Pathogen Detection in Khasi Mandarin Orange using Serological and Electronic Nose Diagnostic Technique
Sudipta Hazaika1, Rajdeep Choudhury2, Sarat Saikia3, Utpal Sarma4

1Sudipta Hazarika*, Dept. of Instrumentation & USIC, Gauhati University, Guwahati, Assam, India.
2Rajdeep Choudhury, Dept. of Instrumentation & USIC, Gauhati University, Guwahati, Assam, India.
3Sarat Saikia, Horticultural research station, Assam Agricultural University, Guwahati, Assam, India.
4Utpal Sarma, Dept. of Instrumentation & USIC, Gauhati University, Guwahati, Assam, India.
Manuscript received on January 15, 2020. | Revised Manuscript received on January 22, 2020. | Manuscript published on February 10, 2020. | PP: 2981-2985 | Volume-9 Issue-4, February 2020. | Retrieval Number: D2077029420/2020©BEIESP | DOI: 10.35940/ijitee.D2077.029420
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (

Abstract: The inherent ability of most living organisms to perceive their immediate environment based on sensory responses has immensely contributed to their survival in the harshest of conditions. Animals rely on their olfactory sense to assess the quality of food before intake. This paper addresses a technique of using the electronic nose for distinguishing Khasi Mandarin orange plants infected by a virus called Citrus Tristeza Virus (CTV) in terms of their degree of infection. Leaves from 16 plants were collected and, tested for CTV infection using the standard serological test, Enzyme-linked Immunosorbent Assay (ELISA), prior to electronic nose (e-nose) analysis. Essential oil was extracted from the leaves using hydro distillation and the extracted oils were analyzed with commercial e-nose system Alpha MOSFOX 3000 system. Bootstrapped ensemble of support vector classifier was used for classifying the samples. The classifier model was optimized with the best parameters and a kernel specific performance evaluation was done for finding out the best model for classification. Among the linear, radial basis function and polynomial kernels, the linear kernel of the classifier performed the best among all the kernels with an accuracy of 97.67% and a Cohen’s Kappa score of 95.25%. Dimensionality reduction techniques like principle component analysis and linear discriminant analysis were also used for graphical visualization of the classification boundaries. The dimensionally reduced dataset was also fitted to the optimized bootstrap ensemble support vector classifier and the performance of the classifier was analyzed. The performance scores of the classifier models reveal the possibility of using e-nose technique in detecting CTV infected plants. 
Keywords: Volatile Organic Compounds (VOC), Support Vector Classifier (SVC), Citrus Tristeza Virus (CTV).
Scope of the Article: Artificial intelligent methods, models, techniques