- Email: [email protected]
Review of quantitative outcome analysis of cranial morphology in craniosynostosis
+
MODEL
Journal of Plastic, Reconstructive & Aesthetic Surgery (2016) xx, 1e5
Review
Review of quantitative outcome analysis of cranial morphology in craniosynostosis Mark S. Lloyd, Edward P. Buchanan, David Y. Khechoyan* Division of Plastic and Reconstructive Surgery, University of Colorado, Department of Surgery, Aurora, CO, United States Received 18 February 2016; accepted 22 August 2016
KEYWORDS Craniosynostosis; Outcome measures; Point cloud representation; Stereophotogrammetry
Summary Outcome measures in craniosynostosis surgery have progressed from those based on the need for surgical revision to linear anthropometric measurements, 2D CT vector analysis and 3D CT vector analysis. However, finding an objective means to assess postoperative cranial morphological improvement remains challenging. A critical review of previous studies used to measure craniosynostosis surgery outcomes is presented. We also introduce and briefly discuss the key features of the computational algorithm that is being utilized in our center for evaluating craniosynostosis surgical outcomes. This has addressed a number of the previous challenges encountered in quantitative measurement of cranial morphological change. Point cloud representation and 3D stereophotogrammetry have made it possible to compare pre and post-operative images of children undergoing surgical correction for craniosynostosis. These pre- and post-operative images can also be compared to age, sex and race-matched controls throughout the patient’s lifetime allowing longitudinal changes to be measured on follow up. ª 2016 British Association of Plastic, Reconstructive and Aesthetic Surgeons. Published by Elsevier Ltd. All rights reserved.
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on need for surgical revision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on linear anthropometric measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on CT vector analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on 3D vector analysis with point cloud representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes using surface analysis without landmark registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
00 00 00 00 00 00
* Corresponding author. 13123 East 16th Avenue, B467, Aurora, CO 80045, United States. Fax: þ1 720 478 7070. E-mail address: [email protected] (D.Y. Khechoyan). http://dx.doi.org/10.1016/j.bjps.2016.08.006 1748-6815/ª 2016 British Association of Plastic, Reconstructive and Aesthetic Surgeons. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
2
M.S. Lloyd et al. Recent Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Point set registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Financial disclosures/commercial associations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sources of support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Products/devices/drugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction The goal of this paper is to provide a historical review of the various methods used to provide a quantitative outcome analysis of cranial morphology in craniosynostosis. Novel techniques are then described showing the future direction of quantitative outcome analysis. Hankinson et al. set the scene e “Pediatric craniofacial surgeons have not, however, agreed upon objective means to assess postoperative cranial morphological improvement. We should therefore endeavor to agree upon objective craniometric tools for the assessment of operative outcomes, allowing us to accurately compare the various surgical techniques that are currently available”.1
Outcomes based on need for surgical revision The Whittaker categorization was the first system to be used by several studies to quantitatively look at outcome measures. Recognition was made of how “the excellence of correction is ultimately a subjective aesthetic judgment to which both the surgeon and the patient or family contribute”.2 An attempt was made to move away from subjective measurement. A total number of 164 patients were categorized into those with a symmetrical versus asymmetrical deformity. The authors subdivided the end points of correction into four categories based on whether the patients warranted further revision surgery. Category I included those patients in whom no refinements or surgical revisions were considered advisable or necessary by the surgeon or the patient. In category II, soft-tissue or lesser bone-contouring revisions were desirable whether performed or not. Category III included patients in whom major alternative osteotomies or bone grafting procedures were needed or performed for example further repositioning of the orbits to improve residual exorbitism or onlay bone grafts. Category IV included patients in whom a major craniofacial procedure was necessary essentially duplicating or exceeding in extent the original surgery.2 In a landmark post-operative review of patients with non-syndromic and syndromic synostosis, McCarthy et al. used the Whittaker classification to analyze results in 104 patients treated over a 20-year period.3 According to this study 87.5% were in categories IeII and 11.5% were in categories IIIeIV. The patients included in the study included those with diagnoses of bicoronal synostosis (10 cases),
00 00 00 00 00 00 00 00
unilateral coronal synostosis (57 cases), metopic synostosis (29 cases), and sagittal synostosis (8 cases). Seruya et al. in their study of 212 patients acknowledged the paucity of outcomes studies for the primary management of craniosynostosis over the most recent decade. Bi-coronal synostosis patients had a higher frequency of Whittaker class III/IV distribution, which was similar to the studies by Pearson and Sloan.4 It also highlighted the problem in comparing cranial form between different surgical techniques such as endoscopic to total cranial vault remodeling. Sloan et al. modified the Whittaker system into a seven-category outcome classification system to allow recognition of more subtle differences in surgical results. Further studies highlighted the difficulties in analysis of “cranial form” in craniosynostosis and the complexity of the deformities.5e8 The studies which used the Whittaker categorization were all retrospective with the data being derived from the notes but not from photos or by serial examination of the patient themselves. There were no pre-operative assessments made of cranial form, so comparison could not be made between before and after results. The success of the eventual outcome could also not be stratified by severity of pre-operative deformity.
