Accepted Manuscript Title: The use of craniofacial superimposition for disaster victim identification Author: Caroline Wilkinson Amy Lofthouse PII: DOI: Reference:
S0379-0738(15)00129-2 http://dx.doi.org/doi:10.1016/j.forsciint.2015.03.023 FSI 7948
To appear in:
FSI
Received date: Revised date: Accepted date:
24-6-2013 21-10-2014 24-3-2015
Please cite this article as: C. Wilkinson, A. Lofthouse, The use of craniofacial superimposition for disaster victim identification, Forensic Science International (2015), http://dx.doi.org/10.1016/j.forsciint.2015.03.023 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1
THE USE OF CRANIOFACIAL SUPERIMPOSITION FOR DISASTER VICTIM IDENTIFICATION
ip t
Comparison of manual and computerised craniofacial superimposition for DVI Twenty skulls compared to face pools
cr
Profile and three-quarter views were the most valuable facial images
Poor identification rate suggests MPUB database images would not be optimal
Ac ce p
te
d
M
an
us
Multiple ante-mortem images with a variety of facial expression should be utilised
Page 1 of 19
2
THE USE OF CRANIOFACIAL SUPERIMPOSITION FOR DISASTER VICTIM IDENTIFICATION Caroline Wilkinson1 and Amy Lofthouse1
cr
ip t
Centre for Anatomy & Human Identification MSI Building Dow Street University of Dundee UK DD1 5EH [email protected] [email protected]
Ac ce p
te
d
M
an
Corresponding author: Prof Caroline Wilkinson Centre for Anatomy & Human Identification MSI Building Dow Street University of Dundee UK DD1 5EH Tel: 01382386324 [email protected]
us
1
Page 2 of 19
3
THE USE OF CRANIOFACIAL SUPERIMPOSITION FOR DISASTER VICTIM IDENTIFICATION
ip t
ABSTRACT
Ac ce p
KEYWORDS
te
d
M
an
us
cr
Skull-to-face comparison is utilised for human identification where there is a suspected identity and the usual methods of identification, such as DNA or dental comparison, are not possible or practical. This research aimed to compare the reliability of manual and computerised craniofacial superimposition techniques and to establish the application of these techniques for disaster victim identification, where there may be a large database of passport-style images, such as the MPUB Interpol database. Twenty skulls (10 female; 10 male) were utilised from the William Bass Skeletal Collection at the University of Tennessee and compared to face pools of 20 face photographs of similar sex, age and ethnic group. A traditional manual photographic method and a new 3D computer-based method were used. The results suggested that profile and three-quarter views of the ante-mortem face were the most valuable for craniofacial superimposition. However, the poor identification rate achieved using images in frontal view suggests that the MPUB Interpol database would not be optimal for disaster victim identification, and passport-style images do not provide enough distinguishing facial detail. This suggests that multiple ante-mortem images with a variety of facial expression should be utilised for identification purposes. There was no significant difference in success between the manual and computer methods.
Craniofacial superimposition, identification, disaster victim
INTRODUCTION
Facial anthropology involves biological analysis and interpretation of the skull and/or the face and can be applied to identification of the living (biometrics, image analysis) and the dead (facial depiction, craniofacial superimposition, osteology, anatomy)[1]. Craniofacial analysis of the dead can be utilised for single unidentified human remains, multiple victims of disasters and mass graves. In the majority of forensic investigations there will be a suspect in relation to identity; in single forensic cases there may be a missing persons list, in mass disasters there may be a closed list (such as a passenger list for a transport system) or an open list (reported missing by families and/or employers) and for mass graves there may be whole missing populations. In these circumstances there may be ante-mortem data available relating to the suspects and this might include biological profiles (age, sex, stature, ethnic group), personal information (body modifications, identifying marks, clothing, jewellery, hair style etc.),
Page 3 of 19
4 photographs (ID cards, passport images, family albums or snapshots), hair samples (collected from a hairbrush etc.), dental records, clinical images and/or medical records [2].
ip t
However, much of this ante-mortem data may not be available, for example where the individual did not have a history of dental or medical treatment, or the available data may not be useful, for example where multiple members of the same family are missing (such as in a mass grave or mass disaster) and DNA analysis cannot separate family members. This may be significant in areas/countries with high levels of poverty, low socio-economic status and poor medical/dental practise.
us
cr
Where ante-mortem images are present, craniofacial superimposition may be effective as it does not require expensive or invasive techniques and is cost and time efficient. Craniofacial superimposition is the process where ante-mortem images are aligned and matched to the skull in order to assess the relationship between the hard and soft tissues of the face. This analysis may allow positive identification, especially where multiple ante-mortem images are available and this has been accepted in international courts as a method of identification [3-8].
an
Traditionally craniofacial superimposition has been carried out by forensic anthropologists or anatomists [5, 7, 9] and the techniques incorporate similar anatomical principles [10] and anthropological standards [11] as utilised in facial depiction. Numerous computerised systems have been developed for skull-to-face alignment [8, 12-15].
