Work

Towards three-dimensional temporal monitoring of naevi: a comparison of methodologies for assessing longitudinal changes in skin surface area around naevi

Year: 2016

Type: article

Abstract: Dear Editor, Although change in a skin lesion is a sensitive clue to the diagnosis of melanoma, all skin lesions grow at some point during their lifespan. Few studies have assessed how skin lesions change in size relative to the body. Rhodes et al.1 estimated the relative growth of congenital naevi in infants by comparing changes in naevus surface area (referred to hereafter simply as ‘area’) with regional body area changes calculated using height and weight data and adjusted for age. They found disproportionate growth in the first 6 months followed by proportionate growth thereafter. As body sites change at different rates and are influenced by age, sex and environmental factors, it is unclear whether this approach is uniformly valid or whether other methodologies might be more applicable.1,2 Until recently, measuring change in skin area was not readily feasible (i.e. paper cast or palm techniques),3 but the advent of three‐dimensional (3D) stereophotogrammetry‐based imaging in dermatology clinics4 now offers an opportunity for precise study of perilesional area changes. Here we compared 3D‐imaging‐observed area changes with those calculated from two‐dimensional (2D) imaging or inferred from height and weight data. This is part of the Study of Nevi in Children (SONIC), which to our knowledge is the first to follow naevi longitudinally using 3D imaging. SONIC was approved by the Harvard University Institutional Review Board and adhered to the guidelines of the Declaration of Helsinki. Participant accrual and data collection have been reported previously.5,6,7,8 Relevantly to this study, participants had 180° anterior and posterior 3D body images and 2D overview images of the back and legs captured at baseline (age ~14 years) and 3 years later using standardized poses, along with height and weight measurements. Three‐dimensional images were obtained using a customized Vectra Body System (Canfield Scientific, Inc., Fairfield, NJ, U.S.A.), which had a stereocapture array of twelve 21‐megapixel digital single‐lens reflex cameras configured to meet the study capture field, and photographic texture‐resolution requirements. The calibrated dimensional imaging system collected each image in under 4 ms to produce a 3D model. Vectra systems have been validated for accuracy and are in clinical use.4,9,10,11,12,13 Digital 2D images were obtained using a Phase One P25 Camera Back, Hasselblad 503w Camera System, 2kWatt Studio Flash System (Canfield Scientific, Inc.). The area around naevi (or ‘perilesional area’) was measured on 3D and 2D images using four neighbouring landmark naevi (‘anchors’) that surrounded the naevus. These ‘anchors’ had to be clearly visible at both time points (Fig. 1). When four neighbouring landmark naevi could not be identified, the index naevus was used as the fourth landmark. In a pilot experiment using 3D images of a single participant taken 15 min apart, we explored the precision and repeatability of the 3D Vectra body system and use of perilesional area measurements for 10 back naevi. We found a mean coefficient of variability of 1·04%, supporting the reproducibility of this approach. Perilesional area measurements on three‐dimensional (3D) and two‐dimensional (2D) images. (a) A 3D back image. The perilesional area measurement of the naevus with adjacent teal‐coloured sticker is shown with blue shading and delineated by four neighbouring or ‘landmark’ naevi. (b) Rotated image of part (a) that visually depicts how 3D area measurements account for surface topography. (c) Identical perilesional area measurement performed on a 2D back image. To explore further the area change measurements, we calculated the 3D and 2D regional area, defined as the area of the upper back, lower back or lower leg (Fig. 2), and the total body surface area (TBSA) (TBSA = xHyMz; where x, y and z are constants equalling 0·02667, 0·38217 and 0·53937, respectively; H is height and M is mass).14,15 A convenience sample of 93 naevi from 65 participants (33 male) was chosen from the upper back (n = 29), lower back (n = 21), upper leg (n = 22) and lower leg (n = 21). Feasibility of naevus perilesional area measurements was required for inclusion, and we overselected naevi on the lower back and legs to include similar proportions by anatomical site (Appendix S1 and Tables S1–3; see Supporting Information). Regional area measurements on three‐dimensional (3D) and two‐dimensional (2D) images. (a) A 3D back image. The regional area measurement of the upper back is shown with blue shading and delineated by the nape of the neck to the bilateral acromion processes to the level of the lower tips of the scapulae. (b) Rotated image of part (a) that visually depicts how 3D area measurements account for surface topography. (c) Identical regional area measurement performed on a 2D back image. The mean 3D perilesional area increased after 3 years by 16·1% (95% confidence interval 13·0–19·3). There was nearly perfect inter‐ and intraobserver reliability (intraclass correlation coefficient of 0·99 for both measures). To assess the accuracy of the 3D imaging system and measurement tool, coloured adhesive dots that were originally placed adjacent to naevi during photography sessions for tracking purposes were randomly selected from 10 participants, and their diameters were measured at both time points. The mean diameter was 12·83 ± 0·13 mm, comparable with the manufacturer's reported specification of 12·70 mm. Nonvisible landmark naevi located on curved surfaces precluded 2D perilesional area measurements in 29 cases (31%). For the 64 naevi with both 2D and 3D perilesional area measurements, 3D perilesional area measurements were consistently greater than 2D (baseline 69·0 vs. 59·4 cm2; follow‐up 87·5 vs. 72·6 cm2; P < 0·001), but there was no difference in the mean change between 3D and 2D perilesional area (17·0% vs. 17·6%, P = 0·8). The change in regional area differed from the change in 3D perilesional area but did not reach statistical significance in this small sample (3D perilesional area vs. 3D regional area was 17·9% vs. 20·5%, P = 0·07; 3D perilesional area vs. 2D regional area was 17·9% vs. 13·6%, P = 0·05). However, we did identify a significant difference in the change between 3D perilesional area and TBSA (16·1% vs. 13·3%, P = 0·02). This study is the first to compare methodologies for calculation of area changes around naevi using a reference standard of 3D stereophotogrammetry‐based imaging. We found a significant difference in the mean percentage area change between calculated TBSA and direct perilesional area measurements. This difference can be attributed to disproportionate growth of anatomical sites. Consistently with this interpretation, we found regional area change to be a closer approximation to perilesional area change than TBSA. Although 2D perilesional area appears to be a similarly valid and reliable method for monitoring change in area over time compared with 3D perilesional area, it is not always feasible for lesions bordering sharply curved anatomical sites when landmark lesions cannot be visualized. A limitation of this study is that the perilesional area may notably not be applicable to individuals with few skin lesions. Further study is needed to examine whether the differences in methodologies identified in this study are clinically important. Knowledge of changes in the size of skin lesions relative to their surroundings may someday aid automated melanoma detection efforts, and also provide insight into naevus growth dynamics. E.C. and M.A.M. contributed equally. Funding sources: Research reported in this publication was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748 and supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under Award Number R01‐AR049342, as well as the National Institute of General Medical Sciences of the National Institutes of Health under grant number P41GM103545. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Canfield Scientific, Inc. provided no funding for this work and did not participate in the design or interpretation of the study. Conflicts of interest: A.C.H. reports receiving consulting fees from Canfield Scientific, Inc. D.D. reports being an employee of Canfield Scientific, Inc. Appendix S1. Materials and methods. Table S1. Intra‐ and interobserver agreement measures. Table S2. Summary of paired two‐ and three‐dimensional surface area measurements by location and grade. Table S3. Paired comparisons of average percentage change in surface area between 8th and 11th grades.

Source: British Journal of Dermatology

DOI: 10.1111/bjd.14700