burns 41 (2015) 1043–1048
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/burns
ICG angiography predicts burn scarring within 48 h of injury in a porcine vertical progression burn model§ Mitchell S. Fourman 1,*, Peter McKenna, Brett T. Phillips 2, Laurie Crawford, Filippo Romanelli 3, Fubao Lin, Steve A. McClain, Sami U. Khan, Alexander B. Dagum, Adam J. Singer, Richard A.F. Clark Stony Brook, New York, NY 11794, United States
The current standard of care in determining the need to excise and graft a burn remains with
Accepted 2 November 2014
the burn surgeon, whose clinical judgment is often variable. Prior work suggests that
the predictive capabilities of Laser Doppler Imaging (LDI) and indocyanine green dye (ICG)
angiography in the prediction of burn scarring 28 days after injury using a previously
minimally invasive perfusion technologies are useful in burn prognostication. Here we test
validated porcine burn model that shows vertical progression injury. Twelve female York-
shire swine were burned using a 2.5 2.5 cm metal bar at variable temperature and
application times to create distinct burn depths. Six animals (48 injuries total) each were
analyzed with LDI or ICG angiography at 1, 24, 48, and 72 h following injury. A linear regression was then performed correlating perfusion measurements against wound contraction at 28 days after injury. ICG angiography showed a peak linear correlate (r2) of .63 (95% CI .34 to .92) at 48 h after burn. This was significantly different from the LDI linear regression ( p < .05), which was measured at r2 of .20 (95% CI .02 to .39). ICG angiography linear regression was superior to LDI at all timepoints. Findings suggest that ICG angiography may have significant potential in the prediction of long-term burn outcomes. # 2014 Elsevier Ltd and ISBI. All rights reserved.
Over 175,000 burns received hospital management in the United States in 2013, with average per-patient costs of nearly, $100,000 . Without early intervention, some partial thickness
burns progress to full thickness injuries, exposing the patient to substantial morbidity and undergoing excision and grafting . Despite multiple advances in acute treatment and surgical management, the decision to debride a burn still largely relies on the experience and visual perception of the individual surgeon . Because early clinician accuracy is limited, ranging
§ Study performed at: T19-069 Health Sciences Center, Department of Surgery, Stony Brook University Medical Center, Stony Brook, New York, 11794, USA. * Corresponding author at: Kaufmann Medical Building, University of Pittsburgh Medical Center, Suite 911, 3471 Fifth Avenue, Pittsburgh, PA, 15213, United States. Tel.: +1 631 513 9369. E-mail address: [email protected]
(M.S. Fourman). 1 Current address: University of Pittsburgh Medical Center, Pittsburgh, PA, United States. 2 Current address: Duke University Medical Center, Durham, NC, United States. 3 Current address: New York College of Osteopathic Medicine, Old Westbury, NY, United States. http://dx.doi.org/10.1016/j.burns.2014.11.001 0305-4179/# 2014 Elsevier Ltd and ISBI. All rights reserved.
burns 41 (2015) 1043–1048
from 50 to 75% [4,5], it is often necessary to await burn margin declaration prior to surgical management. Delayed treatment and inaccurate diagnosis increases hospital stays, patient morbidity and increased hypertrophic scarring . Without definitive treatment, current work has focused on early diagnosis of burn wound depth to optimize the management of burns. Perfusion analysis appears to be a viable modality to this end and Laser Doppler Imaging (LDI) has long been considered the gold standard of burn perfusion measurement [7,8]. Since the benchmark work of Pape et al., in which LDI was shown to predict burn depth with a 97% accuracy , multiple clinical and animal trials have shown that LDI can predict burn qualities at a far greater accuracy rate than visual assessment alone [10,11]. Work by members of our group and others suggests that enhanced accuracy with LDI peaks at roughly 48 h following injury, as compared to four days or more with clinical judgment alone . However, as LDI demands a motionless patient and requires a significant scan time, clinical implementation remains problematic. Indocyanine green dye (ICG) angiography has recently been presented as an alternative to LDI in burn analysis. A near infrared (800–830 nm) fluorescent dye previously FDA approved for coronary artery bypass graft patency analysis, ICG angiography is widely applied off-label in plastic and vascular surgery [13,14]. It also benefits from rapid image acquisition and movie sequence video capture. In prior work by our group, we derived an algorithm for the prediction of lateral burn progression in a validated porcine ‘‘hot comb’’ model that’s effective as early as one hour after-injury . While the ability of ICG angiography to demarcate horizontal burn margins is important to definitive burn management, it fails to fully characterize the utility of the technology in predicting partial to full thickness vertical burn progression. Furthermore, no technology to our knowledge is capable of predicting wound contraction and depth of scarring. We therefore sought to (1) Assess a relationship between perfusion readings using LDI and ICG angiography at early time points and 28 day burn contraction and (2) Determine the comparative efficacy of LDI and ICG angiography to determine 28 day burn scar depth.
