Simultaneous morphological analysis of large numbers of hair cross sections as a tool for investigation of population‐level trends

Abstract

Objective

Characterizing the fibre properties of individuals with different hair textures across several ethnicities is important for understanding how hair shape varies within and between groups, and how these may influence consumer needs. Here, we present a high‐throughput scanning electron microscope (SEM) method for simultaneous measurement of cross‐sectional single hair shape parameters from hundreds of hairs per sample, which has not been feasible previously. We demonstrate the power of the method through application on a population with diverse hair types.

Methods

Scalp hairs were collected from individuals located in the United States of America. Each hair sample (consisting of up to several hundred fibres) was classified using two different methods, one during clinical collection [hair texture Types 1–4] and later another blind standard laboratory method [hair curliness classification Types I–VIII]. Additional clinical data were collected on age and self‐identified ethnicity. Hair shape parameters (cross‐sectional area, ellipticity, shape factors) were measured using a SEM sample preparation, imaging and image analysis method. SEM data were analysed with respect to clinical texture, age and self‐identified ethnicity and subsequent hair curliness classifications.

Results

The SEM method generated sufficient data from each sample to identify trends, and we found some statistically significant differences between SEM hair shape parameters and clinical sample types, as well as with laboratory curliness classifications. In the curliness classification, there was an expected tendency between hair curliness and aspect ratio: curlier hairs were more elliptical than straight hairs. In terms of the hair grouping types, in the age group, older individuals had thinner hairs than young ones. In the texture group, individuals in Texture Type 1 had thinner hairs than Texture Types 2, 3 and 4. Texture Types 3 and 4 had hairs with a more elliptical profile than individuals in Texture Types 1 and 2.

Conclusion

The SEM method was reliable to quantify cross‐sectional hair parameters within populations of donors with different types of hair. This approach corroborates clinically assessed hair type and curliness classification systems and provides a more thorough characterization of hair shape variation between and within individuals and ethnicities.

Here we present a method for simultaneous measurement of cross‐sectional single hair shape parameters from hundreds of hairs per sample. This method is reliable for quantifying hair parameters within populations of donors with different types of hair, and it generates sufficient data from each sample for statistical analysis.

INTRODUCTION

Hair shape, as subjectively experienced on the head, is influenced by a wide range of parameters. There is great interest in understanding how single hair fibre dimensional properties influence on‐the‐head hair texture, how these may be linked to ethnicity or age and how they might drive consumer needs. However, it is difficult to obtain objective measurements of single hair shape at a statistical level suitable to explain the mechanisms that underpin variation in on‐the‐head hair texture in a population [1]. Macroscopic on‐the‐head hair phenotype is typically assessed during clinical trials by expert clinicians or by study participants (i.e. self‐perception) to rapidly classify participants into hair texture types (e.g. straight, high‐curl, wavy, coiled) [2]. However, innate variation needs to be understood at the level of single hairs shape, structure and molecular composition because on‐the‐head hair texture is an emergent effect of many single hairs interacting with each other [3, 4, 5]. There is a lack of efficient methods for assessing and characterizing single hair properties, such as morphology or shape, in a population. Also, effects caused by technology (e.g. hair products) are a consequence of chemical or physical processes within single hairs [6, 7]. This leads to a problem of scale in which it is necessary to measure hair samples from many individuals to understand inter‐individual variation and necessary to measure from many single hairs per participant to understand intra‐individual variation. In short, there may be a need to measure thousands or tens of thousands of single hairs per study.

Single hair three‐dimensional shape can be broken down into components such as curvature, torsion and diameter which varies along the length of the hair [3, 5, 8]. Precise three‐dimensional measurement of single hair morphology is a slow process [2, 5, 9]. Rapid on‐the‐head hair qualitative classification used by hairdressers, for example, the Andre Walker types [2], is insufficient to describe single‐hair diversity. Multifactor systems that integrate quantitative approximations from single hairs of curvature and torsion, such as the “L'Oréal” eight‐type system [10, 11] provide a more precise connection between single hairs and macroscopic phenotype, but the process of classification is manual and relatively slow.