Outcomes based on linear anthropometric measurements The next development in outcome analysis has come through the use of two dimensional (linear) direct anthropometric measurements, with the most common being the cranial-index. Fearon et al. analyzed 296 patients with single suture craniosynostosis retrospectively to assess long term growth using anthropometric measurements taken up to 11 years post-operatively with the mean follow up time of 4 years.9,10 This study raised the important question of how authors select and justify the specific points and indices used for measurements. Fearon et al. stated, “four specific measurements were chosen for this analysis, for they most accurately defined the dimensions of the cranial vault that are involved in the single sutural synostoses. These measurements were minimum frontal breadth (ft-ft), head circumference, maximum cranial length (g-op) and maximal cranial breadth (eu-eu). However, it is well documented that the selection of cephalometric landmarks incurs errors that authors may not account for in their final calculations.11 The major limitation of direct patient
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
Quantitative Outcome Analysis in Craniosynostosis anthropometry is the inter- and intra-observer error as well as logistical difficulty in the acquisition of the measurements in a young, uncooperative pediatric patient. However, this was one of the first studies to implement comparison of anthropometric outcomes in cases to published normal controls.9 Other authors have used linear anthropometric measurements for quantitative outcome analysis with varying results.12e14 Mendonca et al. demonstrated that caliper measurements were found to underestimate the anterioreposterior and bi-parietal distances as compared to measurements derived from computed tomography and three-dimensional stereophotogrammetry.15 The most obvious limitation of linear anthropometric measurements stems from the fact that two-dimensional measurements fail to accurately represent threedimensional morphology.16 Although simple measurements and ratios had been used to characterize specific aspects of skull shape in the past, cranial deformities are complex, three-dimensional structures. When specific indices are used individually, they overly simplify; and when used collectively, they are burdensome and difficult to interpret. For example, the cranial index is a simple cranial width-tolength ratio often used as a severity index in brachycephaly and scaphocephaly. It works well in these phenotypes for the following reasons: it describes a specific, obvious feature of dysmorphology; it is easy to acquire from CT or skull radiograph; there is a large normative and comparative data set for analysis. The cranial index does have limitations that are indicative of all anthropometry. It has no predefined origin, and it cannot provide qualitative regional assessment, especially since frontal and occipital bossing can both independently affect the cranial index. It also cannot indicate the presence of any regional variations other than the overall length or width of the skull.16
Outcomes based on CT vector analysis Mid-sagittal vector analysis (MSVA) is a method to evaluate dysmorphology in sagittal synostosis with quantitative and qualitative comparison before and after cranial vault remodeling.16 This involved patients undergoing pre-op and post-op CT one year after surgery. MSVA uses a radial vector analysis in which distances to the cranial surface are measured from a single reference point origin in the sagittal plane (sella turcica). The limitation of this technique was that it only assessed the dysmorphology in a single, 2D plane and also required the patient to be subjected to an additional CT scan and associated radiation exposure post-operatively. Transverse narrowing e a cardinal feature of scaphocephaly e as would be characterized by axial-plane measurements could not be assessed by this technique; also, the other limitation of the study was that there were no age-and gender-matched controls included for comparison. CT vector analysis was used to assess the changes in frontal morphology after single-stage open posteriormiddle vault expansion for sagittal craniosynostosis with comparison to age-matched normal controls.17 This study built on the technique described in the previous MSVA study by being able to collect measurements in both mid-sagittal
3 and axial plane to characterize frontal bossing in a large cohort of patients with sagittal synostosis. A point is made when comparing frontal breadth measurements to the Fearon study in that, “frontal breadth measurement does not measure the same anatomic point on repeated measures as the location of minimum frontal breadth can theoretically change with time. Since the study by Fearon et al. took the measurements at different times, the high variability in timing of measurement can make comparison of means difficult”. The limitation in the study by Khechoyan et al. was that the infant and toddler normal control groups were not paired and represented two distinct, disparate groups. Acquisition of normative, longitudinal CT data is difficult. The study did demonstrate that normal age-comparable control group databases could be built from normal patients undergoing CT imaging to rule out acute intracranial process following head trauma.
Outcomes based on 3D vector analysis with point cloud representation The technique of 3D vector analysis uses a program called 3DVA to generate cranial point cloud representations using imported CT scan DICOM data from cranial/craniofacial CT scans taken pre- and post-operatively of the patient.18 This process defined point clouds based on automatically generated vectors from the dorsum sella to the cranial surface of the CT scan. Each cloud consisted of more than 37,000 points and was orientated according to userselected locations of the dorsum sella, nasion and vertex. Cloud data for each skull was saved as an array of direction and magnitude information for the vectors defining each point. The authors recognized the importance of comparing pure shape differences between the patients’ skulls to the normative controls; they also normalized all point clouds for controls to account for overall size differences and subtle morphological variation between the skulls. This normalization process allowed the authors to focus on pure shape (and not size) analysis. A color histogram overlaid over the three-dimensional point cloud was utilized to represent deviation from the norm. A follow-up study by the same group applied the 3-D point cloud quantitative method to analyzing the cranial deformity encountered in patients with combined metopic and sagittal sutural fusion, establishing that 3D analysis is a valid method for evaluating cranial morphology.19
Outcomes using surface analysis without landmark registration CT vector analysis and 3D vector analysis studies have come closest to achieving comprehensive quantitative analysis of cranial morphology but still highlight three major problems. First, to measure the change in head shape all methods have relied on user-specified landmark selection. Second, the limitation of 3D vector analysis is that every surface point is referenced to an intra-cranial landmark. This approach ignores the relationship between two surface points and intervening variations in shape and curvature. Third, to monitor longitudinal post-operative changes in