animals was in accordance with National Research Council guidelines .
Experimental burn creation was performed using a methodology previously described by Singer et al. . Burns were created using a 150 g, 2.5 2.5 cm2 aluminum bar preheated in a 100 8C water bath, wiped dry prior to skin application. The aluminum bar was applied using a previously described spring-loaded device designed to maintain a controlled pressure on the skin of 2 kg/6.25 cm2. A total of eight burns were placed at equal increments along each animal’s paramedian spine, at one of four temperature/time combinations—predetermined prior to the experiment. These consisted of 6 wounds at 70 8C for 20 s, 6 at 70 8C for 30 s, 28 at 80 8C for 20 s, and 8 at 80 8C for 30 s. The non-uniform breakdown of burn intensity was to create a wide range of scar outcomes, as previously reported. The non-regular distribution of burn type on each animal was as a result of the joint requirements of ours and other studies utilizing these burn models. These other studies had aims that were incongruent to our own, and with protocols that did not alter the outcome of this study. The outline of each burn was demarcated using a black tattoo pen. Wounds were covered with a polyurethane occlusive dressing (TegadermTM, 3 M, St. Paul, MN), before being wrapped with gauze and adhesive bandage. Dressing changes were performed daily for the first 3 days after burn, then at 7, 10, 14, 21 and 24 days. Intramuscular buprenorphine 0.01 mg/kg was administered as needed if the animals showed signs of pain and discomfort. All pigs were euthanized using intravenous pentobarbital 28 days following injury. Wound contraction was measured via photographic assessment. Images of specific wounds were taken with a Canon Powershot Duo (Canon, Melville, NY). Changes in the surface area of each tattooed burn margin were measured using NIH ImageJ (Bethesda, MD), with percentage change attributed to wound contraction. Normalization between acquisition distances was performed using ImageJ readings of a ruler present in each image.
Materials and methods
We performed a pilot study utilizing blinded perfusion measurements of burns created with a previously published and validated vertical progression porcine burn model . All animal care and procedure protocols were approved by the Institution Animal Care and Use Research Committee (IACUC).
All experimental procedures were performed at our Division of Laboratory Animal Research (DLAR). Twelve young female Yorkshire swine (20–25 kg) were used in this study. Animals were fed a standardized diet for several days to acclimate, and then fasted overnight before procedures. Housing and care for
Burn creation and contraction measurement
Scar depth measurement
Prior to euthanasia, 6 mm punch biopsies were taken from the center of each burn wound and fixed, dehydrated, and H&E stained as previously described . Scar tissue depth was determined by vertical measurements from the basement membrane to the interface between scar and subcutaneous tissue (Fig. 1). Three separate vertical measurements were performed on each bisected face of the 28 d burn scar: one at the center of the image and 2 midway between the center and edge of the specimen. The average of these 6 measurement was defined as burn scar depth. All measurements were performed by a board-certified dermatopathologist (S.A.M.) blinded to burn conditions.
LDI and ICG angiography were performed 1, 24, 48, and 72 h after burn utilizing a technique previously described in our lab
burns 41 (2015) 1043–1048
Fig. 1 – Specimen obtained from a scar 28 days followingburn and stained with H&E. From each bisected half of the biopsy, scar depth measurements were obtained: one from the center and two mid-way between the center and edge of the image (vertical black thin lines). Average of the six vertical lengths was defined as burn depth. Bar = 1 mm.
