Sociedade Brasileira de Dermatolodia Surgical & Cosmetic Dermatology


ISSN-e 1984-8773

Volume 1 Number 1

Back to summary


Original Article


Avaliação da intensidade das rugas periorbitais por processamento digital de imagens: um estudo de validação

Vínicius C. Brand1, Rosilene C. A. de Souza1, Hélio Pedrini1, Hermênio C. Lima1

Received on 12/02/2009
Approved on 15/02/2009
We declare no conflict of interest.



Introduction: Dermatologists get an idea of how old a patient is by looking at his/her facial wrinkles. There are several methods to quantify skin microrelief and wrinkles.
Objective: To use imaging processing technology to assess the wrinkle area and its interrelation with dermatologists subjective impression.
Material and methods: One hundred seventy-one volunteers, randomly selected, aged 18-69 years, both gender, Fitzpatrick skin type II-IV, with Glogau scale index I-IV were selected. Digital photography was obtained in two different conditions: relaxed or contract wrinkle condition (CWC), and at least 6 repeated pictures for checking the repeatability of each method. Six participating dermatologists evaluated the pictures and ascertained patients’ wrinkle using a digital pen. Wrinkle intensity was done by 3 different image processing methods. Quantitative assessment of facial skin folds intensity, and inter- plus intramethod consistency were determined and compared to manual evaluation.
Results: Dermatologists variability was 61.49%. Intratests agreement varied from 4.4 to 31%. The gradient method had the best results and detected the wrinkle conditions (p < 0.001). Also, this method had a positive correlation with the manual assessment.
Conclusions: Periorbital wrinkles could be determined by digital image processing. Dermatologists had internal precision but low accuracy. Sobel operator digital processing is a valid and reliable instrument for quantitative wrinkle assessment.


Skin changes in structure and elasticity is part of the physiological aging process.1 The rate of this process depends upon genetic aspects and the effects of sunlight exposure.2 Skin aging is divided into chronological aging and photoaging, caused by the cumulative effects of solar ultraviolet radiation.3

Wrinkles are a ridge or crease of skin surface and the most apparent sign of aging. They appear as a result of the aging process itself, habitual facial expressions, sun damage, smoking, poor hydration, among other factors. Wrinkles may cause cosmetic disability and psychological distress.4 Reduction of wrinkles is one of the most sought effects after aesthetic procedures.5

However, the need for evidence-based evaluation of aesthetic procedures dictates the development of more objective and quantitative measures of wrinkles and treatment outcome.6 Several techniques have been developed.7 Some are subjective. Therefore, they are ineffective due to the lack of accuracy and precision.8 Others have been developed which determines directly (in vivo) or indirectly.9,10 However, some are complex to apply in a daily bases due to cost or to the difficulty in obtaining the appliance.11 Nonetheless, digital imaging is a reliable method for storage and retrieval of skin data, image analysis, and allows data interpretation with objectivity by an automated computerized image process.12

The present study was undertaken for evaluation of wrinkle severity determination methods for a general population, comparing the digital and the manual determination.


Study objective
The purpose of this study was to evaluate and compare methods of assessment of periorbital wrinkle by digital image processing.

Study design
This was an experimental, Descriptive, comparative study of wrinkle severity rating based on digital photographs. The study was conducted at a dermatological center in Brazil.
The methods were evaluated and compared in terms of reproducibility and accuracy in assessing the periorbital wrinkle area from digital photos. This study was approved by the appropriate institutional review boards and conducted in accordance with Good Clinical Practice guidelines.

The subjects were screened for eligibility between October 2007 and July 2008. To be eligible for the study, volunteers of both gender had to fulfill the following requirements at the time of the first evaluation: age 18 or older, no history of systemic or topic therapy that may affect the area studied, good general health as determined by history and a physical examination, I to IV at Fitzpatrick phototype13, and photoaging with periorbital wrinkle classified as I to IV at Glogau scale, and be able to refrain from use of any product or system that could intervene with the study. All volunteers were required to give signed consent.

Inclusion and exclusion criteria
At start of the study, the anamnestic data were evaluated including former therapies, skin type classification, and use of any medication (including the use of oral contraceptives). A standardized documentation for basic demographic data was obtained after the volunteer has been selected and their informed consent sign obtained.