The wool industry developed automated high‐throughput methods for collecting population‐level data on wool curvature and diameter during the 1980s that used automated microscopy and image analysis to measure from short snippets of wool fibres [12]. The resultant instrument, the Optical Fibre Diameter Analyser (OFDA) is the standard method within the wool industry and has been effectively used on human scalp hair [13]. A recent laboratory bench high‐throughput quantitative approach expanded this method by including the analysis of hair cross sections [1]. Cross‐section analysis has largely been missing from high‐throughput methods with the exception of fibres mounted for automated testing in, for example, Dia‐Stron mechanical testing instruments [14].

Single hairs cross‐sectional shape varies considerably and has long been of interest because the derived measurement of normalized hair cross‐sectional ellipticity varies along single hairs [15] and correlates with hair curvature and varies with genetic background [16, 17, 18]. Although ellipticity is unlikely to be the primary cause of hair curvature [1, 5], the effects of ellipticity on modulating curl‐related performance or having a subtle role in modulating on‐the‐head hair shape are not fully resolved [19]. Studies on single hair cross‐sectional shape, like those of hair curvature, require a large sample size and therefore high‐throughput methods.

Here we present a high‐throughput method for sample preparation, imaging and image analysis that allows for the simultaneous measurement of cross‐sectional single hair parameters (cross‐sectional area, ellipticity, shape factors) from hundreds of hairs per sample. We demonstrate the power of this method by applying it to a diverse population of 191 individuals with various hair types across multiple ethnicities.

METHODS

Hair sample collection

One hundred and ninety‐one hair samples from different individuals were collected in the United States of America following ethical approval (Advarra Institutional Review Board, Columbia, MD, USA). Hair samples were collected from a 1.5 × 1.5 cm area on the scalp of each participant. The sampling area was defined to be approximately the same location on the scalp for each participant. Hair samples were clipped as close as possible to the scalp by trained clinical study technicians, and they were mounted on a polystyrene holder to facilitate the identification of the root and tip ends of the hair.

Individuals were categorized according to self‐identified ethnicity (White/Caucasian, Hispanic/Latino and Black/African American) and clinically assessed on‐the‐head hair texture by trained assessors (Texture Type 1, 2, 3 or 4, based on the Andre Walker hair typing system; similar to Gaines et al. [2]). White/Caucasian volunteers were included if they had Texture Type 1 hair, Hispanic/Latino volunteers if they had Texture Type 2 hair, and Black/African American volunteers if they had Texture Type 3 or Type 4 hair. The collection plan aimed to independently balance samples across two criteria—texture and age (Figure 1). Participant age was classified as two groups—Young (18–36 years) and Older (37–55 years).

FIGURE 1.

Proportions of individuals (%) analysed in age, participant‐defined ethnicity and clinician‐defined texture grouping types. Labels are category names used during measurement and analysis.

In keeping with good experimental practice, all measurements, testing and analyses were performed blind. All individual hair samples were additionally assigned random labels ahead of single hair measurements. Image analysis was automated by computer scripting. Group codes and sample codes were maintained during population‐level statistical analysis, with the definition of categories revealed only after the data were analysed.

Curliness classification

Hair types were defined according to specific shape criteria following the standard method described in Loussouarn et al. [10], and De La Mettrie et al. [11]. In this method, three descriptors are measured to classify the hairs into eight curliness types (Types I to VIII, with very straight hairs described as Type I and the curliest hairs as Type VIII). Following this method, three single hairs per individual (triplicate samples from 191 individuals) were measured and categorized using the three descriptors: (1) curve diameter (CD); (2) curl index (i) and (3) number of waves (w). The first four types of curliness (I–IV) are distinguished by CD values, and the combination of i and w differentiates the other four types (V–VIII), as summarized in the decision tree in Figure S1 and Table S1 (Supplementary S1).