(Fig. 2). At each time point burns were evaluated for associated trauma, significant tissue compromise, or infection. In these cases, the wound was excluded from evaluation. In all cases, an operator blinded to the protocol acquired all images. Perfusion analysis was performed by a different, similarly blinded, investigator. To allow for independent comparison of perfusion measurements, only one perfusion technology was utilized on each pig. LDI imaging was performed using a Moor LDI scanner (Moor Instruments, Wilmington DE), with the camera centered at the animal’s spine at a distance of 30.48 cm from the skin. The scanner allowed for an image size of 16 cm 10 cm, with a resolution of 229 143 pixel. This permitted simultaneous capture of all burns in a single image (total = 56 burns total).
Laser ramp speed was 10 ms/pixel, with total scan time of roughly 6 min. Analysis was performed utilizing Moor LDI Image Processing v5.3 (Moor Instruments, Wilmington, DE). Analysis consisted of the creation of ‘‘regions of interest’’ (ROI), in which the edges of the wound were demarcated and a perfusion average was returned. In each case, the tattooed margins were utilized. No normalization was performed on these values, as optic feedback and anatomic plane was deemed consistent for each group of burns. ICG angiography analysis was performed using the LifeCell SPY-Elite (LifeCell, Branchburg, NJ), with the camera centered 30.48 cm from the skin. For each of eight wounds (n = 56 wounds), the camera was shifted to center directly perpendicular to each wound, with consistent target distance aided by a built in laser target. A total of 3 cm3 of 2.5 mg/ cm3 ICG was injected via an ear vein venous catheter. Image acquisition was permitted no sooner than 60 s after injection, and no later than 2 min following injection. This is in line with a ‘‘plateau perfusion range’’ found in prior work by our group . Each wound was scanned for at least 3 s to obtain a fixed signal return. SPY-Q Software (LifeCell, Branchburg, NJ) was used to analyze specific points of interest along burn wounds. For each burn, no fewer than 6 independent points of interest were defined, with the geometric location of these points consistent between wounds. To allow for the appropriate normalization given a variable ICG dose and anatomic plane for each injury, all measured perfusion values were reported as a percentage of the maximum perfusion of normal skin measured in the specific view in which the burn was analyzed.
Burn perfusion was plotted as a function of 28 day contraction for each independent animal. A linear regression was calculated for the burns on each animal, and average r2 was calculated and compared between LDI and ICG angiography using Prism 6.0 (GraphPad, LaJolla, CA). Composite data-set measurements for each technology were obtained as well. Significance was determined using a One-Way ANOVA with Tukey range test to assess for significance ( p < .05) between means.
Fig. 2 – Experimental burn immediately following wound creation (left), with LDI and ICG angiography imaging showing changes in blood flow from 1 to 72 h from burn. Image of 28 day clinical outcome on right.
burns 41 (2015) 1043–1048
Table 1 – Comparative p and r-squared values of composite data set, blood flow measurement vs. 28 day wound contraction. 1h r Laser Doppler ICG angiography *
24 h 2
48 h 2
72 h 2
p-Value .0009* .0010*
Statistical significance p < .05.
Six pigs were analyzed using ICG angiography, while a different six were monitored using LDI. This was to minimize extensive anesthesia delays due to scan setup and time. Porcine weight and age, were similar between groups. No infectious or significant traumatic complications were noted for animals in either group. When all data points from all animals were assessed as a composite set, analysis of LDI and ICG angiography perfusion versus wound contraction resulted in poor linear regression outcomes. Peak r2 values were observed 72 h from injury at .23 for ICG angiography and .29 for LDI (Table 1). We therefore chose to assess linear regressions performed within each individual animal. While LDI failed to show a strong adherence to a linear trend at any time point using this technique, ICG angiography findings improved to a peak r2 of .63 (95% CI .34 to .92) at 48 h after burn. This was significantly different from the LDI linear regression ( p < .05), which was measured at r2 of .20 (95% CI .02 to .39). ICG angiography r2 remained high at 72 h following injury, measuring .51 (95% CI .25 to .77) (Fig. 3). No significant
Fig. 3 – Graph showing comparative adherence of LDI and ICG angiography perfusion measurements to a linear trend (r2) as a function of 28 day wound contraction, when each animal was assessed as a separate dataset. Significance ( p < .05) was observed between LDI and ICG angiography at 48 h after injury.
differences were measured between the average ICG angiography linear regression at 24, 48, or 72 h following injury. When perfusion trends with either device were compared with scar depth as measured histologically, a poor adherence to a linear trend was found when all points were assessed as a composite set (data not shown). When each animal was considered separately, findings were still poor. ICG angiography showed a peak adherence to linear trend of .36 (SEM .14) at 24 h after injury. LDI performed poorly at all time points (Fig. 4). A linear regression performed between 28 day wound contraction and scar tissue depth was also limited, with an r2 of .14.