Ac ce p
te
d
M
The first documented use of craniofacial superimposition for identification in a medico-legal investigation was in 1935 [3], with a particular case involving a Lancastrian GP and the mysterious disappearance of his wife. Dr Ruxton claimed that his wife had left him for another man, but two weeks later two dismembered bodies were found in Glasgow. Police recovered two human heads and over seventy body parts wrapped in newspaper. The newspaper was from a special edition that was distributed only in the area where the Ruxtons lived. Dr Buck Ruxton had killed his wife and her maid and removed the eyes, noses, lips, skin and teeth to avoid identification of the bodies. The police suspected that the maid and the wife were the victims and a craniofacial superimposition was carried out using ante-mortem images of the two women and photographs of the skulls [16]. The method used in this case was photographic; employing enlargement, measurable objects, anatomical landmarks and craniophore orientation [17]. Known objects in the photographs (e.g. a tiara and a picket fence) were used to enlarge the faces to life size in order to identify Mary Rogers and Isabella Ruxton. There have been other significant forensic cases where craniofacial superimposition has been utilised for identification of human remains, including the identification of the remains of Josef Mengele [7, 18] and the identification of the victims of the serial killers, Fred and Rosemary West, in the UK [19]. Other significant cases are the Dobkin case [4], the Worlkersdorfer case [5] and the Howick Falls murder case [6]. Detailed reviews of the techniques can be found in the literature [17, 20-27] and many case studies have also been published [8, 13, 28-42]. Craniofacial superimposition development has passed through three technological phases: photographic, video and computer-assisted [17, 25]. The photographic technique was pioneered in the 1930s [3-5], the video technique was developed in the 1970s [43-45] and the computerassisted technique was introduced in the 1980s [12, 36, 46]. It is professionally agreed that craniofacial superimposition is of greater value for exclusion than positive identification [17] and that where single facial images are utilised, more than one person may appear consistent with the skull in question. Therefore, forensic practitioners must be well trained in anatomy and anthropology for the effective utilization of craniofacial superimposition and multiple ante-mortem images of the suspect should be analysed. When evaluating anatomical consistency special attention is paid to the cranial outline, the soft tissue thickness at various anthropometric landmarks on the skull and feature relationships between the skull and the face [47].
Page 4 of 19
5
us
cr
ip t
It is vitally important to practitioners and law enforcement agencies to establish the level of accuracy of these craniofacial techniques, especially in the case of craniofacial superimposition where this may be utilised in court for positive identification. The credibility of craniofacial superimposition was first established in a study using a database of 52 skulls from the Smithsonian collection [8]. An unknown skull was shown to match an ante-mortem facial photograph, while the four most similar skulls from the collection showed distinct differences to the facial photograph. Another blind study [48] used three skulls and compared them to 97 lateral view and 98 frontal view photographs of subjects (including the targets); in total 585 superimpositions. False matches were recorded for 9.6% of the lateral view and 8.5% of the frontal view superimpositions and all 3 targets were correctly matched. The incidence of false matches was reduced to 0.6% when both frontal and lateral view photographs of the same individual were used, suggesting that multiple photographs are optimal to prove or disprove identity by craniofacial superimposition. A further study [17] found that the outline from trichion to gonion in the lateral or oblique view was the optimal portion for personal identification.
M
an
Technological advances in the field of facial imaging over the past 20 years have produced new opportunities for research in facial morphology and facial growth assessment [49]. Surface scanning technologies such as laser scanners, photogrammetry cameras and other threedimensional systems (including three-dimensional digitizers and structured light systems) have advanced craniofacial landmark location [50-52] and facial anthropometry in both adults and juveniles [53-58]. These systems have benefited the field of orthodontic and plastic surgery with the ability to record subtle changes in surface morphology, especially important in the monitoring orthodontic treatment, facial growth and maxillofacial intervention [59, 60].
te
d
Numerous accuracy studies have been conducted on different laser scanners with regard to the accuracy from three-dimensional capture of faces of the resulting scan measurements when compared with anthropometric measurements from the face [52, 60-62]. These studies all found that laser scanners are a reliable and fast means of capturing a three-dimensional reproduction of a facial surface whilst maintaining high levels of accuracy with regard to the difference in resulting measurements between the scan and the subject. However, Aung et al, [52] did indicate that accurate location of landmarks and operator skill are important factors in achieving reliable results.