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
4 cranial morphology, the patient is required to undergo additional (potentially un-indicated) CT imaging study, with the attendant risk of ionizing radiation exposure. Other important issues that need further investigation are that outcome analysis using the techniques of anthropometry has been done using either the bony skeleton or the surface soft tissue anatomy. Several studies have asked which of these represents cranial morphology best e the patient’s soft tissue surface anatomy or their bony anatomy? The advantage of 3D stereophotogrammetry and indirect anthropometry is that the quantitative, morphological evaluation could be deferred to another time point and the logistical difficulties of direct anthropometry in a pediatric patient are avoided. Landmark identification on the cranium alone was shown not always to be representative of the soft tissue envelope.20 Of practical importance, 3D stereophotogrammetric images can be an adequate patient surrogate allowing automatic high-density registration quickly without the need for landmark registration and also allowing quantitative analysis of complex facial traits.21 The quality of these images has also evolved over time allowing high-definition image capture quickly.22 With these features of stereophotogrammetry, outcome analysis of cranial morphology using surface analysis without landmark registration has been studied. Amm et al., in their study, acquired CT scans with threedimensional reconstruction on patients with sagittal synostosis at age 4e6 weeks, 1 year post-operatively, and yearly thereafter. The authors utilized three-dimensional surface analysis software, but primarily relied on the twodimensional cephalic index to quantitate outcomes.23 Non-invasive laser shape digitizer commonly known as a “STARscanner” has been described as a way of quantifying head shape in children with deformational plagiocephaly using four variables that the authors thought important in assessing head shape: the cranial vault asymmetry index, radial symmetry index, posterior symmetry ratio, and overall symmetry ratio. This data set was then used to design an orthotic molding helmet.24 However, when the utility of the STARscanner for evaluation of surgical outcomes in metopic synostosis patients was studied, the anterior symmetry ratio, posterior symmetry ratio, overall symmetry ratio, cranial vault volumes, cranial vault asymmetry index and cephalic ratio were found to be individually informative but were not amenable to composite integration to assess overall cranial form.25 Progress has been made in a 3D automated procedure for the characterization of deformational plagiocephaly that did not depend on landmark selection since this was thought to be “subjective, time-consuming and potentially unreliable”. Using 3D surface meshes of infant’s skulls, a comparison was computed between vectors that were perpendicular to a tangent plane at a particular point on a 3D surface mesh. This was based on the assumption that the surface normal vectors of 3D points lying on the flat surfaces would be more convergent, whereas surface normal vectors of 3D points that lie on rounded surfaces would be more divergent. The authors hypothesized that these measures would differentiate plagiocephaly from more normal head shapes.26
M.S. Lloyd et al.
Recent Techniques Point set registration Point set registration is a novel technique that may be the solution to the problems discussed above. The goal of point set registration is to assign correspondences between two sets of points and to recover the transformation that maps one point set to the other.27 The “points” in a point set are often features extracted from an image, such as the locations of corners, boundary points, or other important regions. The points can represent both geometric and intensity characteristics of an image. Practical point set registration algorithms should have desirable properties which the authors state as: 1) the ability to accurately model the transformation required to align the point sets with tractable computational complexity; 2) the ability to handle possibly high dimensionality of the point sets; and 3) robustness to degradations such as noise, outliers, and missing points that occur due to imperfect image acquisition and feature extraction. Point set registration can be rigid or non-rigid. In the rigid form it only allows for translation, rotation and scaling. The non-rigid transformation occurs in many real-world problems including shape recognition and medical image registration. Multidimensional point sets are common in many real world problems. The authors also state that degradations such as noise, outliers, and missing points significantly complicate the problem. Outliers are the points that are incorrectly extracted from the image; outliers have no correspondences in the other point set. Missing points are the features that are not found in the image due to occlusion or inaccurate feature extraction. A point set registration method should be robust to these degradations. Philips et al. have used this technique to create a 3D point cloud for normal skulls allowing the authors to manufacture a frontal bandeau template so that the deformed bandeau in the patient with craniosynostosis can be re-modelled more accurately.28
Future directions With imaging quality of 3D stereophotogrammetry and the use of non-landmark-based registration, it is now becoming possible to more accurately compare pre-operative to post operative images of children undergoing surgical correction for craniosynostosis. The pre-operative and post operative images can also be compared to age-, sex- and race-matched controls across the treatment period and at long-term followup. The 3D cranial form may then be represented both qualitatively and quantitatively with more fidelity utilizing techniques of differential geometry, such as 3D curvature computation and comparison.
Conclusion In summary, the measures of cranial morphology for craniosynostosis surgical outcomes have generally evolved from linear direct anthropometrics to indirect CT-based craniometric analysis to, more recently, 3D vector analysis
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
Quantitative Outcome Analysis in Craniosynostosis and surface analysis with non-landmark-based image registration. Future directions will likely involve utilization of automated image registration, with techniques such as point set registration, along with computational strategies that measure cranial form in 3D. Finally, comparison to agematched normal controls would be routinely done for infants and older children with craniosynostosis to stratify the pre-operative severity and evaluate surgical correction, respectively.
Financial disclosures/commercial associations None.
Sources of support None.
Products/devices/drugs None.