Specific criteria for exclusion of the study were: history of skin treatment that interfere in the analysis of the study; skin active disease in the visit of selection; cutaneous signs in the experimental area that could interfere in the evaluation of the skin; continuous treatment with another topical use product in the periorbital areas; pregnant women; nursing mothers; participation in any other clinical trial; any treatment of periorbital wrinkles which may affect the natural development of the wrinkle; any condition that, according to the researchers, could compromise the evaluation of the study; history of lack of adherence to medical regimens or lack of willingness to accede to the study protocol; be relative of members directly involved in the study and their family members; refusal to sign the end of free and informed consent; as well as not present conditions of reproducibility in the collection of digital photos.

Registration database
The center maintained a volunteers database for storage and retrieval of data collected. Each volunteer received a unique number to ensure confidentiality until the end of the study. Each center was responsible for maintaining all the original data, which included the initials and the number of subject of research. Demographic data and analysis of each subject were also collected and included: date of birth, age, phototype according to the Fitzpatrick classification, medical history, diseases, and medications that might interfere with the trial.

Digital pictures
Digital photographs from periorbital area of both sides were obtained for wrinkle evaluation. This approach allowed the investigator to determine the wrinkle intensity based on in the number of pixels in the area and their change over two basic conditions. Briefly, pictures were taken with a Canon EOS Digital Rebel XT with 18 mm-to-55 mm lens. All pictures were audited for any discrepancy in resolution, manipulation, and date or other modification using Zoombrowser EX® freeware. The cameras were adapted on a metal structure and fixed at 50 cm from the subject’s face, under controlled luminosity. The volunteer’s chin was placed on a support and the head fixed by two angle holders (Figure 1A). Lighting conditions were standardized using a GE 250 watt daylight photofood® bluelight bulb placed above the cameras towards the subject’s periorbital area in dark conditions. The whole periorbital area was represented in the picture by centering on the external corner of the eye,. All subjects were photographed without any make-up. Digital images with 1280 x 1024 effective pixels or 1.3 Mb were taken from both sides of the subject’s face at 45 degrees angle. All pictures were used under the same resolution. Two to six sequential repeated pictures from 60 volunteer were used for precision and reproducibility measurements.

Wrinkle determination
Digital pictures from both sides of the face were analyzed. Wrinkle determination was performed as described. A rectangular studied area covering the periorbital area of each side was initially selected from the original picture (Figure 1B). The amount of pixels from the studied area was determined by MATLAB (Figure 1C). The pixels corresponding to the wrinkles were obtained after image processing.

Imaging processing
Manual wrinkle determination: Six participating dermatologists evaluated the pictures in a random sequence and ascertained patients’ wrinkles by using a digital pen on Toshiba R15-S829 Satellite Tablet PC. For tracing, observers were not provided with any wrinkles definition.

Poster Edge: Contrast-enhancing filter Poster Edges from Adobe Photoshop CS2® (Adobe Systems Incorporated, USA) was used to produce special effects on wrinkles inducing edges sharper in the study area of the picture (Figure 2A). Three different filters patterns were applied.12

Sobel operator (gradient operator): The Sobel operator is used in image processing for edge detection algorithms. It calculates the gradient of the image intensity at each point. The result therefore shows image changes at a higher gradient point, which represents the edge of the wrinkle. It has been widely used in neurology field, specifically in MRI image processing.14 Three variations in filters layers were used in the study (Figure 2B).

Canny: The Canny edge detection operator uses a multiple stage algorithm to detect a wide range of edges. Several stages for Canny algorithm application are applied.15 Six different canny gradients have been used (Figure 2C).

MATLAB (MATrix LABoratory – The MathWorksTM, USA): MATLAB is a numerical computing environment and programming language. Although it specializes in numerical computing, an optional toolbox interfaces with the Maple symbolic engine, allowing it to be part of a full computer algebra system.16

Wrinkle area determination
Briefly, manual wrinkle determination, Poster Edge, Sobel operator (gradient operator) or canny running in a MATLAB® (R14) were used in image processing for edge detection algorithms. We calculated the gradient of the image intensity at each pixel point. The higher gradient points, which represent the edge of the wrinkle, were marked in black. The space in the margins was also marked as showed in Figure 2C. Finally, the number of pixels marked was obtained from the software and adjusted to each original selected total of pixels for comparison. The volunteer’s wrinkle area was obtained after calculating the average wrinkle pixels from each side.

