Submitted to: National Cotton Council Beltwide Cotton Conference
Publication Type: Proceedings
Publication Acceptance Date: 1/15/2010
Publication Date: 4/19/2010
Citation: Suh, C.P., Medrano, E.G., Lan, Y., Hall, D.L. 2010. Detecting boll rot of cotton with an electronic nose. National Cotton Council Beltwide Cotton Conference. p. 215-218. Interpretive Summary: Early and rapid detection of diseased cotton bolls is often complicated by the absence of external symptoms on infected bolls. The potential of using electronic nose (E-nose) technology to detect bolls infected with the bacteria Pantoea agglomerans, a causative agent of South Carolina Boll Rot, was investigated. The results showed that infected bolls produce an odor that is quite different than the odor emitted from healthy bolls. More importantly, it was found that a commercially-available E-nose could detect these odors and be trained to recognize and discriminate between infected and non-infected bolls. Based on these promising but preliminary findings, E-nose technology shows potential as a non-destructive screening tool for rapidly detecting diseased bolls and further evaluation of this technology is warranted.
Technical Abstract: Early and rapid detection of diseased cotton bolls is often complicated by the absence of external symptoms on infected bolls. A preliminary study was initiated in 2009 to examine the potential of using an electronic nose (E-nose) to detect volatiles emitted from bolls infected with the opportunistic bacterium Pantoea agglomerans. Bolls on greenhouse-grown plants were inoculated at 2 weeks postanthesis with a suspension of P. agglomerans (2 x 103 colony forming units (CFU) per boll or sterile water only (control group). An E-nose (Cyranose 320) was trained to recognize the “smell print” of volatiles emitted from bolls two weeks after inoculation. Canonical projection plots and principal component analysis of the smell prints obtained during the training session showed distinct separation between P. agglomerans- and non-infected bolls. Cross-validation of the sensor data indicated the E-nose was 90% accurate in discriminating between these treatment groups. However, upon testing the bolls used to train the E-nose, only 66% of the bolls were correctly identified. Despite this marginal performance, it is anticipated that a higher level of accuracy could be achieved with minor adjustments to the training procedures as well as E-nose detection and data processing settings. Based on these preliminary results, the E-nose shows promise as a non-destructive screening tool for rapidly detecting diseased bolls, and continued investigation of this technology is warranted.