Cotton is a preeminent natural fiber crop that is extensively cultivated in nearly 80 countries worldwide for its natural fiber [1]. In addition to its lint, the cottonseed is also a valuable component of the plant and serves as a vital source of oil and protein [2]. The lint, which is the cotton fiber, is primarily used in the textile industry for clothing. Linters, known as short fuzz on the seed, provide cellulose for making explosives, plastics, and many other products [3]. The cottonseed, on the other hand, offers oil, meal, and hulls [3]. The global annual consumption of cotton fiber is approximately 27 million metric tons, or approximately 115 million bales [4]. Cotton breeders strive to improve fiber quality and yield to provide a better alternative to synthetic fibers derived from petroleum, which has significant environmental impacts [5]. Researchers are also exploring methods to modify cottonseed for human consumption and livestock feed [6].

In the United States, cotton production primarily comes from Gossypium hirsutum (Upland cotton) and Gossypium barbadense (Pima cotton), which account for 16.1 million bales and 400 thousand bales, respectively [7]. The widespread cultivation of Upland cotton can be attributed to its better adaptability to harsh environmental conditions, high production, and greater yield potential [8]. Conversely, Pima cotton is known for its superior fiber quality attributes such as length, strength, and fineness. Pima cotton is distinct from Upland cotton in various traits, including growing habits, adaptations, and yield [9]. Attempts to introgress genes for elite Pima fiber traits into high-yielding commercial Upland lines have been limited due to the incompatibility of their genomes [10].

Cotton fibers arise as single-celled structures emerging from epidermal cells of ovules [11]. The differentiation and development process encompasses four overlapping stages: initiation, elongation involving primary cell wall biogenesis, Secondary Cell Wall (SCW) biosynthesis, and fiber maturation [12,13]. Elongation involves the accumulation of crystalline cellulose within the secondary cell wall, forming the lumen wall that encloses an internal space called the lumen. The lumen, an essential conduit for nutrient transfer during growth [14], is characterized by a hollow canal. Initially, during the growth phase, the fiber exhibits an approximately circular cross-section. However, as the fiber undergoes desiccation, its shape becomes irregular, assuming a kidney-shaped appearance that is contingent upon the thickness of the secondary cell wall [15]. Specific genetic mechanisms regulate key fiber attributes such as length, strength, uniformity, and micronaire values [16]. Micronaire measures the air permeability of a consistent mass of compressed cotton fibers within a fixed volume [17]. Among these attributes, fiber strength and length hold paramount importance for yarn quality [18], whereas micronaire value impacts dyeing consistency and processing (spinning). It is worth noting that the micronaire value serves as an indicator of both fiber maturity and fineness [19,20,21]. However, it should be emphasized that micronaire readings represent a relative scale and cannot be solely relied upon to assess either fineness or maturity. Higher micronaire values are associated with more mature and coarser fibers, whereas lower micronaire values indicate finer and less mature fibers [20,21].

Assessing cotton fiber quality is a complex process that requires precise and accurate measurement of various fiber quality traits [22]. For over two decades, the High-Volume Instrument (HVI) has been the primary source of measurement for fiber quality and selection improvement in the cotton industry. However, the HVI system has limitations in evaluating the contribution of fiber length variations within a sample [23]. The Advanced Fiber Information System (AFIS) has emerged as a complement to the HVI for fiber quality measurement, especially fiber length distribution within a sample, providing more accurate single-fiber quality information [24]. Several studies have demonstrated that the AFIS is an effective tool for evaluating yarn quality and spinning performance, surpassing the HVI in terms of measuring mean fiber values and distributions and precisely estimating fineness and maturity through cross-sectional analysis [23]. Nevertheless, further research is needed to fully understand the potential for the AFIS to provide more helpful information regarding fiber length distribution and how it can be utilized in the cotton industry.