References 1. Hankinson TC, Fontana EJ, Anderson RC, et al. Surgical treatment of single-suture craniosynostosis: An argument for quantitative methods to evaluate cosmetic outcomes. J Neurosurg Pediatr 2010 Aug;6(2):193e7. 2. Whitaker LA, Bartlett SP, Schut L, et al. Craniosynostosis: An analysis of the timing, treatment, and complications in 164 consecutive patients. Plast Reconstr Surg 1987 Aug;80(2): 195e212. 3. McCarthy JG, Glasberg SB, Cutting CB, et al. Twenty-year experience with early surgery for craniosynostosis: II. The craniofacial synostosis syndromes and pansynostosis: Results and unsolved problems. Plast Reconstr Surg 1995 Aug;96(2): 284e95. discussion 296e8. 4. Seruya M, Oh AK, Boyajian MJ, et al. Long-term outcomes of primary craniofacial reconstruction for craniosynostosis: A 12year experience. Plast Reconstr Surg 2011 Jun;127(6):2397e406. 5. Pearson GD, Havlik RJ, Eppley B, et al. Craniosynostosis: A single institution’s outcome assessment from surgical reconstruction. J Craniofac Surg 2008 Jan;19(1):65e71. 6. Sloan GM, Wells KC, Raffel C, et al. Surgical treatment of craniosynostosis: Outcome analysis of 250 consecutive patients. Pediatrics 1997 Jul;100(1):E2. 7. Selber JC, Brooks C, Kurichi JE, et al. Long-term results following fronto-orbital reconstruction in nonsyndromic unicoronal synostosis. Plast Reconstr Surg 2008 May;121(5):251ee60e. 8. Bastidas N, Mackay DD, Taylor JA, et al. Analysis of the longterm outcomes of nonsyndromic bicoronal synostosis. Plast Reconstr Surg 2012 Oct;130(4):877e83. 9. Fearon JA, McLaughlin EB, Kolar JC. Sagittal craniosynostosis: Surgical outcomes and long-term growth. Plast Reconstr Surg 2006 Feb;117(2):532e41. 10. Fearon JA, Ruotolo RA, Kolar JC. Single sutural craniosynostoses: Surgical outcomes and long-term growth. Plast Reconstr Surg 2009 Feb;123(2):635e42.
5 11. Vincent AM, West VC. Cephalometric landmark identification error. Aust Orthod J 1987 Oct;10(2):98e104. 12. Hansen M, Padwa BL, Scott RM, et al. Synostotic frontal plagiocephaly: Anthropometric comparison of three techniques for surgical correction. Plast Reconstr Surg 1997 Nov;100(6): 1387e95. 13. David LR, Plikaitis CM, Couture D, et al. Outcome analysis of our first 75 spring-assisted surgeries for scaphocephaly. J Craniofac Surg 2010 Jan;21(1):3e9. 14. Skolnick GB, Naidoo SD, Patel KB, et al. Analysis of digital measures of cranial vault asymmetry for assessment of plagiocephaly. J Craniofac Surg 2014 Jul;25(4):1178e82. 15. Mendonca DA, Naidoo SD, Skolnick G, et al. Comparative study of cranial anthropometric measurement by traditional calipers to computed tomography and three-dimensional photogrammetry. J Craniofac Surg 2013 Jul;24(4):1106e10. 16. Marcus JR, Stokes TH, Mukundan S, et al. Quantitative and qualitative assessment of morphology in sagittal synostosis: Mid-sagittal vector analysis. J Craniofac Surg 2006 Jul;17(4): 680e6. 17. Khechoyan D, Schook C, Birgfeld CB, et al. Changes in frontal morphology after single-stage open posterior-middle vault expansion for sagittal craniosynostosis. Plast Reconstr Surg 2012 Feb;129(2):504e16. 18. Marcus JR, Domeshek LF, Loyd AM, et al. Use of a threedimensional, normative database of pediatric craniofacial morphology for modern anthropometric analysis. Plast Reconstr Surg 2009 Dec;124(6):2076e84. 19. Domeshek LF, Das RR, Van Aalst JA, Mukundan Jr S, Marcus JR. Influence of metopic suture fusion associated with sagittal synostosis. J Craniofac Surg 2011 Jan;22(1):77e83. 20. Dias GJ, Premachandra IM, Mahoney PM, et al. A new approach to improve TMJ morphological information from plain film radiographs. Cranio 2005 Jan;23(1):30e8. 21. Guo J, Mei X, Tang K. Automatic landmark annotation and dense correspondence registration for 3D human facial images. BMC Bioinform 2013 Jul 22;14:232. 22. Heike CL, Upson K, Stuhaug E, et al. 3D digital stereophotogrammetry: A practical guide to facial image acquisition. Head Face Med 2010 Jul 28;6:18. 23. Amm CA, Denny AD. Correction of sagittal synostosis using foreshortening and lateral expansion of the cranium activated by gravity: Surgical technique and postoperative evolution. Plast Reconstr Surg 2005 Sep;116(3):723e35. 24. Plank LH, Giavedoni B, Lombardo JR, et al. Comparison of infant head shape changes in deformational plagiocephaly following treatment with a cranial remolding orthosis using a noninvasive laser shape digitizer. J Craniofac Surg 2006 Nov; 17(6):1084e91. 25. Weathers WM, Khechoyan D, Wolfswinkel EM, et al. A novel quantitative method for evaluating surgical outcomes in craniosynostosis: Pilot analysis for metopic synostosis. Craniomaxillofac Trauma Reconstr 2014 Mar;7(1):1e8. 26. Atmosukarto I, Shapiro LG, Starr JR, et al. Three-dimensional head shape quantification for infants with and without deformational plagiocephaly. Cleft Palate Craniofac J 2010 Jul; 47(4):368e77. 27. Myronenko A, Song X. Point set registration: Coherent point drift. IEEE Trans Pattern Anal Mach Intell 2010 Dec;32(12): 2262e75. 28. Saber NR, Phillips J, Looi T, et al. Generation of normative pediatric skull models for use in cranial vault remodeling procedures. Childs Nerv Syst 2012 Mar;28(3):405e10.
