Researchers at TRI, in partnership with Newtone Technologies, have pioneered the use of an image analysis algorithm to glean further insight into hair breakage and support claim substantiation. The number of broken hairs, the length of individual broken hairs, and the total length of broken hair can now be determined quickly and efficiently in one test, creating new insights and new claim possibilities.
Many consumers comb their hair on a daily basis, and in some cultures combing hair, or grooming, is part of a complex ritual or bonding experience. In terms of sensory attributes and perceived product performance, the experience of the consumer during combing may influence if they decide to (re)purchase a particular product, or not. For example, ease-of-combing or fewer residual hairs on a brush or comb after grooming may be perceived by a consumer to show a quality product.
The process of grooming leads to a combination of bending, torsion, and inter-fiber friction that can result in localized stresses, leading to fiber weakening or breakage. The degree of fiber breakage can be measured using a repeated grooming test and is a common technique for characterising or benchmarking products against each other in relation to tangle-reduction, conditioning, or substantiating claims. During this test, tresses are brushed or combed up to 10,000 times and the number of hairs that break off are counted manually. Although this simple test provides a numerical value for the number of broken hairs collected, valuable information, for example the length distribution of the broken hairs, which could provide richer, more robust data, is overlooked due to time and labor constraints.
In recent years, TRI Princeton, in conjunction with Newtone Technologies, have re-visited this established technique, to glean further insight into hair breakage, further support claim substantiation, and modernise this testing stalwart. The crux of the improvement is the use of an algorithm to analyse images of the broken hairs collected, Figure 1a and 1b. This allows both the number of broken hairs, as well as the length of each broken hair, to be ascertained, enabling extensive interrogation of differences between products, and generation of data to support claims.
To validate this new technique, experiments were undertaken where broken hairs were both counted manually and by using the image analysis algorithm; the correlation between these two techniques was excellent, Figure 1c. In addition, the length of each broken hair could be determined, which was previously difficult due to time and labor constraints, allowing for even more rigorous analysis and interrogation of results.
From our experiments there were three main findings:
1. The lengths of broken hairs collected do not follow a ‘normal’ distribution
It was observed that most broken hairs were short in length and a small number were very long in length, leading to a large skew of the data, Figure 2a. This distribution also meant that there was a large discrepancy between the mean break-length (indicated by the horizontal ◊) and the median (indicated by the horizontal line in the box). When the log of each hair’s length was taken, the values were more Gaussian in character, meaning that the mean and median data values were closer to each other. Use of logged data allowed the use of standard parametric tests, such as ANOVA and t-Tests, to interrogate the data.
2. The quality of hair conditioning leads to differences in breakage length
It was noted during initial experiments that tresses that had been subjected to conditioning experienced shorter breakages, on average, than the control tresses, Figure 3a. For example, after 2000 combing cycles the SLES-treated control had average breakage length of 6.24 mm, whereas the tress treated with shampoo had average breakage length of 5.12 mm and the tress treated with shampoo and conditioner had average breakage length of 4.53 mm. When final data are compared directly, Figure 3b, the differences are clear. The error bars in Figure 3b represent geometric standard deviations and are skewed representing the skew in the raw data. Since the number of broken fibers counted was very high (n > 1300), the statistical differences between all the treatments were very strong (p<0.05).
3. The total length of broken hair can be quantified
The total length of broken hair can provide another means for quantification of hair breakage and comparison between samples. Totalling the lengths of hair lost after each treatment (rather than just the number of broken hairs, as in the traditional technique) allows for quantification of the total amount of hair lost. This can reveal significant differences between samples, aiding with claim substantiation. For example, during an experiment where a control tress was compared with a tress treated with both shampoo and conditioner, the cumulative number of hairs lost (i.e. the ‘traditional method’) showed a 3-fold reduction in breakage, but the total length of hair lost showed a 5-fold reduction in breakage.
In summary, the use of an image analysis algorithm within the repeated grooming test provides valuable insights into the performance of a product and allows for facile comparison with competitors. Statistical analysis of the length of each broken fiber, and the total length of all broken fibers, is game-changing, providing robust, quantifiable data from one simple test. This work is likely to be of interest to new product developers, formulation scientists and during claim substantiation.
Comments