RESUMO
Recent advances in phenotyping techniques have substantially improved the ability to mitigate type-II errors typically associated with high variance in phenotyping data sets. In particular, the implementation of automated techniques such as the High-Volume Instrument (HVI) and the Advanced Fiber Information System (AFIS) have significantly enhanced the reproducibility and standardization of various fiber quality measurements in cotton. However, micronaire is not a direct measure of either maturity or fineness, lending to limitations. AFIS only provides a calculated form of fiber diameter, not a direct measure, justifying the need for a visual-based reference method. Obtaining direct measurements of individual fibers through cross-sectional analysis and electron microscopy is a widely accepted standard but is time-consuming and requires the use of hazardous chemicals and specialized equipment. In this study, we present a simplified fiber histology and image acquisition technique that is both rapid and reproducible. We also introduce an automated image analysis program that utilizes machine learning to differentiate good fibers from bad and to subsequently collect critical phenotypic measurements. These methods have the potential to improve the efficiency of cotton fiber phenotyping, allowing for greater precision in unravelling the genetic architecture of critical traits such as fiber diameter, shape, areas of the secondary cell wall/lumen, and others, ultimately leading to larger genetic gains in fiber quality and improvements in cotton.