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
MODEL
Journal of Plastic, Reconstructive & Aesthetic Surgery (2016) xx, 1e5
Review
Review of quantitative outcome analysis of cranial morphology in craniosynostosis Mark S. Lloyd, Edward P. Buchanan, David Y. Khechoyan* Division of Plastic and Reconstructive Surgery, University of Colorado, Department of Surgery, Aurora, CO, United States Received 18 February 2016; accepted 22 August 2016
KEYWORDS Craniosynostosis; Outcome measures; Point cloud representation; Stereophotogrammetry
Summary Outcome measures in craniosynostosis surgery have progressed from those based on the need for surgical revision to linear anthropometric measurements, 2D CT vector analysis and 3D CT vector analysis. However, finding an objective means to assess postoperative cranial morphological improvement remains challenging. A critical review of previous studies used to measure craniosynostosis surgery outcomes is presented. We also introduce and briefly discuss the key features of the computational algorithm that is being utilized in our center for evaluating craniosynostosis surgical outcomes. This has addressed a number of the previous challenges encountered in quantitative measurement of cranial morphological change. Point cloud representation and 3D stereophotogrammetry have made it possible to compare pre and post-operative images of children undergoing surgical correction for craniosynostosis. These pre- and post-operative images can also be compared to age, sex and race-matched controls throughout the patient’s lifetime allowing longitudinal changes to be measured on follow up. ª 2016 British Association of Plastic, Reconstructive and Aesthetic Surgeons. Published by Elsevier Ltd. All rights reserved.
Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on need for surgical revision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on linear anthropometric measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on CT vector analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes based on 3D vector analysis with point cloud representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outcomes using surface analysis without landmark registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
00 00 00 00 00 00
* Corresponding author. 13123 East 16th Avenue, B467, Aurora, CO 80045, United States. Fax: þ1 720 478 7070. E-mail address: [email protected] (D.Y. Khechoyan). http://dx.doi.org/10.1016/j.bjps.2016.08.006 1748-6815/ª 2016 British Association of Plastic, Reconstructive and Aesthetic Surgeons. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
2
M.S. Lloyd et al. Recent Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Point set registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Financial disclosures/commercial associations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sources of support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Products/devices/drugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction The goal of this paper is to provide a historical review of the various methods used to provide a quantitative outcome analysis of cranial morphology in craniosynostosis. Novel techniques are then described showing the future direction of quantitative outcome analysis. Hankinson et al. set the scene e “Pediatric craniofacial surgeons have not, however, agreed upon objective means to assess postoperative cranial morphological improvement. We should therefore endeavor to agree upon objective craniometric tools for the assessment of operative outcomes, allowing us to accurately compare the various surgical techniques that are currently available”.1
Outcomes based on need for surgical revision The Whittaker categorization was the first system to be used by several studies to quantitatively look at outcome measures. Recognition was made of how “the excellence of correction is ultimately a subjective aesthetic judgment to which both the surgeon and the patient or family contribute”.2 An attempt was made to move away from subjective measurement. A total number of 164 patients were categorized into those with a symmetrical versus asymmetrical deformity. The authors subdivided the end points of correction into four categories based on whether the patients warranted further revision surgery. Category I included those patients in whom no refinements or surgical revisions were considered advisable or necessary by the surgeon or the patient. In category II, soft-tissue or lesser bone-contouring revisions were desirable whether performed or not. Category III included patients in whom major alternative osteotomies or bone grafting procedures were needed or performed for example further repositioning of the orbits to improve residual exorbitism or onlay bone grafts. Category IV included patients in whom a major craniofacial procedure was necessary essentially duplicating or exceeding in extent the original surgery.2 In a landmark post-operative review of patients with non-syndromic and syndromic synostosis, McCarthy et al. used the Whittaker classification to analyze results in 104 patients treated over a 20-year period.3 According to this study 87.5% were in categories IeII and 11.5% were in categories IIIeIV. The patients included in the study included those with diagnoses of bicoronal synostosis (10 cases),
00 00 00 00 00 00 00 00
unilateral coronal synostosis (57 cases), metopic synostosis (29 cases), and sagittal synostosis (8 cases). Seruya et al. in their study of 212 patients acknowledged the paucity of outcomes studies for the primary management of craniosynostosis over the most recent decade. Bi-coronal synostosis patients had a higher frequency of Whittaker class III/IV distribution, which was similar to the studies by Pearson and Sloan.4 It also highlighted the problem in comparing cranial form between different surgical techniques such as endoscopic to total cranial vault remodeling. Sloan et al. modified the Whittaker system into a seven-category outcome classification system to allow recognition of more subtle differences in surgical results. Further studies highlighted the difficulties in analysis of “cranial form” in craniosynostosis and the complexity of the deformities.5e8 The studies which used the Whittaker categorization were all retrospective with the data being derived from the notes but not from photos or by serial examination of the patient themselves. There were no pre-operative assessments made of cranial form, so comparison could not be made between before and after results. The success of the eventual outcome could also not be stratified by severity of pre-operative deformity.
