Challenges with textured hair analysis: single fibre and hair assembly assessments and statistical explorations

Introduction: Very curly hair, often referred to as textured hair, is defined by the fibre’s highly elliptical cross-sectional shape, and irregular twists and kinks. This results in hair assembly characteristics that are challenging to assess in laboratory conditions using traditional instrumental methods such as combing and single fibre mechanical tests. These tests are reliant on standardised fibre and tress length and the interpretation of data is based on assumptions of equally distributed forces (cross sectionally and longitudinally). Therefore, new approaches to textured hair testing are needed reflecting the magnitude of effect of hair geometry and interactions with products.

Methods: Tress testing (n=8): Wet hair combing and dry friction tests before (t0) and after multiple product and heat styling applications (t1) were performed. Total Work of Combing (TWC) and Coefficient of Friction (COF) at t0 and t1 were analysed using t-tests and PCA. One-way independent ANOVA was applied to the ΔTWC and ΔCOF with t0 values as co-variates.

Single fibre testing (n=80): Single fibres before (t0) and after treatment with silicone oil (t1) were photographed. Image features describing the fibre 2D form were extracted using computer vision processes. t-tests, correlations and PCA of the futures were conducted.

Results: The TWC and COF for tresses decreased after heat styling indicating permanent loss of curl pattern (t-test), however two active treatments mitigated this effect (ANOVA). The PCA analysis identified two components which accounted for the differences between treatments. The main component is hypothesized to be related to the curl pattern, the second component as the product effect on the hair surface. Combing test results showed higher level of dispersion than friction.

Single fibre feature analysis produced significant difference in t0 and t1 features related to the estimated number of small turns along a single fibre (decreased) and average overall fibre curvature (increased). The PCA analysis identified five components accounting for >80% of the data variability which did not change between and after treatment. The curvature of the lower two quartiles along the fibre’s length appeared to have the highest loading, meaning the effect treatment had on them.

Discussion and conclusion: Combing and friction of textured hair are most significantly impacted by fibre curvature modifications. Accounting for tress variability via covariates and exploring combinations of tests and statistical analyses mitigates the challenges with experimental control and allow for differentiation of cosmetic effect on the hair assembly. In this case, protection against permanent hair deformation was detected. The analysis of single fibre’s form captured via 3D/2D scanning/imaging to quantify the scale and dimensions of curl modification effects was complementary. The variance distribution in the PCA before and after oil treatment is small suggesting an overall small effects on the extracted features. To increase sensitivity, imaging and data extraction processes could be improved. We illustrate how textured hair experiments should combine single and tress data of relevance to its core characteristic. The application of computer vision and machine learning is essential for improving the process of curly hair assessment at fibre level.