Proof-of-Concept: NIR Spectroscopy Can Detect Rice Pathogen, Burkholderia glumae , in Artificially Inoculated Rice Seeds

Strong evidence that near-infrared transmittance and reflectance spectroscopy can detect B. glumae contamination in bacterial suspensions and in individual rice seeds, respectively, was shown in this proof-of-concept study. For B. glumae suspension (0.0 to 1.42 x 10 4 CFU mL -1 ), NIR transmittance spectroscopy using partial least squares (PLS) regression calibration model (1000-1650 nm) showed a coefficient of determination (R 2 ) = 0.984 and root mean square error (RMSE) = 0.031. For rice seeds treated in varying dilutions of B. glumae (0.0 to 4.52 log 10 CFU/mL bacterial load in seeds), NIR reflectance spectroscopy using a selected PLS second derivative with Savitzky-Golay smoothing calibration model (1000-1650 nm) resulted in coefficient of determination of calibration (R 2 Cal) = 0.97, root mean square error of calibration (RMSEC) = 0.06, coefficient of determination of cross-validation (R 2 CV) = 0.93, standard error of cross validation (SECV) = 0.08 at 7 factors; the independent validation showed coefficient of determination of validation (R 2 Val) = 0.83 and standard error of prediction (SEP) = 0.11. Two-category qualitative PLS calibration model (1000-1650 nm) correctly classified 94.7% of uninoculated and 100% of inoculated seeds with independent validation of 90% and 100%, respectively. Predictions may be attributable to differences in aliphatic hydrocarbons, cellulose, amide, oil, protein, and starch contents across rice seeds that are uninoculated and inoculated at varying dilution levels. Developing NIR-based instrumentation for individual seed segregation of healthy from contaminated rice seeds can support a clean seed program and can be a useful tool for quarantine officers for seed exchange and farmers’ seed production.