Maturity and fineness are two critical parameters used to evaluate cotton fiber quality. Highly mature fibers produce good-quality yarns and have good dye affinity. However, despite the significance of these fiber characteristics, no measurement methods are satisfactory, quick, and reliable [25]. The lack of reference standards for maturity has made it impossible to validate the existing instruments, such as double compression airflow instruments and the maturity module in the AFIS [24]. There are many indirect methods for measuring cotton fiber maturity; however, direct determinations can be performed only by microscopic processes, which is the definitional reference for measuring maturity and fineness [26]. In 1956, Lord developed 100 cottons to validate micronaire [25]. Gravimetric fineness was determined by the “cut and weigh” method. However, this method introduces bias into the measurements, as the sample taken is not independent of the fiber length [24]. In the 1980s, the International Textile Manufacturers Federation (ITMF), in collaboration with other organizations, established a set of nine calibration cottons [26]. In 1999, Thibodeaux and Rajasekaran [27] used a commercial image analysis system to measure the cross-sectional images of ~50 cotton varieties showing a wide range of genetic fineness. Despite these significant efforts, none of these methods have achieved widespread adoption today. The Commonwealth Science and Industrial Research Organization (CSIRO) developed the Cottonscan instrument [28,29,30,31,32] for the direct measurement of fiber fineness and the Siromat instrument for measuring fiber maturity distribution using polarized light microscopy [33,34,35,36]. In 2018, ITMF recognized the commercialization of Cottonscan/Siromat. This instrument measures the diameter of thousands of cotton fiber snippets in just 30 s and estimates the fiber maturity ratio using two procedures. Although it produces results within a few seconds without the need for resin embedding, it is worth noting that the sodium hydroxide solution used in procedure 1 is caustic and corrosive. This method does not provide direct measurements, unlike the approach presented in this paper [28,29].

Cross-sectional analysis offers direct and precise measurements of fiber maturity and fineness, serving as reference data to calibrate or validate other indirect measurements of these fiber characteristics. However, in spite of the significance, cross-sectional measurements have not been widely applied to cotton fiber quality evaluations due to tedious procedures for both sample preparation and image analysis [30]. The technique of directly measuring fiber cross-sections under an optical microscope, known as first-principles analysis, poses significant technical challenges that demand a skilled operator. Additionally, this method is labor-intensive and time-consuming. As a result, until recently, there was a lack of a universally accepted reference cotton data set with assigned values for fiber maturity and fineness [31]. Hequet et al. addressed this challenge with an extensive and detailed study aimed to create a set of 104 reference cottons for fiber maturity measurements [24]. In their study, they used a modified version of the methacrylate embedding method for fiber embedding and cross-sections developed by Boylston at the United States Department of Agriculture Southern Regional Research Center (USDA-SRRC) in New Orleans, Louisiana, USA [32,33] and the Fiber Image Analysis Software (FIAS) image analysis system developed by Xu and colleagues [30,34,35]. Although this method is more accurate, it is tedious, requires harmful chemicals, and cannot be used to rapidly assess many samples.

This study presents a novel approach for simplified and efficient fiber histology, cross-sectioning, and image analysis of individual fibers. The method involves detailed analysis of latitudinal fiber cross-sections acquired through a combination of super-resolution light microscopy and machine learning techniques. This approach establishes a first-generation analytical tool capable of automatic detection of viable fibers in an image, enabling accurate determination of key measurements while reducing experimental error. The obtained measurements provide precise quantitative data on fiber properties, specifically (1) fiber circumference and area, (2) lumen circumference and area, and others.

The primary objective of this study was to establish a standardized protocol for obtaining consistent and high-quality measurements of individual cotton fibers, with validation conducted using a subset of the reference cotton material generated by Hequet et al. [24]. To accomplish this, we devised an embedding technique where mature cotton fibers were embedded within a resin medium that preserves the natural sample structure. This approach facilitated efficient embedding of a large number of fibers within a single section while enabling individual fiber analysis. Given the inherent variability in cotton fiber sizes, even within the same boll, a substantial number of fibers were analyzed to ensure precise estimation of fiber properties. Key morphological characteristics of interest such as the diameter and area of the crystalline structure, as well as the lumen, were measured. To evaluate critical fiber traits, ten distinct cotton lines were examined, and subsequent statistical analysis was used to compare fiber properties within these lines. Notably, the experiment was conducted under standard laboratory conditions, with temperatures maintained within the range of 20 °C to 25 °C, and humidity levels between 30% and 50%.

This method was designed to detect quantitative differences between different lines of cotton fiber. High-resolution images of individual cotton fibers were acquired and had the lumen area, lumen perimeter, total area, fiber area, fiber perimeter, and circularity measured. The total area had the lumen area subtracted to obtain the true fiber area. These measurements were then used to calculate θ, which is the degree of cell wall thickening, which is calculated by (θ = 4π(Fiber True Area)/(Fiber Perimeter²)) [33]. Fiber fineness was determined by calculating the product of the Lumen Area and the density of cotton fibers (1.52 g/cm³) [33]. Standard fineness was the final measurement, which is calculated by multiplying fiber fineness by 0.577 then dividing that by θ [33]. All individual measurements were compiled based on their fiber line, creating an average measurement meant to be representative of the line as a whole. Statistical analysis was conducted to ensure that statistically significant differences in measurements were detectable. Each line had around 100 fibers imaged and analyzed to produce the average measurements. An independently analyzed data set from Cotton Incorporated using the same cotton lines was obtained. The data set was used to see if results aligned with a preestablished baseline. The data set only analyzed eight fibers per line, so the data will not fully align due to the higher variability. After the initial analysis was conducted, a histogram was generated to identify outlying data, which removed significantly outlying data from the data set. As fibers were manually selected for measurement analysis, human error could mistakenly include damaged or improperly embedded fibers. This lowered the variability of the measurements without much change to the average values. Algorithmic binning of fibers would remove the need for this process, as a standard will be consistently met.