Outcomes based on linear anthropometric measurements The next development in outcome analysis has come through the use of two dimensional (linear) direct anthropometric measurements, with the most common being the cranial-index. Fearon et al. analyzed 296 patients with single suture craniosynostosis retrospectively to assess long term growth using anthropometric measurements taken up to 11 years post-operatively with the mean follow up time of 4 years.9,10 This study raised the important question of how authors select and justify the specific points and indices used for measurements. Fearon et al. stated, “four specific measurements were chosen for this analysis, for they most accurately defined the dimensions of the cranial vault that are involved in the single sutural synostoses. These measurements were minimum frontal breadth (ft-ft), head circumference, maximum cranial length (g-op) and maximal cranial breadth (eu-eu). However, it is well documented that the selection of cephalometric landmarks incurs errors that authors may not account for in their final calculations.11 The major limitation of direct patient
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
Quantitative Outcome Analysis in Craniosynostosis anthropometry is the inter- and intra-observer error as well as logistical difficulty in the acquisition of the measurements in a young, uncooperative pediatric patient. However, this was one of the first studies to implement comparison of anthropometric outcomes in cases to published normal controls.9 Other authors have used linear anthropometric measurements for quantitative outcome analysis with varying results.12e14 Mendonca et al. demonstrated that caliper measurements were found to underestimate the anterioreposterior and bi-parietal distances as compared to measurements derived from computed tomography and three-dimensional stereophotogrammetry.15 The most obvious limitation of linear anthropometric measurements stems from the fact that two-dimensional measurements fail to accurately represent threedimensional morphology.16 Although simple measurements and ratios had been used to characterize specific aspects of skull shape in the past, cranial deformities are complex, three-dimensional structures. When specific indices are used individually, they overly simplify; and when used collectively, they are burdensome and difficult to interpret. For example, the cranial index is a simple cranial width-tolength ratio often used as a severity index in brachycephaly and scaphocephaly. It works well in these phenotypes for the following reasons: it describes a specific, obvious feature of dysmorphology; it is easy to acquire from CT or skull radiograph; there is a large normative and comparative data set for analysis. The cranial index does have limitations that are indicative of all anthropometry. It has no predefined origin, and it cannot provide qualitative regional assessment, especially since frontal and occipital bossing can both independently affect the cranial index. It also cannot indicate the presence of any regional variations other than the overall length or width of the skull.16
Outcomes based on CT vector analysis Mid-sagittal vector analysis (MSVA) is a method to evaluate dysmorphology in sagittal synostosis with quantitative and qualitative comparison before and after cranial vault remodeling.16 This involved patients undergoing pre-op and post-op CT one year after surgery. MSVA uses a radial vector analysis in which distances to the cranial surface are measured from a single reference point origin in the sagittal plane (sella turcica). The limitation of this technique was that it only assessed the dysmorphology in a single, 2D plane and also required the patient to be subjected to an additional CT scan and associated radiation exposure post-operatively. Transverse narrowing e a cardinal feature of scaphocephaly e as would be characterized by axial-plane measurements could not be assessed by this technique; also, the other limitation of the study was that there were no age-and gender-matched controls included for comparison. CT vector analysis was used to assess the changes in frontal morphology after single-stage open posteriormiddle vault expansion for sagittal craniosynostosis with comparison to age-matched normal controls.17 This study built on the technique described in the previous MSVA study by being able to collect measurements in both mid-sagittal
3 and axial plane to characterize frontal bossing in a large cohort of patients with sagittal synostosis. A point is made when comparing frontal breadth measurements to the Fearon study in that, “frontal breadth measurement does not measure the same anatomic point on repeated measures as the location of minimum frontal breadth can theoretically change with time. Since the study by Fearon et al. took the measurements at different times, the high variability in timing of measurement can make comparison of means difficult”. The limitation in the study by Khechoyan et al. was that the infant and toddler normal control groups were not paired and represented two distinct, disparate groups. Acquisition of normative, longitudinal CT data is difficult. The study did demonstrate that normal age-comparable control group databases could be built from normal patients undergoing CT imaging to rule out acute intracranial process following head trauma.
Outcomes based on 3D vector analysis with point cloud representation The technique of 3D vector analysis uses a program called 3DVA to generate cranial point cloud representations using imported CT scan DICOM data from cranial/craniofacial CT scans taken pre- and post-operatively of the patient.18 This process defined point clouds based on automatically generated vectors from the dorsum sella to the cranial surface of the CT scan. Each cloud consisted of more than 37,000 points and was orientated according to userselected locations of the dorsum sella, nasion and vertex. Cloud data for each skull was saved as an array of direction and magnitude information for the vectors defining each point. The authors recognized the importance of comparing pure shape differences between the patients’ skulls to the normative controls; they also normalized all point clouds for controls to account for overall size differences and subtle morphological variation between the skulls. This normalization process allowed the authors to focus on pure shape (and not size) analysis. A color histogram overlaid over the three-dimensional point cloud was utilized to represent deviation from the norm. A follow-up study by the same group applied the 3-D point cloud quantitative method to analyzing the cranial deformity encountered in patients with combined metopic and sagittal sutural fusion, establishing that 3D analysis is a valid method for evaluating cranial morphology.19
Outcomes using surface analysis without landmark registration CT vector analysis and 3D vector analysis studies have come closest to achieving comprehensive quantitative analysis of cranial morphology but still highlight three major problems. First, to measure the change in head shape all methods have relied on user-specified landmark selection. Second, the limitation of 3D vector analysis is that every surface point is referenced to an intra-cranial landmark. This approach ignores the relationship between two surface points and intervening variations in shape and curvature. Third, to monitor longitudinal post-operative changes in