Sample preparation consisted of single hairs clipped near the root end and attached to glass slides using double side tape. Hairs were fully extended until they were straight, after which they were cut to a length of 60 mm from the root end (L ₂ value). Single hairs were immersed in 1% v/v neutral shampoo solution (12% sodium lauryl ether (1EO) sulfate and 1.6% cocamidopropyl betaine) for 3 min, rinsed in tap water for 3 min and then dried for a minimum period of 18 h on a paper towel. The hair was carefully separated from the glass slide, laid on another glass slide free of mechanical stress (to allow a return to natural shape) and a second glass slide was gently lowered onto the hair. CD values (Figure S2), i and w (Figure S3) parameters were classified accordingly Table S1 [10, 11].

High‐throughput cross‐sectional measurements

Using a small industrial knitting hook (e.g. latch crochet hook), between (approx.) 80 and 200 single hair fibres per sample from near the root end of collected samples were pulled into generic brand heat‐shrink plastic tubes (initial internal diameter 1.5 mm, shrinkage 50%, polyolefin) (Figure 2a). Available hairs per sample may vary in clinical studies, and while it is possible to vary hair, heat‐shrink tube diameter and hook sizes, we found for our tubes that a maximum of about 200 hair fibres sufficiently stabilized samples while still being big enough to handle easily. Because the sample is pulled through in a loop, each hair is ultimately represented twice as a cross‐section (e.g. 200 hairs results in 400 cross‐sections). For this study, one plastic tube was prepared per individual. Before loading the sample inside the heat‐shrink tubes, hair fibres were combed with a fine‐tooth comb to align them parallel to the tube. Tubes containing the doubled‐up hair samples were accumulated in a grid and their coordinates recorded (Figure 2b) before being heated in plastic laboratory tubes containing distilled water at 90°C for 2 min (Figure 2c). After removal from heating, tubes were returned to their grid positions and then each cut at one end (ultimately the non‐imaging end), and a small quantity (small paintbrush touch) of generic brand clear nail varnish (primary ingredients, nitrocellulose and formaldehyde resin in solvent mainly composed of ethyl acetate) as a glue was applied across the cut end, returned to the grid and left to dry for at least 1 h. To generate an imaging surface, each tube was cut 10 mm from the glued end and then within the glued region using a clean single‐edge razor blade (Figure 2d). The tube was immobilized in a rig for this step, to ensure that the cuts were straight and parallel to one another (Figure S4 in Supplementary S2), generating two half‐tubes for each individual. After cutting, half‐tubes were placed at their grid coordinates, with the nail varnish containing end down, within a custom aluminium sample holder containing multiple samples (Figure 2e). Holders containing the half‐tubes used for imaging were also used as a sample archive. The half‐tubes without nail varnish were also archived. Two variations of the sample processing method were developed to overcome challenges presented by very high‐curl hair and for samples that were sparse (e.g. fewer than 100 hairs).

FIGURE 2.

Sample processing workflow. (a) Hair sample and insertion into a tube using an industrial knitting machine hook. (b) Grid for temporary storage of samples during processing. (c) Thermo‐mixer for heating samples in individual tubes of water. (d) Rig for fast and precise cutting of 5 mm samples for imaging. (e) Sample holder for positioning samples during imaging.

Especially curly hair samples resulted in poor data quality using the above‐described method because many individual hairs would move and tilt following cross‐sectioning and because many of these samples had higher levels of adhered material that was resistant to removal by the shampoo treatment. These hairs were cleaned in 0.15% v/v polyethylene glycol nonylphenyl ether in distilled water at 60°C and then 40°C, followed by washing in distilled water at 40°C and 60°C for 2 min in each condition to remove residual cosmetic products. Following drying, the hairs were held at one end and straightened using a sequence of passes using a comb followed by a flat iron (T > 200°C), repeated 10 times, after which they were loaded into tubes as described in the previous paragraph.

Trained clinical assessors collected hair samples from a defined sampling area containing a variable number of hairs. For some individuals, the sampling area contained fewer than 100 hairs. In these cases, another method variant for sparse hair fibre samples was applied; 30% of the hair samples in this work were collected with less than 100 hair fibres. The low packing density of sparse samples resulted in hair movement, slippage and tilting that resulted in reduced data quality. To overcome this, nylon yarn with a cross‐sectional shape different from that of human hair (Figure 3d) was combined with the sparse hairs to fill the heat‐shrink plastic tubes and processed as described previously.

