COLOR RETENTION AND BLEEDING POTENTIAL OF A LIPSTICK 405 STATISTICA L ANALYSIS Binomial p robability method was used to test the statistical signifi cance of the panel perception results. Linear interpolation for threshold value between two measured datapoints was performed by determining the lowest possible percentage of panelists who perceived a color parameter acceptable while maintaining the statistical signifi cance (p 0.05 at the 95% confi dence level). RESULTS AN D DISCUSSION QUANTIFICA TION OF COLOR RETENTION AND BLEEDING Our automa ted lip detection algorithms adequately detected the lip area from the clinical im- ages, such as those illustrated in Figures 1 and 2, which made it possible to measure color reten- tion and determine bleeding potential 8 h after application. A mean color retention of 82.67% with a standard deviation of 8.27% was obtained from the test population in this study, as shown in Figure 3A. The observed variability was believed to come from the life habits of partici- pants who paid variable attention to keeping the color on their lips in their daily activities. For color bleeding potential (Figure 3B), the results fell in a linear scale ranging from zero to about 17, with a mean of 2.43 and a standard deviation of 2.97. It was noticed that most of the volunteers did not show much color bleeding around the lips, although the phenomenon was signifi cantly visible on individuals with excessive wrinkles in the mouth area, skewing the data toward the higher end of the bleeding potential range. PANEL PERC EPTION The result s of panel-perceived acceptability on lip color retention are shown in Table I, where the percentages of the panel who perceived pairs of before and after images as “acceptable” are displayed together with the corresponding lip images. The p values of binomial probability tests were also shown to validate the perception results with statistical signifi cance. For color bleeding, although the perception survey asked for unacceptability, we converted the results to acceptability, that is, %Panel (acceptable) = 100%-%Panel (unacceptable), to make the format of analysis consistent. The results are shown in Table II in the fashion similar to those of color retention. Table I Peception Results on Acceptability of Lip Color Retention Levels

JOURNAL OF COSMETIC SCIENCE 406 CONSUMER RE LEVANCE OF THE QUANTIFIED LIP COLOR PROPERTIES Combining p anel perception results with their corresponding levels of color reten- tion and bleeding potential measured objectively, we obtained charts to demonstrate consumer relevance. As shown in Figure 4A, the percentage of panelists who per- ceived the levels of color retention acceptable decreased with the levels of the objec- tively measured property. Between color retention levels of 85% and 77.1%, the percentage of acceptability dropped dramatically from 79.8% to 13.1%, indicating the existence of a threshold in the region. For the ble eding potential, the percentage of panelists who perceived the levels of color bleeding acceptable decreased with increase in the bleeding potential measured objec- tively. Between the levels of 1.94 and 2.72, the percentage of acceptability dropped dramatically from 86.7% to 38.9%, suggesting a threshold in the region, as shown in Figure 4B. Interpolati ng within the threshold region using a binomial probability test at the given population of each perception study and under the condition of p 0.05, we fi rst deter- mined the minimal percentages of panel acceptability to be 60.12% and 60.56% for color retention and bleeding potential, respectively. The corresponding threshold values of color retention and bleeding properties were then derived from the linear equation, with the values being 82.67% and 2.37 for color retention and bleeding potential, respectively. Plot- ting the color parameters of all 30 participants in this study, we obtained Figure 5A and B to show the color properties relative to their threshold of acceptability. Sorting the color retention levels from high to low, we see in Figure 5A a gradual decrease in the property among different participants, which reaches the threshold at 82.67%, beyond which the color retention levels become unacceptable. Similarly, in Figure 5B, we see a gradual in- crease in bleeding potential reaching the threshold at 2.37 beyond which the bleeding level is too high to be acceptable. With these charts, we would be able to quantitatively evaluate the performance of a new lip color formulation during product development and fi nal claims’ substantiation. After using it for an expected time period such as 8 h in a usability study, should the measured mean value of color retention be higher than the threshold of 82.67%, or the mean value of bleeding potential be lower than 2.37, it would indicate with statistical confi dence that a claim of “long-lasting” or “non-bleeding” be substantiated. Table II Perception Results on Acceptability of Lip Color Bleeding Levels