Table 2 shows three Tukey–Kramer HSD tests for three different properties of cotton fibers. There is no overall grouping between any of the lines. Some lines are more similar to each other than others, but this is just due to similarity between lines, which is inevitable when using multiple samples. The various connecting letter reports show that there is no consistent grouping between lines across measurements. There are somewhat correlated groupings such as with lumen area and lumen perimeter, but these measurements are naturally related to each other, so these measurement groupings are expected.

Measurements each had an all-pairs Tukey–Kramer HSD test conducted to detect statistically significant divergences between the various cotton lines. All measurements had at least three significant groupings among the five lines. The inherent characteristics of cotton fibers resulted in significant variation in measurements among all the lines studied. Surprisingly, the measurements from one line exhibited greater similarity to measurements from other lines than to its own line. Consequently, the data displayed high standard deviations, indicating the substantial variability. Fortunately, the extensive range of measurements conducted demonstrated that no two lines could be grouped together consistently across all the measurements taken based on a Tukey–Kramer HSD test (Table 2). An automated approach for fiber identification and measurement was evaluated. The model had an 84% classification accuracy on the validation data set. Large quantities of analyzed fibers and objective fiber binning would reduce the overlap between fiber line phenotypes.

At least 60 fibers were measured from at least 2 cross-sections for each reference line. Table 3a highlights the differences between the embedding methods for London Resin (LR) embedding and Polymethacrylate (PMMA) embedding. The comparison results ensured that any future LR White results would not have a strong bias in data collection and did not introduce contamination for the measurements. PMMA used a method that required a long curing time that used ultraviolet light in order to polymerize. LR White allows for a quicker polymerization time, and avoiding the use of ultraviolet light allows this technique to be used on more sensitive material. Table 3b shows five unique cotton lines that display different measurements that separate out independently from each other. Each replication verified the embedding technique as not introducing bias into any measurements. For example, the lines Beta and Delta are considerably different, with Beta having a 23% larger lumen area than Delta (Beta:14.83 vs. Delta:12.02) but being within 1% of the fiber fineness (Beta: 241.84 vs. Delta: 244.31). Such different measurements can be found throughout Table 3b, further showing how little bias is introduced via LR White embedding.

Table 3a,b shows that there is a pattern when comparing measurements of the same cotton line. All of the differences in measurements from the LR White lines to the PMMA lines are scaled at a similar rate, as opposed to the different cotton lines seen in Table 3b, which have differently scaled measurement differences for the various lines.

The objective of this study was to simplify the embedding technique for improved analysis of individual cotton fibers. Previous methods utilized PMMA, which is a more intensive method. The methods involving PMMA require the use of UV light and more reactive chemicals, such as benzoyl peroxide. Additionally, these methods require a longer time for polymerization, as PMMA takes around 150 min, while LR polymerization takes less than 20 min. Drying of cross-sections from PMMA can take anywhere from overnight to several days, depending on the embedding thickness. In contrast, the currently proposed method allows the cross-sections to be imaged within 5 min since we are using charged slides, which helps them stick quickly and dry rapidly. Initially, paraffin was employed as the embedding material, and scanning electron microscopy (SEM) was utilized to capture images, followed by ImageJ for image analysis. This method, however, was found to be time-consuming and produced highly variable results with low fiber density per image (Figure 8). Paraffin is a commonly used and cost-effective agent for embedding mammalian tissue due to its simplicity, affordability, and rapid application. However, paraffin embedding can obscure the materials of interest under the microscope and necessitates a deparaffinization process to selectively retain the target material. Furthermore, paraffin embedding lacks the ability to control the orientation of fibers during the sectioning process, often resulting in sections containing horizontally or obliquely arranged fibers (Figure 9A,B).