FIGURE 3.

Scanning electron microscopy (SEM) image of straight hair (a) and sparse hair combined with nylon yarn (d). Examples of image analysis using FIJI software (b, c, e and f).

Cross‐sectional images were obtained under vacuum using a benchtop scanning electron microscope (SEM, TM3030Plus, Hitachi, Japan) in the backscatter electron imaging mode at an accelerating voltage of 15 kV, Figure 3a,d. Images were obtained at 80× magnification without metal coating. Image analysis of SEM micrographs was automated via a script implemented as a plugin for ImageJ2 or FIJI (Fiji Is Just ImageJ, details in Supplementary S3, Figure S5 and Table S2) to generate cross‐sectional area (CSA) values (based on pixel measurements) and aspect ratio (AR) values calculated from the ratio of the major axis to the minor axis acquired from a best‐fit ellipse (more elliptical hair fibres have values higher than one). The script was written in Python, run using ImageJ2/FIJI's Jython interpreter and calling on standard Python and ImageJ2 libraries (Figure 3b,c,e,f). A variation on the standard script for analysis of curly and straight hair samples was made for sparse samples in which analysis is essentially the same but a smaller region is analysed due to the tendency of sparse sample tubes to shrink and the algorithm to produce poor results (Supplementary S3).

All results were checked by a trained operator as part of a defined quality control (QC) process. The QC process consisted of manual scanning for errors in each result, errors were largely caused by material adhering to hairs, debris, excess glue, tilted hairs or other damage. Samples failed the QC check if there were more than one error in every 10 hairs. Where possible, samples failing a QC check were re‐imaged (other half‐tube without nail varnish) and/or the sample preparation repeated. Any that failed the QC check twice were not included in the data analysis. An average of 160 cross‐section hair fibres per individual were measured in this work.

Statistical analysis

Statistical analysis was completed using the R statistical programming language; version R Core Team (2022). The cross‐sectional area (CSA) and aspect ratio (AR) values obtained from measurements described in the cross‐sectional measurements were analysed according to their hair grouping types (curliness classification, age, ethnicity and texture). In this study, hair fibres were doubled‐up inside the shrinktube and most of them were measured twice (two observations per hair); the variation along a hair fibre had been investigated during the method development, and it was found to be similar to the variation between different hairs, so each observation was considered to be independent [Unpublished work]. Parametric statistics were used only after normality testing of CSA and AR data confirmed that data per group was normally distributed (histograms for each individual are in Supplementary S4). Means and standard deviations (SD) for individuals have been used for the analysis (SD plots are in Supplementary S5, Figures S7 and S8).

Hair grouping type was investigated by ANOVA to test for significant differences (at p < 0.05 level) of CSA and AR. Tukey adjustments [20] for the levels of significance were applied, post hoc, to allow for the number of tests performed (Supplementary S6).

RESULTS AND DISCUSSION

Curliness classification

In this group of individuals (Figure 4), the population distribution was more represented by Curliness Types II to VI, not very straight and not very curly. Most of the individuals were Curliness Types III and VI. Samples classified as Curliness Types I, VII and VIII were rare or absent in the 191 individuals analysed in this study.

FIGURE 4.

Distribution of curliness type classifications of 191 individuals, population from the United States of America. Curliness Types I, VII and VIII were rare or absent.

The following sections present the results and discuss how the sampling classifications (age, texture and ethnicity) relate to the blind‐analysed curliness classification (Curliness Types I–VIII), which is followed by results and discussion of the high‐throughput cross‐section method applied to that data set. Plots of the proportions of age (young and older) are presented in Figure 5a, texture (Types 1–4) in Figure 5b and ethnicity (White/Caucasian, Hispanic/Latino and Black/African American) in Figure 5c. Tabular data are presented in a supplement (Table S3).

FIGURE 5.

