ABSTRACT
La determinación de la apariencia del grano de arroz es un aspecto clave para evaluar su calidad. Generalmente, este análisis es realizado de manera visual empleando analistas expertos, sin embargo debido a la naturaleza subjetiva de su determinación, los resultados pueden ser divergentes. Con el objetivo de evaluar la concordancia entre analistas de laboratorios latinoamericanos de calidad de arroz en la determinación de la apariencia del grano de arroz pulido con ayuda de imágenes digitalizadas, se realizó un ensayo interlaboratorio con diez analistas e imágenes de 90 granos, capturadas mediante scanner de alta resolución. Los granos fueron clasificados en cuatro categorías incluyendo grano traslúcido, grano yesoso, grano panza blanca y granos dañados. La categorización fue analizada mediante la moda, frecuencia, concordancia relativa y coeficiente de concordancia Kappa. Adicionalmente, se elaboró una galería referencial de imágenes de granos típicos de cada categoría, basada en la frecuencia de modas. Los resultados revelaron un valor de Kappa de 0,49 que corresponde a una reproducibilidad moderada, atribuida a la subjetividad del análisis visual de las imágenes. Los resultados evidencian la necesidad de uniformizar criterios de evaluación entre analistas para mejorar la confiabilidad en la determinación de la apariencia de granos de arroz.
Concordance among analysts from Latin-american laboratories for rice grain appearance determination using a gallery of digital images. The appearance of rice grain is a key aspect in quality determination. Mainly, this analysis is performed by expert analysts through visual observation; however, due to the subjective nature of the analysis, the results may vary among analysts. In order to evaluate the concordance between analysts from Latin-American rice quality laboratories for rice grain appearance through digital images, an inter-laboratory test was performed with ten analysts and images of 90 grains captured with a high resolution scanner. Rice grains were classified in four categories including translucent, chalky, white belly, and damaged grain. Data was categorized using statistic parameters like mode and its frequency, the relative concordance, and the reproducibility parameter kappa. Additionally, a referential image gallery of typical grain for each category was constructed based on mode frequency. Results showed a Kappa value of 0.49, corresponding to a moderate reproducibility, attributable to subjectivity in the visual analysis of grain images. These results reveal the need for standardize the evaluation criteria among analysts to improve the confidence of the determination of rice grain appearance.
Subject(s)
Humans , Image Processing, Computer-Assisted/statistics & numerical data , Laboratories/standards , Oryza/anatomy & histology , Latin America , Observer Variation , Quality Control , Species Specificity , Statistics as Topic/methodsABSTRACT
The genetic relationships between amylose content (AC) and appearance quality traits of indica rice (Oryza sativa L.) were investigated using conditional analysis and unconditional analysis in present experiment. The results from the unconditional analysis indicated that AC of rice positively correlated with brown rice (BR, i.e., dehulled but unmilled rice) length (BRL), width (BRW) and thickness (BRT), but was negatively correlated with the ratio of length to width (RLW). The conditional analysis showed that weight of brown rice (WBR) negatively affected the genetic relationships between AC and the appearance quality traits of rice except between AC and BRW, while the genetic relationships between AC and most appearance quality traits were negatively affected by protein content (PC). However, these influences were not apparent due to the impact of WBR or PC on the most covariance components of the different genetic systems between AC and the appearance quality traits. The conditional analysis showed that it was possible to improve AC while significantly reduce BRL and BRT under maintaining WBR. Furthermore, AC could be improved when BRL was reduced under maintaining PC, but BRW and BRT could be significantly increased.