Burn contracture remains a major challenge in terms of both wound management and long term patient morbidity. As no therapy to reduce scarring has been fully validated, attention has instead turned to the early prediction and excision of burns likely to progress to hypertrophic scarring and contracture. Here we compared the efficacy of ICG angiography and
Fig. 4 – Graph comparing the adherence of LDI and ICG angiography to a linear trend as a function of burn scar depth as measured 28 days following injury by a blinded dermatopathologist. No significant differences between either technology was noted, and overall trend resolution was poor.
burns 41 (2015) 1043–1048
LDI in the projection of long term scarring in a validated porcine vertical progression burn model. ICG angiography exhibits a strong adherence to a linear trend in the prediction of wound contraction 48 h after burn. However, we found that the individual animals themselves displayed radically different relationships between wound perfusion and 28 day wound contraction . While linear trends between ICG angiography measurements and contraction existed in each animal, the slope and degree of this relationship prevented effective composite analysis. While further investigation is warranted to fully elucidate the nature of this variance, we may attempt to equilibrate this high variability with the phenomenon of hypertrophic burn scarring in humans. We suggest that quantitative blood flow measurements may be only part of a greater formula necessary to predict long term clinical outcome. This assertion is aided by our finding that quantitative scar depth is of limited value in this assessment, and poorly correlated with both perfusion measurement and wound contraction. We found that LDI is of limited predictive value in evaluating the scarring of non-surgically managed burns. While LDI’s clinical potential for highlighting skin injury has been repeatedly touted in prospective studies, it appears to be in fact quite limited in addressing more severe injuries. Recent work by Park et al. and Stewart et al. use both clinical decisionmaking and quantitative histology to suggest that while partial thickness burns with extremely high LDI perfusion measurements appear to be correlated with non-operative management as per physician decision, intermediate perfusion measurements are highly variable and of no clinical utility [10,18]. We also found a similar difficulty interpreting intermediate and low LDI perfusion measurements in our previous work with a porcine horizontal burn progression model, a difficulty not observed with ICG angiography. A reasonable explanation for this obstacle is the penetration depth of both technologies. LDI has been shown to measure perfusion up to one mm beneath the skin surface, while ICG angiography has a reported penetration depth of up to 1 cm . This supports our belief that LDI may be able to measure the hyperemic response of very superficial injuries, but is unable to distinguish between deeper partial thickness injuries. While we are encouraged by the potential suggested by our early findings, we acknowledge several limitations to our measurements. First, we accept that full scarring does not take place in burn wounds after only 28 days. We noted that in a majority of burns, fresh granulation tissue persisted at the centers. Therefore, we acknowledge that our findings are functions related to an incomplete scarring process, although there is no indication that the relative size of these scars would be significantly different. Future work would benefit from a longer time course allowing for complete, definitive scarring and supplemented by more detailed tissue histomorphologic scoring. In addition, perfusion measurements from LDI and ICG angiography were analyzed using very different techniques—absolute perfusion for LDI vs relative perfusion with ICG angiography. While this is a reflection of techniques that have either worked for our group clinically or on prior studies [12,13], it is difficult to say with absolute certainty whether our mode of analysis was the most effective for each technology.
It therefore may be reasonable that other forms of analysis may yield different results for the same set of wounds. Pigs are a reasonable representation of human skin, and widely used to simulate human dermal wound healing. However, we lack a full clinical correlate for the utility of burn projections using either technology. We therefore plan to justify our findings here using a prospective clinical study in our burn unit. Finally, we describe experimental measurements of burn wounds using these technologies. These measurements were only correlated with the contraction of an experimental wound. However, we cannot make the assumption in our findings that contracted vs. preserved injuries are black and white concepts and universally translatable into the clinic and surgical decision-making. Logic would suggest that were clinical assessments made of our standardized burns, it would be determined that occasional burns that go on to significant contracture would not benefit from excision, while other preserved injuries may be marked for surgical management. Indeed, we hope our findings serve to highlight that the scar process after burns is a complex and multi-factorial process. While we hope our technique is one piece of the puzzle, it is surely not the whole answer. As we did not attempt a clinical correlate, we do not expect our findings to discredit or supplant prior evidence that LDI is effective clinically. Instead, we hope that our findings serve as the basis for a large scale and well-designed clinical study–where such conclusions may be proposed.
