Recipient Organization
UNIVERSITY OF NORTH TEXAS
1155 UNION CIR #305250
DENTON,TX 76203-5017
Performing Department
(N/A)
Non Technical Summary
Phenotypic traits of cotton fibers, including biological fineness and ribbon shape (maturity level), provide measures of fiber quality which are needed for cotton improvement efforts. The current fiber testing methods are unable to provide both rapid and accurate measurements of individual fibers. In this project, we will develop a microscope-based imaging system to produce high-fidelity images and 3D surface maps of fibers in the longitudinal view, permitting quick sample preparations and high-throughput fiber phenotyping. The system will be equipped with a newly modified pneumatic cutter that can cut and spread fiber snippets onto a slide in seconds, and a light microscope with a 3-axis motorized stage and high-resolution digital camera to automate positioning the slide and focusing the images. The software will be customized by adopting deep learning and image fusion algorithms to automatically classify ribbon shapes, scan ribbon profiles, and calculate phenotypic features in the 3D space. Two diversity panels of cotton (Gossypium species and Upland) will be grown to validate the fiber phenotypic measurements of the proposed system by comparing them with those from traditional testing systems. A successful 3D phenotyping tool will enable cotton geneticists to measure the traits of fiber fineness and ribbon shape with a speed consistent with the short timeframe available for making decisions in a cotton breeding program. The proposal is submitted in response to the commodity board topic #2 "Develop advanced, high-throughput, visual fiber phenotyping microscopy tools that ultimately lead to superior cotton fiber quality."
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Goals / Objectives
Cotton fiber phenotyping refers to the measurement of the geometric features of the fiber, e.g., biological fineness and ribbon shape. Fiber biological fineness is calculated by the perimeter of the fiber cross-section or the diameter of an equivalent circle having the same perimeter as that of the cross-section. Fiber ribbon shape is a measure of the deformation of the cylindrical structure of the fiber due to flattening and twisting. These phenotypic measurements are characteristics associated with a cotton variety or species [1] and as such are important data for cotton breeders in their efforts to find genetic linkages to improve fiber quality of U.S. cotton. this proposed project is aimed to create a high throughput fiber phenotyping tool to accurately measure cotton biological fineness and ribbon shape in the longitudinal view for cotton breeding, parent germplasm selection and fiber development research. To achieve this goal, we will develop a microscope-based imaging system and an operational protocol with the following objectives and tasks:To construct a fiber cutter and spreader driven by a pneumatic force to automatically prepare a sample slide that can carry over 10,000 well-spread fiber snippets.To equip a light microscope with a 3-axis motorized stage, high-resolution digital camera, and specialized software to automate the motion of the stage and camera functions to generate sequential fiber images in the z direction at each stage stop.To create image-processing software with machine learning, image fusion, and other advanced algorithms to detect in-focus pixels from sequential images to form high-fidelity images and z-maps (i.e., 3D) of fiber ribbons.To develop image-analysis software to classify ribbon shapes, scan ribbon profiles, and calculate phenotypic features (e.g., perimeter, twists/mm) based on ribbon shape models.To establish two fiber diversity panels of cotton (Gossypium species and Upland) to validate fiber phenotypic measurements of the proposed system, and to compare them with those from traditional testing systems (e.g., AFIS and HVI).
Project Methods
The proposed project will be conducted by creating a high throughput fiber phenotyping system and establishing two panels of diverse cotton samples to validate phenotypic measurements. The development of the new phenotyping system will be based on a light microscope equipped with a 3-axis motorized stage and high-resolution digital camera to acquire z-series image sequence of fibers in the longitudinal view (i.e., side view), and image fusion and deep learning algorithms to produce high-fidelity images and 3D surface maps of fibers. This longitudinal fiber imaging approach eliminates the need for the laborious fiber cross-sectioning procedure currently employed, allowing for swift sample preparations and facilitating high-throughput image analysis. Another distinctive feature of this approach is its capability to generate 3D fiber images in the longitudinal view, facilitating precise measurements of fiber biological perimeter, maturity, and ribbon twists, which are invaluable for cotton breeding, parent germplasm selection, and developmental research.Efforts:Our efforts will encompass the identification, tailoring, and application of computational methods to process cotton fiber images with needed efficacy and efficiency. This includes the incorporation of image fusion and classification algorithms. For image fusion, we will introduce a crucial deep learning model (U-net) designed for extracting in-focus pixels, a pivotal step in 3D image generation. In the realm of image classification, we will utilize another deep learning model, the few-shot learning prototypical network, for accurately categorizing cotton ribbon shapes, directly linked to the maturity levels of fibers. These AI-driven algorithms will greatly enhance the .Another pivotal initiative is to establish two cotton diversity panels, focusing on Gossypium species and Upland cotton--the cultivated species globally and in the U.S. Each accession within these panels will be characterized by a restricted set of molecular markers and unique fiber traits. Establishing the fundamental connections between molecular markers and phenotypic features will enhance geneticists' comprehension of the associations between genotype and phenotype, particularly in relation to fiber fineness within the diversity panels.Evaluations:The fiber phenotyping system will be evaluated by comparing the phenotypic traits, such as biological fineness and maturity, in two diverse cotton sample panels measured by the proposed system against those assessed using traditional fiber testing methods--the High-Value Instrument (HVI) and the Advanced Fiber Information System (AFIS). Correlation analyses will be conducted using analytical software such as Excel, SPSS, and others to examine the relationships between the phenotypic traits obtained from the proposed system and the AFIS fineness and maturity ratios, as well as the HVI micronaire. A strong correlation between these systems will serve as validation for the proposed system. Following validation, we will delve into the associations of phenotypic traits with SSR allele frequencies from prior research, specifically distinguishing between categories of high and low-value accessions. This analysis aims to deepen our understanding of the connections between genotype and phenotype within the diversity panels.