Relaxed wrinkle condition (RWC) x contract wrinkle condition (CWC)
One hundred seventy-one digital photographs from periorbital area of both sides in two different conditions were obtained. The volunteer under relax conditions was named “relaxed wrinkle condition” (RWC). The same volunteer was requested to contract its periorbital area at maximal strength. This was called “contract wrinkle condition” (CWC) (Figure 3). Pair-wise difference was used to detect the gradient methods difference by the best graded image process (see Sobel results).

Statistical analysis
The primary outcome was analyzed on quantitative assessment of facial skin folds intensity, and inter- plus intramethod consistency were determined and compared with manual evaluation. Gage R&R was used for measuring how much of the variability was due to operator variation (reproducibility) and to the variation itself (repeatability). Correlations and multivariate techniques were used to determine the interactions among the tests themselves and manual determination. Student t-test was used to determine the RWC and CWC differences. JMP® 6.0 – SAS (SAS Institute Inc., Cary, NC, USA). Statistical differences were considered significant when p = 0.05 for all data.


Manual wrinkle evaluation
When manual wrinkle determination was used, we found that dermatologists’ variability was 61.49%. This indicates a high variability and low accuracy among themselves. However, the mean internal reproducibility was 15.28%. Therefore, they keep marking the same area in different pictures from the same volunteer.

Image processing computer based methods
Intratest (test-retest) agreement varied from 4.4 to 13% among the testes. However, the gradient method (Sobel operator) with two filters had the best results for repeatability, 4.4%. The gradient method with one or two filters had a positive correlation with the manual assessment. The coefficient of correlation was r = 0.771 and 0.662 (p < 0.001) with one or two filters, respectively. From linear regression technique, we have obtained an adjusted R2 of 0.5934 (p < 0.001).

Relaxed (RWC) x Contract (CWC)
We decided to apply the RWC-CWC pair-wise difference that could be detected by the gradient methods (Sobel operator). Correspondence could differentiate from 417.52 ± 40.3 pixels (Mean ± and Std error) to 789.09 ± 41.98 pixels, respectively (Figure 4; p < 0.001). Further analysis had shown that a difference lower than 10% from the previous picture could be detected by the Sobel operator, which means that the sensibility of the method is lower than 0.01.


This trial raised some methodological issues. What are the right outcome criteria for studies evaluating the efficacy of products for wrinkles? There is no standardized system for recording and interpreting wrinkles intensity, although some methods and devices have been developed.17 Moreover, most of those methods for wrinkles evaluation are not practical for use in a busy clinical practice.18,19 One may wonder whether the efficacy of the cosmetic products might be compromised by applying methods for wrinkle analysis that are based on subjective approach.20 For instance, analysis using the Glogau scale has not shown any difference among the groups (data not shown). Therefore, precision varies considerably among different studies, and this variability may be related to patient and physician opinion. In this study, we used an objective, precise, practical method for wrinkles analysis by digital photography developed from a method for objective evaluation of weal and flare skin reaction.12 This method reduces the inter- and intraobserver variability inherent to any method that requires a human component and allows different research centers and physicians to compare their results in a multicenter, randomized clinical trial. This was especially true for Sobel operator running in MATLAB. It showed high accuracy, precision and sensitivity. The high sensitivity could explain the ability to detect variations in a short-term therapy (studies not shown).

There are limitations on this digital photographic system. Good quality and reproducible pictures conditions are necessary for analysis from each independent center. We attempted to maintain the same standard and reduce inadvertent mistakes by training the group of investigators. In a prospective study, this aspect was responsible for 48% of the uniformity lost among the centers and treatment groups of the initial potential sample (paper submitted). Moreover, even when good conditions were obtained, some variations in data collection that could affect the wrinkles intensity could not be avoided, such as smiling behavior or periorbital muscles contraction. However, the sample size and data analysis obtained results point to a correct estimated and statistically significant result.

Finally, we could determine from this study that dermatologists showed high internal precision but low accuracy to evaluate wrinkles. Sobel operator (gradient operator) digital processing is a valid and reliable instrument for quantitative assessment of facial skin wrinkles, with good inter- and intraobserver consistency. Since it allows objective and reproducible grading of data, the gradient method is a useful clinical tool for assessing the effectiveness of wrinkle treatment or other facial procedures.


1 . Oikarinen A, Karvonen J, Uitto J, Hannuksela M. Connective tissue alterations in skin exposed to natural and therapeutic UV-radiation. Photodermatol. 1985;2(1):15-26.

