial intelligence Electrical and Electronic Engineering business Instrumentation |
Zdroj: | IEEE Transactions on Instrumentation and Measurement. 69:9010-9018 |
ISSN: | 1557-9662 0018-9456 |
DOI: | 10.1109/tim.2020.2997064 |
Popis: | Olfaction is one of the primary senses of a living organism and has been used as first aid for inspecting freshness, edibility, and the quality of food. This article addresses a technique where the biological olfaction process has been mimicked by electronic nose (E-Nose) to detect a pathogen named Citrus Tristeza Virus (CTV) in Khasi Mandarin Orange plants. The proposed technique may be used as an alternative to currently used traditional serological or molecular tests, which requires costly and elaborate laboratory infrastructures. In this article, leaves from 62 plants were collected and examined for viral infection using the gold-standard polymerase chain reaction test. A commercial E-Nose system, Alpha MOS Fox 3000, was used for hypothesis testing, resulting in an accuracy of 95.30% using random forest classifier. These results were used to identify the most prominent sensors and their target gases responding to the CTV infection and were considered as a base for sensor selection for custom developing a prototype. The prototype was designed based on simulations for optimum airflow in the sensor chamber. Bootstrap ensemble of $k$ -nearest neighbors and adaptive boost ensemble of decision tree classifier distinguished the data generated from the prototype with an accuracy of 99.36% and 97.58%, respectively. The performance of the classifiers was visualized graphically using dimensionality reduction techniques. These results indicated that the E-Nose technique employed could differentiate the healthy and infected samples and the prototype could serve as an effective first-hand diagnosing tool for large-scale orchards. |
Databáze: | OpenAIRE |
Externí odkaz: |
Pro tento záznam nejsou dostupné žádné jednotky.