Here we present evidence that indocyanine green dye angiography possesses the ability to project burn scarring and outcome at early time points. This long-term prognostic utility is significantly greater than laser Doppler, the traditional gold standard. Our studies add to the growing body of literature that ICG angiography can yield quantitative, highresolution data that can be applied towards acute and long term wound healing prognostic outcomes. However, the lack of a definitive, long term clinical study severely limits the utility of any perfusion technology.
Conflict of interest statement No author has any financial, personal, or other relationship with any individuals or organizations that could inappropriately influence this work. Funding source had no involvement in this study.
 Association AB. 2013 National Burn Repository: Report of Data from 2003–2012. American Burn Association; 2013, http://www.ameriburn.org/2013NBRAnnualReport.pdf.  Moritz AR, Henriques FC. Studies of thermal injury: II. The relative importance of time and surface temperature in the causation of cutaneous burns. Am J Pathol 1947;23(5): 695–720.
burns 41 (2015) 1043–1048
 Mandal A. Burn wound depth assessment—is laser Doppler imaging the best measurement tool available? Int Wound J 2006;3(2):138–43.  Jaskille AD, et al. Critical review of burn depth assessment techniques: Part II. Review of laser Doppler technology. J Burn Care Res 2010;31(1):151–7.  Sharma VP, O’Boyle CP, Jeffery SL. Man or machine? The clinimetric properties of laser Doppler imaging in burn depth assessment. J Burn Care Res 2011;32(1):143–9.  Aarabi S, Longaker MT, Gurtner GC. Hypertrophic scar formation following burns and trauma: new approaches to treatment. PLoS Med 2007;4(9):e234.  Rousseau AF, et al. Toward targeted early burn care: lessons from a European survey. J Burn Care Res 2014;35(Jul–Aug (4)):72–7.  Gill P. The critical evaluation of laser Doppler imaging in determining burn depth. Int J Burns Trauma 2013;3(2):e234–9. http://www.ncbi.nlm.nih.gov/pubmed/?term=.+Toward +targeted+early+burn+care%3A+lessons+Q4+from+a +European+survey#.  Pape SA, Skouras CA, Byrne PO. An audit of the use of laser Doppler imaging (LDI) in the assessment of burns of intermediate depth. Burns 2001;27(3):233–9.  Park YS, et al. The impact of laser Doppler imaging on the early decision-making process for surgical intervention in adults with indeterminate burns. Burns 2013;39(4): 655–61.
 Nguyen K, et al. Laser Doppler imaging prediction of burn wound outcome in children: is it possible before 48 h? Burns 2010;36(6):793–8.  Fourman MS, et al. Indocyanine green dye angiography accurately predicts survival in the zone of ischemia in a burn comb model. Burns 2013.  Fourman MS, et al. Comparison of laser Doppler and laserassisted indocyanine green angiography prediction of flap survival in a novel modification of the McFarlane flap. Ann Plast Surg 2014.  Fritz JR, et al. Comparison of native porcine skin and a dermal substitute using tensiometry and digital image speckle correlation. Ann Plast Surg 2012;69(4):462–7.  Singer AJ, et al. Validation of a vertical progression porcine burn model. J Burn Care Res 2011;32(6):638–46.  Garber J, et al. Guide for the care and use of laboratory animals, vol. 8. Washington DC: The National Academic Press; 2010. p. 220.  Phillips BT, et al. Intraoperative perfusion techniques can accurately predict mastectomy skin flap necrosis in breast reconstruction: results of a prospective trial. Plast Reconstr Surg 2012;129(5):778e–88e.  Stewart TL, et al. The use of laser Doppler imaging as a predictor of burn depth and hypertrophic scar postburn injury. J Burn Care Res 2012;33(6):764–71.  Marshall MV, et al. Near-infrared fluorescence imaging in humans with indocyanine green: a review and update. Open Surg Oncol J 2010;2(2):12–25.