Proportions of grouping types in the curliness classification (Types I–VIII). (a) Age group: Young (18–36 years old) and Older (37–55 years old). (b) Texture group: Type 1, Type 2, Type 3 and Type 4. (c) Self‐identified ethnicity group: White/Caucasian, Hispanic/Latino and Black/African American.

Age did not correlate with curliness class

We observed close to a 1:1 split between Age Young (18–36 years old) and Older (37–55 years old) in all Curliness Types (Figure 5a).

Expert subjective classification of hair texture trends towards curliness class

Texture (Figure 5b) Types 1 and 2 were primarily found in straighter hair types (Curliness Types I to III), while Texture Types 3 and 4 were found primarily in curlier hairs (Curliness Types V to VII). Curliness Type IV had the four texture types (Texture Types 1 to 4), but most of the individuals were Texture Types 2 and 3.

In the participant‐identified ethnicity category (Figure 5c), White/Caucasian was primarily associated with straighter hairs (Curliness Types I and II), while Black/African American was primarily associated with curlier hairs (Curliness Types IV–VII). Hispanic/Latino was mostly associated with wavy hairs (Curliness Types II–IV).

At‐collection classification has value and is an adequate indicator of curliness classification

Clinical classification is an established method to differentiate hair into groups quickly and inexpensively at the point of collection. This method provides valuable information about how hair samples are distributed across different categories during sampling, indicating whether the collection procedure is effectively targeting the intended individuals for a specific study. In this method, on‐the‐head hairs are qualitatively separated into groups by way of observations of experts. We conclude from this study that the exercise was a useful first step in the sorting of hair into different categories, such as texture, ethnicity and age. While parameters that are self‐classified, such as ethnicity, can be biased by social circumstances, the classifications for hair texture generally agreed with the subsequent blind‐tested curliness classifications (which included replicates). At the point of collection, on‐the‐head hair is classified into four types of texture. However, such broad classification does not fully capture the complexity of human hair's biological diversity. In the curliness classification, single hairs are classified based on measurements of physical features, such as curve diameter and number of waves, resulting in eight distinct categories. For example, hair classified as Texture Type 1 in the clinical classification (Figure 5b) was distributed across four different Curliness Types (I, II, III and IV) after measurements using the curliness classification method.

Large‐scale hair fibre cross‐sectional measurements, scanning electron microscopy (SEM) method

The scanning electron microscope (SEM) approach allowed us to measure single hair fibre cross‐sectional area (CSA) and aspect ratio (AR) for large numbers of hair fibres from small tresses of each donor. Accurate and precise measurement of hair fibre cross‐section dimensions, when hairs are handled individually, is typically time‐consuming. These types of measurements for single hairs are generally obtained using light microscopy imaging or laser diffraction pattern techniques. The SEM approach enables measurement of variability between hair fibres from single individuals while being rapid enough to be used for population studies. Here we correlated the SEM method results with curliness classification and clinical team classifications.

Cross‐sectional measurements from 170 individuals were obtained using the SEM method, while 21 individuals could not be measured even with our quality control system where the imaging and/or sample preparation was repeated. The number of individuals in age, texture and ethnicity grouping is well distributed, with representation within all type categories (Table S4 in Supplementary S7). The 21 individuals that failed the quality control are mainly from Age Older samples (older 37–55 years old), Texture Type 4, Ethnicity Black/African American and Type V and Type VI in the curliness classification. Curlier hair samples had more damage compared to other hair types, and these groups had higher levels of adhered material, which contributed to poorer data quality.

We found that hair cross‐sectional dimensions varied considerably within individuals. Figure 6 shows randomly selected graphs of CSA and AR distributions, means and standard deviations from five individuals from our larger data set (full set in Supplementary S4); more than 100 hairs per individual were measured. Data in the graphs were spread out, showing the high variability between hairs. The results suggest to us that measuring large numbers of single hairs per individual is crucial for characterizing hair shape in population studies.

FIGURE 6.

Variability between single hair fibres from the same individual. Distribution in cross‐sectional area (a) and aspect ratio (b) data in different individuals. Example of 5 individuals (more than 100 hairs per individual were measured), box indicates average variable from 170 (all histograms in Supplementary S4).

