116 Parameter Expert image score Svm Sz Visual grade Age Thigh subregion % fat Ultrasound surface area Elastic deformation Energy absorption JOURNAL OF COSMETIC SCIENCE Table V Significant Correlations "" 0.4 or Higher (n = 56 subjects) Positively correlated with .... BMI, weight, thigh circumference, Svm, Sz, Spm, Sa, Sq, Sp, St, visual grade, thigh % fat, thigh subregion % fat, red-band SD, energy absorption BMI, weight, thigh circumference, expert image score, visual grade, thigh % fat, thigh subregion % fat, ultrasound surface area, stiffness (neg),* energy absorption, red-band SD BMI, weight, thigh circumference, expert image score, visual grade, thigh % fat, thigh subregion % fat BMI, weight, thigh circumference, Sa, Sq, Spm, Svm, Sz, thigh % fat, thigh subregion % fat, ultrasound surface area, red-band SD, stiffness (neg), energy absorption % elasticity (neg) BMI, weight, thigh circumference, Sa, Sq, St, Spm, Svm, thigh % fat, ultrasound surface area, red-band SD, elastic deformation, energy absorption BMI, weight, thigh circumference, Svm, Sz, thigh % fat, red-band SD, elastic deformation, energy absorption, stiffness (neg) BMI, weight, thigh circumference, Spm, thigh % fat, ultrasound surface area, laxity %, laxity (mm), stiffness (neg), energy absorption, elasticity (mm) BMI, weight, thigh circumference, St, Spm, Svm, Sz, thigh % fat, ultrasound surface area, laxity (mm), elastic deformation, stiffness (neg), elasticity (mm) * Neg: negative correlation. (p = 0.02). The equation was: Svm = -2.2 + (0.08 · Subregion % Fat) + (0.006 · D-S surface area), with R, R 2 , and adjusted R2 values of 0.68, 0.46, and 0.44, respectively. MALE AND FEMALE COMPARISONS Ten females, matched for BMI and age, were selected at random for comparison to the males. A second group of BMI- and age-matched females was used to further verify the results. The average BMI (kg/m2) for the males was 33.2 ± 6, compared to 33.4 ± 6 and 33.0 ± 5.9 for the two female groups. T-test analyses confirmed that the groups were comparable for BMI and age. The males and females differed significantly in body composition (p 0.001) (Table IV). The mean subregion % fat was 20.3 ± 5.4% for the males and 36.S ± 7 .2% and 33.8 ± 9.3% for the females. The mean thigh % fat was 29.0 ± 6.1 % for males and 44.5 ± 5.8% and 42.0 ± 7.3% for females. The mean live visual cellulite grade was directionally higher (p = 0.10) for one female group than for the males. No significant differences were found for the expert image score, thigh circumference, dermal thickness, dermal­ subcutaneous surface areas, and biomechanical properties. With respect to cellulite severity, the males and females differed significantly in Sdr (p 0.005), with substan­ tially higher values for females (Figure 5). DISCUSSION Three-dimensional skin surface microstructures, e.g., wrinkles, have been quantified by various techniques (27-30). Many use surface replicas coupled with mechanical, laser,

QUANTITATIVE MODEL OF CELLULITE 117 60 160 ... 0 140 • 0 50 ... ... t .. 120 • "'iii "O • LL Cf) ... Q) 100 • • C 40 ... ... Q) • 0 ... ... E 80 ·a, ... • � co ..c co 60 :::J (• a.. Cf) 30 tJ) • • ♦ ♦ tJ) • .c Q) 40 • Ol ♦ C i: .c • I- ♦ Ol 20 20 ♦ ♦ :::J • 0 ♦ 0:: 0 ♦ ♦ 10 20 25 30 35 40 45 20 25 30 35 40 45 BMI (kg/m2) BMI (kg/m2} ♦ Male ■ Males ... Females • Females Figure 5. Male and female comparisons. Males had a significantly lower percent fat in the thigh subregion and significantly lower values for Sdr at comparable body mass indices. The graph shows Sdr · 100 for clarity of illustration. For thigh subregion fat versus BMI, the correlation coefficients were 0. 77 (p = 0.0 l) for males and 0.78 (p 0.001) for females. For roughness Sdr versus BMI, the correlation coefficients were 0.69 (p = 0.03) for males and 0.77 (p 0.001) for females. optical or interference fringe profilometry to determine roughness parameters. For cel­ lulite, the area of interest is large and the limitations of replica methods become significant. Akazaki et al. (28) described an optical system for direct skin surface mea­ surements over a 6.4-mm2 area with a high resolution of 12.5 µm. Optical profilometry with CCD sensors allows measurement of depths up to 6 mm (27), and interference fringe projection methods have a 1-mm depth of field (29). These distances are smaller than those encountered with cellulite. Three-dimensional skin surface features have been reported using the non-contact PRIMOS system, which projects parallel stripes onto the surface and determines the third dimension from differences in elevation between the projections and the skin (31,32). Quantitative roughness parameters in the µm range and within a sampling area of 2.4 x 3.0 cm can be measured by this method, indicating the suitability for microtextures. The three-dimensional features of the face and head have been measured with non­ contact laser surface-scanning systems that record the x,y,z coordinates of multiple points across the surface (33,34). Changes in the range of 2-3 mm could be accurately measured over a relatively large surface area. Rohmer et al. (3 5) recently used fringe projection methods to quantify the roughness parameters and volume of cellulite­ affected skin. A system to directly measure 3D surface features and quantify wrinkle depth and width has been reported (28). Wrinkle depth was defined as the maximum distance from top to bottom, similar to St (vertical distance from the highest peak to the deepest valley) in this report. We used a non-contact three-dimensional laser surface scanner (resolution of 0. 5 mm) to compute roughness and to provide a standardized, reproducible measure of severity. This system has been used to quantify the surface features of wounds and burn scars and to generate 3D data for the construction of burn masks (36). Since the laser-scanning technique does not use shadows to create the 3D