Skin imaging modalities quantify progression and stage of infantile haemangiomas

DEAR EDITOR, Infantile haemangiomas (IH) are deep, superficial or mixed vascular neoplasms with rapidly proliferating endothelial cells that stabilize and involute with diminishing cellular activity, apoptosis and resolution over 7–10 years. Clinicians evaluate progression, treatment response and stage by inspection, photography (colour, size, shape) and palpation (temperature, deformability). Lightening, flattening, reduced temperature and softness signal involution. The limitations of subjective methods and need for objective metrics are recognized. Colour, infrared (IR) and three-dimensional (3D) imaging have been applied to IHs. Photographs are compared for improvement as stable/worse (0%), slight (< 25%), moderate (25–50%), good (50–75%) and excellent (> 75%) using the Visual Analog Scale. However, images may be nonstandardized (position, lighting) and results subjective as improvement criteria may be inconsistently applied. IR surface intensity and distribution are affected by tumour depth, size, metabolism, vasculature and perfusion. Dynamic IR applies a temperature stress and the rewarming pattern provides physiological information. We conducted a prospective observational proof of concept study to determine the utility of standardized skin imaging of colour, IR thermography and 3D shape for quantifying IH progression [Clinical Trial Registration: www.clinicaltrials.gov (Identifier NCT02061735)]. We compared these measurements to clinical stage (proliferating, stable, involuting). Data were stratified by age at evaluation, i.e. 1–2, 3–5, 6–9, 10–19 and ≥ 19 months corresponding to expected growth. Fifty-nine patients with 67 superficial or mixed IHs and at least two visits yielded 250 evaluations over 18 months. Deep IHs were excluded due to the small number. Clinicians from our multidisciplinary Hemangioma and Vascular Malformation Center decided the treatment: propranolol (1–2 mg kg 1 daily), topical timolol (one drop 0 5% gel twice daily) or no treatment, and assessed patients in person. The Institutional Review Board approved the research and parents/guardians provided written informed consent. Standardized colour, IR and 3D images of the IH and contralateral controls were taken (Nikon D90 camera, 60-mm lens, 12 3 megapixels, cross polarization, wireless flash; Nikon Corporation, Tokyo, Japan; FLIR T400 IR camera; FLIR Inc., Wilsonville, OR, U.S.A.; 3D scanner; Artec MHT, Artec Group, San Diego, CA, U.S.A., respectively). IR responses to 30-s cooling (18 0 2 °C, Jack Frost; Cardinal Health, McGaw Park, IL, U.S.A.) and rewarming (30 s) were quantified for a subset of evaluations. Photographs were colour corrected, separated into L*, a* and b* images (ImageJ; NIH, Washington, DC, U.S.A.), coregistered with IR images and analysed with a graphical interface employing landmark-based registration and algorithms (MATLAB ; MathWorks, Natick, MA, U.S.A.). Thresholds were applied to identify features of the IH only. Colour and IR intensities and areas were described as means and number of pixels above threshold, respectively. The highest 10% of pixels isolated the IH highest activity for dynamic IR with behaviour measured as area under the curves [AUCcooling (cool), AUCrewarming (rw)]. Height and volume were determined from 3D scans (3dMD Vultus; 3dMD, Atlanta, GA, U.S.A.). Outcomes by time and stage were analysed using univariate general linear models with treatment and depth (time) and age and treatment (stage) as covariates, post-hoc least significant difference for pairwise comparisons (P < 0 05) (SPSS; IBM, Armonk, NY, U.S.A.). Means were 7 4 months, 8 4 months in study, 3 8 evaluations and 3 months between evaluations. Fifty IHs were treated and 51 were mixed. Clinical stage discrepancies were resolved by consensus. Figure 1 shows IH progression for one subject. Lightness intensity and height were the most discriminating outcomes. Lightness increased and height decreased over time (P < 0 05) (Fig. 2a,b; see also Supporting Information). Red intensity was highest at 2 2 and 4 5 months, decreasing by 8 0 months (P < 0 05). IR intensity was higher for 2 2 and 4 5 vs. 12 8 and 27 7 months (P < 0 05) (Fig. 2b). Changes followed a logarithmic pattern. AUCcool AUCrw differences (dynamic IR) were 1 5 6 2, 17 4 3 6, 34 6 5 3, 28 4 5 5 and 20 6 7 3 for 2 2, 4 5, 8 0, 12 8 and 27 7 months, respectively. The smaller difference for 2 2 vs. all other months (P < 0 05) indicates faster recovery. AUCrw was 1109 23, 1054 12, 1009 12, 1004 18 and 1000 24, respectively and greater at 2 2 months (P < 0 05), indicating faster rewarming. The IR area consistently extended beyond the visual IH boundaries (mean visible : IR ratio of 33 18%), suggesting that a third to a half of potentially relevant information is not visible. The IR maps (Fig. 1) show the regions of greatest thermal activity. The results are consistent with the expected changes as higher proangiogenic factors during proliferation cause increased microcirculation followed by vasoconstriction