Cross‐sectional measurements and curliness classification

When individuals were classified into curliness types, no significant differences were found between CSA (Figure 7a). However, statistically significant differences between types in their AR (Figure 8a) were detected. Types V and VI (curlier hairs) had significantly higher AR than the others. Types II and III (straight/wavy) had lower AR than the others. Types I, IV and VII had AR in between that of Types V and VI and Types II and III, Type I closer to Type II and III and Type IV and VII closer to Type V and VI.

FIGURE 7.

Violin plots for mean cross‐sectional area. (a) Curliness classification (Types I–VII). (b) Age group: Young (18–36 years old) and Older (37–55 years old). (c) Texture group: Type 1, Type 2, Type 3 and Type 4. (d) Self‐identified ethnicity group: White/Caucasian, Hispanic/Latino and Black/African American. (e) Mean CSA values for curliness classification, age, texture and ethnicity. *p < 0.05, **p < 0.01 and ***p < 0.001. Plots for standard deviations are in Supplementary S5.

FIGURE 8.

Violin plots for mean aspect ratio. (a) Curliness classification (Types I–VII). (b) Age group: Young (18–36 years old) and Older (37–55 years old). (c) Texture group: Type 1, Type 2, Type 3 and Type 4. (d) Self‐identified ethnicity group: White/Caucasian, Hispanic/Latino and Black/African American. (e) Mean AR values for curliness classification, age, texture and ethnicity. *p < 0.05, **p < 0.01 and ***p < 0.001. Plots for standard deviations are in Supplementary S5.

It is important to view these analyses in light of the underrepresentation of Type I (7 individuals) and VII (1 individual). Nevertheless, in summary, we observed a trend between hair curliness and aspect ratio, with curlier hairs being more elliptical than straight hairs (Figure 9).

FIGURE 9.

Summary of the results: Significant differences between cross‐sectional measurements in the curliness classification and clinical classification (p‐values < 0.05). Statistical analysis is in Supplementary S5.

Our results, and those from other studies on human hair, have found a positive correlation between curliness and cross‐sectional aspect ratio for scalp hair from multiple genetic backgrounds [1, 21, 22]. However, correlation does not imply a causative effect [1, 4, 5, 19]. Animal models with variable hair fibre curliness and a similar internal organization to human scalp hair (e.g. sheep wool) present high‐curl hair fibres with a close to circular cross section. There are also reverse examples where hairs from merino and merino relatives that have a more elliptical profile are less curly than those with circular profiles [23].

Cross‐sectional measurements and clinical classification (hair grouping types)

We found statistically significant differences (at p < 0.05 level) between cross‐sectional measurements in the sample‐collection classification groups: age (CSA in Figure 7b and AR in Figure 8b), texture (CSA in Figure 7c and AR in Figure 8c) and ethnicity (CSA in Figure 7d and AR in Figure 8d). A summary of the results is in Figure 9.

Cross‐sectional area was found to be significantly higher in Age Young samples (18–36 years old) than in Age Older samples (older 37–55 years old) (Figure 7b). No significant differences were found in relation to AR (Figure 8b). The result that older individuals tended to have slightly narrower hair fibres than younger individuals supports earlier findings, suggesting that perceived hair thinning occurs during normal chronological aging, in part because of thinner hair fibres in addition to reduced density [13, 24].

Statistically significant differences between texture types were found with respect to CSA (Figure 7c). Texture Type 2 hair fibres had higher CSA, and Texture Type 1 hairs had lesser CSA. Texture Types 3 and 4 CSA values were in between Texture Type 1 and Type 2. Statistically significant differences with a high level of confidence were found between texture types with respect to AR (Figure 8c); Texture Types 1 and 2 had lower mean AR and Texture Types 3 and 4 had higher mean AR. In summary, individuals classed as Texture Type 1 had narrower hair fibres than Texture Types 2, 3 and 4. Individuals classed as Texture Types 3 and 4 had hair fibres that had a more elliptical profile than individuals in Texture Types 1 and 2.