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
4 cranial morphology, the patient is required to undergo additional (potentially un-indicated) CT imaging study, with the attendant risk of ionizing radiation exposure. Other important issues that need further investigation are that outcome analysis using the techniques of anthropometry has been done using either the bony skeleton or the surface soft tissue anatomy. Several studies have asked which of these represents cranial morphology best e the patient’s soft tissue surface anatomy or their bony anatomy? The advantage of 3D stereophotogrammetry and indirect anthropometry is that the quantitative, morphological evaluation could be deferred to another time point and the logistical difficulties of direct anthropometry in a pediatric patient are avoided. Landmark identification on the cranium alone was shown not always to be representative of the soft tissue envelope.20 Of practical importance, 3D stereophotogrammetric images can be an adequate patient surrogate allowing automatic high-density registration quickly without the need for landmark registration and also allowing quantitative analysis of complex facial traits.21 The quality of these images has also evolved over time allowing high-definition image capture quickly.22 With these features of stereophotogrammetry, outcome analysis of cranial morphology using surface analysis without landmark registration has been studied. Amm et al., in their study, acquired CT scans with threedimensional reconstruction on patients with sagittal synostosis at age 4e6 weeks, 1 year post-operatively, and yearly thereafter. The authors utilized three-dimensional surface analysis software, but primarily relied on the twodimensional cephalic index to quantitate outcomes.23 Non-invasive laser shape digitizer commonly known as a “STARscanner” has been described as a way of quantifying head shape in children with deformational plagiocephaly using four variables that the authors thought important in assessing head shape: the cranial vault asymmetry index, radial symmetry index, posterior symmetry ratio, and overall symmetry ratio. This data set was then used to design an orthotic molding helmet.24 However, when the utility of the STARscanner for evaluation of surgical outcomes in metopic synostosis patients was studied, the anterior symmetry ratio, posterior symmetry ratio, overall symmetry ratio, cranial vault volumes, cranial vault asymmetry index and cephalic ratio were found to be individually informative but were not amenable to composite integration to assess overall cranial form.25 Progress has been made in a 3D automated procedure for the characterization of deformational plagiocephaly that did not depend on landmark selection since this was thought to be “subjective, time-consuming and potentially unreliable”. Using 3D surface meshes of infant’s skulls, a comparison was computed between vectors that were perpendicular to a tangent plane at a particular point on a 3D surface mesh. This was based on the assumption that the surface normal vectors of 3D points lying on the flat surfaces would be more convergent, whereas surface normal vectors of 3D points that lie on rounded surfaces would be more divergent. The authors hypothesized that these measures would differentiate plagiocephaly from more normal head shapes.26
M.S. Lloyd et al.
Recent Techniques Point set registration Point set registration is a novel technique that may be the solution to the problems discussed above. The goal of point set registration is to assign correspondences between two sets of points and to recover the transformation that maps one point set to the other.27 The “points” in a point set are often features extracted from an image, such as the locations of corners, boundary points, or other important regions. The points can represent both geometric and intensity characteristics of an image. Practical point set registration algorithms should have desirable properties which the authors state as: 1) the ability to accurately model the transformation required to align the point sets with tractable computational complexity; 2) the ability to handle possibly high dimensionality of the point sets; and 3) robustness to degradations such as noise, outliers, and missing points that occur due to imperfect image acquisition and feature extraction. Point set registration can be rigid or non-rigid. In the rigid form it only allows for translation, rotation and scaling. The non-rigid transformation occurs in many real-world problems including shape recognition and medical image registration. Multidimensional point sets are common in many real world problems. The authors also state that degradations such as noise, outliers, and missing points significantly complicate the problem. Outliers are the points that are incorrectly extracted from the image; outliers have no correspondences in the other point set. Missing points are the features that are not found in the image due to occlusion or inaccurate feature extraction. A point set registration method should be robust to these degradations. Philips et al. have used this technique to create a 3D point cloud for normal skulls allowing the authors to manufacture a frontal bandeau template so that the deformed bandeau in the patient with craniosynostosis can be re-modelled more accurately.28
Future directions With imaging quality of 3D stereophotogrammetry and the use of non-landmark-based registration, it is now becoming possible to more accurately compare pre-operative to post operative images of children undergoing surgical correction for craniosynostosis. The pre-operative and post operative images can also be compared to age-, sex- and race-matched controls across the treatment period and at long-term followup. The 3D cranial form may then be represented both qualitatively and quantitatively with more fidelity utilizing techniques of differential geometry, such as 3D curvature computation and comparison.
Conclusion In summary, the measures of cranial morphology for craniosynostosis surgical outcomes have generally evolved from linear direct anthropometrics to indirect CT-based craniometric analysis to, more recently, 3D vector analysis
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006
+
MODEL
Quantitative Outcome Analysis in Craniosynostosis and surface analysis with non-landmark-based image registration. Future directions will likely involve utilization of automated image registration, with techniques such as point set registration, along with computational strategies that measure cranial form in 3D. Finally, comparison to agematched normal controls would be routinely done for infants and older children with craniosynostosis to stratify the pre-operative severity and evaluate surgical correction, respectively.
Financial disclosures/commercial associations None.
Sources of support None.
Products/devices/drugs None.