2 . Fisher GJ, Wang ZQ, Datta SC, Varani J, Kang S, Voorhees JJ. Pathophysiology of premature skin aging induced by ultraviolet light. N Engl J Med. 1997;337(20):1419-28.

3 . Gilchrest BA. A review of skin ageing and its medical therapy. Br J Dermatol. 1996; 135(6):867-75.

4 . Sriprachya-Anunt S, Fitzpatrick RE, Goldman MP, Smith SR. Infections complicating pulsed carbon dioxide laser resurfacing for photoaged facial skin. Dermatol Surg. 1997;23(7):527-35; discussion 35-6.

5 . Fodil-Bourahla I, Bizbiz L, Schoevaert D, Robert AM, Robert L. Effect of L-fucose and fucose-rich oligo- and polysaccharides (FROP-s) on skin aging: penetration, skin tissue production and fibrillogenesis. Biomed Pharmacother. 2003;57(5-6):209-15.

6 . Leveque JL. EEMCO guidance for the assessment of skin topography. The European Expert Group on Efficacy Measurement of Cosmetics and other Topical Products. J Eur Acad Dermatol Venereol. 1999;12(2):103-14.

7 . Baret M, Bensimon N, Coronel S, Ventura S, Nicolas-Garcia S, Korichi R, et al. Characterization and quantification of the skin radiance through new digital image analysis. Skin Res Technol. 2006;12(4):254-60.

8 . Glogau RG. Aesthetic and anatomic analysis of the aging skin. Semin Cutan Med Surg. 1996; 15(3):134-8.

9 . Jacobi U, Chen M, Frankowski G, Sinkgraven R, Hund M, Rzany B, et al. In vivo determination of skin surface topography using an optical 3D device. Skin Res Technol. 2004; 10(4):207-14.

10 . Fukuda Y, Fujimura T, Moriwaki S, Kitahara T. A new method to evaluate lower eyelid sag using three-dimensional image analysis. Int J Cosmet Sci. 2005;27(5):283-90.

11 . Lee J, Jung E, Lee H, Seo Y, Koh J, Park D. Evaluation of the effects of a preparation containing asiaticoside on periocular wrinkles of human volunteers. Int J Cosmet Sci. 2008; 30(3):167-73.

12 . dos Santos RV, Titus RG, Lima HC. Objective evaluation of skin prick test reactions using digital photography. Skin Res Technol. 2007;13(2):148-53.

13 . Nagashima H, Hanada K, Hashimoto I. Correlation of skin phototype with facial wrinkle formation. Photodermatol Photoimmunol Photomed. 1999;15(1):2-6.

14 . Zaini MR, Forest GJ, Loshek DD. Image processing for IMRT QA dosimetry. J Appl Clin Med Phys. 2005;6(4):14-21.

15 . Pellegrino FA, Vanzella W, Torre V. Edge detection revisited. IEEE Trans Syst Man Cybern B Cybern. 2004;34(3):1500-18.

16 . Vauhkonen M, Lionheart WR, Heikkinen LM, Vauhkonen PJ, Kaipio JP. A MATLAB package for the EIDORS project to reconstruct two-dimensional EIT images. Physiol Meas. 2001;22(1):107-11.

17 . Friedman PM, Skover GR, Payonk G, Geronemus RG. Quantitative evaluation of nonablative laser technology. Semin Cutan Med Surg. 2002;21(4):266-73.

18 . Friedman PM, Skover GR, Payonk G, Kauvar AN, Geronemus RG. 3D in-vivo optical skin imaging for topographical quantitative assessment of non-ablative laser technology. Dermatol Surg. 2002;28(3):199-204.

19 . Funasaka Y, Sato H, Usuki A, Ohashi A, Kotoya H, Miyamoto K, et al. The efficacy of glycolic acid for treating wrinkles: analysis using newly developed facial imaging systems equipped with fluorescent illumination. J Dermatol Sci. 2001;27 Suppl 1:S53-9.

20 . Harmon CB, Hadley ML. A cosmetic approach to cutaneous defects. Atlas Oral Maxillofac Surg Clin North Am. 2004;12(1):141-62.

Facebook Twitter Linkedin

© 2021 Sociedade Brasileira de Dermatologia - Todos os direitos reservados

GN1 - Sistemas e Publicações