Statistically significant differences between ethnicity types were found with respect to CSA (Figure 7d). Ethnicity Hispanic/Latino had a higher CSA than Ethnicity White/Caucasian and Ethnicity Black/African American; Ethnicity White/Caucasian had a lower CSA than Ethnicity Hispanic/Latino and Ethnicity Black/African American. Ethnicity Black/African American had a mean CSA value in between Ethnicity White/Caucasian and Hispanic/Latino. Differences between texture types in their AR (Figure 8d) were highly significant; Ethnicity Black/African American had higher AR than Ethnicity White/Caucasian and Hispanic/Latino. In summary, individuals of Ethnicity White/Caucasian had the narrowest hair fibres and those from Ethnicity Hispanic/Latino the widest hair fibres, while individuals of Ethnicity Black/African American had hair fibres that were markedly more elliptical in cross‐section than the others.

SEM method to characterize hair fiber cross‐sectional shape in a population

A robust and repeatable sample processing method combined with SEM imaging and automated data extraction is an approach that provided accurate, reliable and reproducible results. In comparison with light microscopy and high‐magnification macro‐photography, we found the main advantages were: (1) Irrespective of hair colour, inclusions or damage, hair fibre and cross‐sectional cut surfaces appear similar in SEM micrographs; (2) resolution (benchtop SEM) at low magnification 80× with approximately 100 hairs produces a pixel width in the captured image of 1700 nm, giving a resolution of around 3.4 μm; (3) other features (medullation, cuticle thickness and number of layers) can be identified without further sample preparation at higher magnifications; (4) the vacuum system in the SEM causes a reliable level of lateral shrinkage (via dehydration) of hair fibres, which increases the gap between adjacent fibres, making them easier to separate for image analysis.

Because hair is highly sensitive to atmospheric water, the vacuum system provides a level of standardization (equiv. relative humidity at close to 0%) that would otherwise require careful maintenance during imaging. However, the potential downside of the vacuum approach is that the standard framework for data comparison for hair studies is generally interpreted at ~50% relative humidity (RH) at ~20°C. The sorption curve for hair is not linear but tends to change more rapidly at low and high humidity levels [25]. The choice of 50% RH coincides with a central sector of relative stability and gentle linear increase in moisture uptake (thus swelling) and increasing RH. The SEM approach would not be suitable, for example, for studies investigating the effects of water swelling on cross‐sectional geometry.

Swelling behaviour of hair fibre cross‐sections according to CSA and AR was verified by measuring the profiles of the same fibres across a range of conditions and using the SEM (Supplementary S8). A range of hair fibres that had diverse cross‐sectional areas, diameters, aspect ratios and shapes were purposely chosen. Results indicated that despite the diversity of shape variation of hair fibres within the sample, swelling was normally distributed and hair fibres swelled laterally the same way despite the diverse cross‐section shapes chosen (Supplementary S8). Correction factors to convert the CSA and AR measured from SEM micrographs to the equivalent values at 50% RH at 21°C (room temperature) could be achieved by further experiments with a larger sample size.

CONCLUSION

The developed scanning electron microscope (SEM) method proved to be a reliable method in the characterization of hair samples within populations of donors with different types of hair. This approach enabled accurate and precise measurements of variability between hundreds of hair fibres from single individuals while being rapid enough to be used for population studies. The SEM technique also corroborates clinically assessed hair type and curliness classification systems, and significant differences were found in the classification into curliness types (I–VIII), age, ethnicity and texture. This method could be applied to capture participant hair cross‐section variability in future studies with participants from other ethnicities or consumer groups to understand hair shapes in different populations.

FUNDING INFORMATION

Unilever R&D.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no competing interests.

Supporting information

Data S1:

Richena M, Noble A, Parker K, Bhogal RK, Messenger D, Clerens S, et al. Simultaneous morphological analysis of large numbers of hair cross sections as a tool for investigation of population‐level trends. Int J Cosmet Sci. 2025;47:676–690. 10.1111/ics.13057