References 1. Hankinson TC, Fontana EJ, Anderson RC, et al. Surgical treatment of single-suture craniosynostosis: An argument for quantitative methods to evaluate cosmetic outcomes. J Neurosurg Pediatr 2010 Aug;6(2):193e7. 2. Whitaker LA, Bartlett SP, Schut L, et al. Craniosynostosis: An analysis of the timing, treatment, and complications in 164 consecutive patients. Plast Reconstr Surg 1987 Aug;80(2): 195e212. 3. McCarthy JG, Glasberg SB, Cutting CB, et al. Twenty-year experience with early surgery for craniosynostosis: II. The craniofacial synostosis syndromes and pansynostosis: Results and unsolved problems. Plast Reconstr Surg 1995 Aug;96(2): 284e95. discussion 296e8. 4. Seruya M, Oh AK, Boyajian MJ, et al. Long-term outcomes of primary craniofacial reconstruction for craniosynostosis: A 12year experience. Plast Reconstr Surg 2011 Jun;127(6):2397e406. 5. Pearson GD, Havlik RJ, Eppley B, et al. Craniosynostosis: A single institution’s outcome assessment from surgical reconstruction. J Craniofac Surg 2008 Jan;19(1):65e71. 6. Sloan GM, Wells KC, Raffel C, et al. Surgical treatment of craniosynostosis: Outcome analysis of 250 consecutive patients. Pediatrics 1997 Jul;100(1):E2. 7. Selber JC, Brooks C, Kurichi JE, et al. Long-term results following fronto-orbital reconstruction in nonsyndromic unicoronal synostosis. Plast Reconstr Surg 2008 May;121(5):251ee60e. 8. Bastidas N, Mackay DD, Taylor JA, et al. Analysis of the longterm outcomes of nonsyndromic bicoronal synostosis. Plast Reconstr Surg 2012 Oct;130(4):877e83. 9. Fearon JA, McLaughlin EB, Kolar JC. Sagittal craniosynostosis: Surgical outcomes and long-term growth. Plast Reconstr Surg 2006 Feb;117(2):532e41. 10. Fearon JA, Ruotolo RA, Kolar JC. Single sutural craniosynostoses: Surgical outcomes and long-term growth. Plast Reconstr Surg 2009 Feb;123(2):635e42.
5 11. Vincent AM, West VC. Cephalometric landmark identification error. Aust Orthod J 1987 Oct;10(2):98e104. 12. Hansen M, Padwa BL, Scott RM, et al. Synostotic frontal plagiocephaly: Anthropometric comparison of three techniques for surgical correction. Plast Reconstr Surg 1997 Nov;100(6): 1387e95. 13. David LR, Plikaitis CM, Couture D, et al. Outcome analysis of our first 75 spring-assisted surgeries for scaphocephaly. J Craniofac Surg 2010 Jan;21(1):3e9. 14. Skolnick GB, Naidoo SD, Patel KB, et al. Analysis of digital measures of cranial vault asymmetry for assessment of plagiocephaly. J Craniofac Surg 2014 Jul;25(4):1178e82. 15. Mendonca DA, Naidoo SD, Skolnick G, et al. Comparative study of cranial anthropometric measurement by traditional calipers to computed tomography and three-dimensional photogrammetry. J Craniofac Surg 2013 Jul;24(4):1106e10. 16. Marcus JR, Stokes TH, Mukundan S, et al. Quantitative and qualitative assessment of morphology in sagittal synostosis: Mid-sagittal vector analysis. J Craniofac Surg 2006 Jul;17(4): 680e6. 17. Khechoyan D, Schook C, Birgfeld CB, et al. Changes in frontal morphology after single-stage open posterior-middle vault expansion for sagittal craniosynostosis. Plast Reconstr Surg 2012 Feb;129(2):504e16. 18. Marcus JR, Domeshek LF, Loyd AM, et al. Use of a threedimensional, normative database of pediatric craniofacial morphology for modern anthropometric analysis. Plast Reconstr Surg 2009 Dec;124(6):2076e84. 19. Domeshek LF, Das RR, Van Aalst JA, Mukundan Jr S, Marcus JR. Influence of metopic suture fusion associated with sagittal synostosis. J Craniofac Surg 2011 Jan;22(1):77e83. 20. Dias GJ, Premachandra IM, Mahoney PM, et al. A new approach to improve TMJ morphological information from plain film radiographs. Cranio 2005 Jan;23(1):30e8. 21. Guo J, Mei X, Tang K. Automatic landmark annotation and dense correspondence registration for 3D human facial images. BMC Bioinform 2013 Jul 22;14:232. 22. Heike CL, Upson K, Stuhaug E, et al. 3D digital stereophotogrammetry: A practical guide to facial image acquisition. Head Face Med 2010 Jul 28;6:18. 23. Amm CA, Denny AD. Correction of sagittal synostosis using foreshortening and lateral expansion of the cranium activated by gravity: Surgical technique and postoperative evolution. Plast Reconstr Surg 2005 Sep;116(3):723e35. 24. Plank LH, Giavedoni B, Lombardo JR, et al. Comparison of infant head shape changes in deformational plagiocephaly following treatment with a cranial remolding orthosis using a noninvasive laser shape digitizer. J Craniofac Surg 2006 Nov; 17(6):1084e91. 25. Weathers WM, Khechoyan D, Wolfswinkel EM, et al. A novel quantitative method for evaluating surgical outcomes in craniosynostosis: Pilot analysis for metopic synostosis. Craniomaxillofac Trauma Reconstr 2014 Mar;7(1):1e8. 26. Atmosukarto I, Shapiro LG, Starr JR, et al. Three-dimensional head shape quantification for infants with and without deformational plagiocephaly. Cleft Palate Craniofac J 2010 Jul; 47(4):368e77. 27. Myronenko A, Song X. Point set registration: Coherent point drift. IEEE Trans Pattern Anal Mach Intell 2010 Dec;32(12): 2262e75. 28. Saber NR, Phillips J, Looi T, et al. Generation of normative pediatric skull models for use in cranial vault remodeling procedures. Childs Nerv Syst 2012 Mar;28(3):405e10.
Please cite this article in press as: Lloyd MS, et al., Review of quantitative outcome analysis of cranial morphology in craniosynostosis, Journal of Plastic, Reconstructive & Aesthetic Surgery (2016), http://dx.doi.org/10.1016/j.bjps.2016.08.006