In this study, we sought to overcome the limitations of traditional embedding methods by exploring the potential of London Resin White (LR White), an acrylic resin that offers numerous polymerization options and exhibits stability at room temperature. LR White resin has been demonstrated to preserve the structural integrity and chemical composition of various biological specimens, making it a promising embedding medium for cotton fibers. The use of N,N-dimethyl-p-toluidine as a chemical accelerator expedited the polymerization process of LR resin, resulting in the preservation of fiber structure within 5–10 min. The resulting embedding medium exhibited a smooth surface after cutting, allowing for high-quality imaging using light microscopy. Both the resin and accelerator used in this study were deemed safe for human use.

LR resin offers superior fiber alignment, resulting in a higher proportion of fibers being optimally oriented (Figure 4A,C). By adopting LR White resin for embedding cotton fibers, we were able to achieve precise control over fiber orientation, improving the accuracy and consistency of fiber measurements. LR White epoxy resin provides a rapid and efficient method for polymerizing biological materials, with several approaches available depending on the material being embedded. Initially, heat-curing was employed but was ultimately deemed unsuitable due to concerns regarding its impact on fiber quality and time efficiency. Instead, the cold-cure method was utilized, incorporating an accelerator to expedite the polymerization time. However, an exothermic reaction can occur during polymerization, necessitating careful experimentation to determine the optimal accelerator concentration that achieves the fastest polymerization rate with minimal heat generation. The use of a thin tube for shaping fibers resulted in significant heat release due to the high surface area-to-volume ratio. Nonetheless, the initial polymerization step ensured the preservation of the fiber shape within the resin, preventing subsequent interference. These results highlight the advantages of using LR White resin for embedding fibrous materials and provide insights for further optimizing polymerization conditions.

Mature fibers are obtained by letting cotton fibers develop from flower to naturally opened cotton boll without outside assistance. Cotton fibers are then allowed to dry without the addition of excessive heat, as this will affect fiber qualities and thus cause a skew in the measurements. The embedding process should result in a pill-sized sample of hardened resin. Be sure to allow the resin to fully polymerize. If not fully polymerized, the sectioning will be problematic and will likely result in poor fiber images in the end (Figure 4B). To verify complete polymerization, the hardness of the resin can be assessed during polymerization by lightly squeezing the gel capsule. Over-polymerization can occur as well, where the resin is very brittle and hard to cut, and this may also result in complications during sectioning. Both issues are related to the concentration of accelerator. Accelerator concentration is the most essential component of embedding, as accelerator-based polymerization is an exothermic reaction. Too much accelerator will release a large amount of heat [40,41] as well as squeeze any of the cotton fiber, both of which will affect fiber properties. Figure 4B represents misformed fibers resulting in excess use of accelerator. Under-polymerization will result in improper sections that will either be impossible to extract or image. Properly polymerized resin should be able to be cut by a straight razor with light force, while not leaving any resin residues after being touched, which is a sign of the resin not being fully polymerized.

After embedding, the fibers were sliced into five-micrometer sections using a glass microtome transferred onto an adhesive microscope slide, and embedding success was determined based on the images obtained from the Olympus LEXT Optical Profiler light microscope. Typical embedding success resulted in 40–60 fibers per section. The glass blades provide ultra-thin sections because they are handmade and sharper than steel blades. Steel blades are unsuitable for achieving a continuously sharp edge due to the crystalline structure of metals [42]. To prevent any disruption to the crystalline structure of the fibers during the embedding process, mature cotton fibers were collected and air-dried at ambient temperature. The current study aims to examine various fiber physical parameters, which is achieved at a 100× resolution. However, the probability of differentiating cell wall layers is minimal at 100× resolution, and this is beyond the scope of this study.

Scanning Electron Microscopy (SEM) was originally used to accurately isolate and measure cotton fibers; however, this approach was time-consuming and costly. In place of SEM imaging, light-based microscopic imaging was utilized. Subsequently, the fibers were manually measured using Adobe Photoshop, which allowed for control over multiple measurements within an image. Despite the method introducing human error, as damaged or obscured fibers had to be excluded, it still produced results similar to an independently analyzed average of the cotton line. Imaging was time-consuming and expensive for SEM, so only healthy fibers were prioritized for imaging, potentially introducing bias into the data set. In contrast, light microscopy can capture larger amounts of fiber simultaneously, allowing a wider variety of fibers to be analyzed. The application of a machine learning approach for fiber phenotyping was explored with positive results. Further adaptation of these models will ultimately allow large numbers of fibers and samples to be rapidly phenotyped and facilitate the adoption of this technique for routine fiber screening. These findings have important implications for the analysis of fibrous materials and provide insights for future studies aiming to optimize cotton fiber phenotyping.