Ac ce p
The ability of a forensic practitioner to reliably cite the accuracy of a particular method employed in forensic human identification is highly desirable. The reason for this is not simply for legal reasons but also as a moral and ethical obligation to the victims and their families. This will ensure that if the practitioner is required to provide an expert witness report to the court (with regard to the identification they have assigned to the deceased) the evidence provided is valid with the support of scientific research. In order to improve craniofacial analysis techniques further rigorous quantitative accuracy studies need to be conducted especially with regard to the scientific method. Improving the accuracy of these techniques and increasing publication of the findings of accuracy studies will allow the use of craniofacial analysis techniques more widely in the field of forensic human identification, especially with regard to disaster victim identification and also increase the acceptance of these methods in a medico-legal investigation. The aim of this research was to compare the accuracy of manual and computerised craniofacial superimposition techniques and to establish the application of these techniques for disaster victim identification, where there may be a large database of passport-style images, such as the MPUB Interpol database.
Page 5 of 19
6
METHOD This study utilised a three-dimensional laser scanning system (FastSCAN™ Polhemus Scorpion™ handheld laser scanner) [63] for collection of 3D skull models for use in craniofacial analysis.
us
cr
ip t
Data collection was conducted on the William Bass donated skeletal collection from the University of Tennessee, USA, following the approval of the University of Tennessee Ethics Board. This collection consists of skeletal material and related ante-mortem images. Twenty (10 male and 10 female) crania and their associated mandibles were scanned using a glass platform and wooden mounting pole to enable 360 degree scanning ability. The crania were scanned separately from the associated mandibles to ensure as much detail was captured as possible. The mandibles were scanned using a glass platform that enabled scanning from all views. Before scanning of the crania and mandible commenced, photographic images were captured of the skeletal remains from numerous angles to display any features that the laser scanner may not be able to acquire. The related ante-mortem images were collected and employed as the target images.
an
Three male face pools and three female face pools were created using the target images in addition to a large number of other ante-mortem images collected from unrelated cases in the William Bass Collection. This enabled the creation of the face pools with consistent demographic profiles. The following face pools were created: Female Face Pool 1 -FFP1 (20 faces – 3 targets, 17 foils)
Female Face Pool 2 - FFP2 (20 faces – 3 targets, 17 foils)
M
Female Face Pool 3 - FFP3 (20 faces – 4 targets, 16 foils) Male Face Pool 1 - MFP1 (20 faces – 3 targets, 17 foils)
d
Male Face Pool 2 - MFP2 (20 faces – 3 targets, 17 foils) Male Face Pool 3 - MFP3 (20 faces – 4 targets, 16 foils)
Ac ce p
te
Within each face pool there were a variety of images, ranging from candid shots to professional portraits. The image quality ranged from poor quality (i.e. photocopies of original documents and low resolution photographic shots – small pixel count) to high quality (i.e. higher resolution photographic shots – large pixel count) and the angle of the faces in the photographs varied from frontal to profile images. The age of the images also varied (with some from the 1960s) although the age of the subjects in the images was similar. The individuals in the face pools also displayed a variety of facial hair and clothing, some of which caused areas of the face to be distorted or obscured. This variation in the face pools represented a typical DVI scenario where photographs would originate from a variety of sources. The twenty skulls were equally divided between the two methods with 10 skulls per method (with 5 females and 5 males in each group). The craniofacial superimposition guidelines used to determine a successful or unsuccessful CFS were based upon the morphology of the skull and the closeness of the match between the skull and the ante-mortem (AM) image (no measurements were taken). The researcher created a set of 3 stages to aid with determination of a match (for both manual and computerised), these were as follows: Stage 1: Individuals were eliminated if the skull did not fit within the facial outline of the AM image. Stage 2: Individuals were eliminated if skeletal features did not match in position or morphology to the AM image. Stage 3: Additional AM images were requested for further analysis and elimination. Not all individuals had additional photographs to reflect a real life DVI scenario. The remaining
Page 6 of 19
7 individuals were then classified as a match or no match to the skull based on all available material. The two methods of craniofacial superimposition (CFS) included: 1. The Manual method was an updated version of a pre-existing method described by Seta and Yoshino [64] for craniofacial analysis that utilises a remote guided motor driven machine to orientate the skull for image capture.
an
us
cr
ip t
The manual method required physical skulls for use with the skull orientation machine. However the original skulls were not available and replicas of the 10 skulls were created by stereolithography (rapid prototyping) [65]. The mandibles were then temporarily fixed to the crania with modelling putty (based upon guidelines produced in [66]). The skull orientation machine was set up in front of a high resolution digital web camera which recorded the movement of the skull model and transmitted the live feed to a PC (see Figure 1). Current literature states that the optimum objective length for photographic superimposition is at 1 metre [23, 24]. Craniofacial superimposition can be conducted with the aid of free online software called Glass 2K (Chime Software Ltd) [67] which enables the user to change the opacity of computer windows. The user can set up the web camera feed and then overlay an image and change the opacity of the image file window to enable the web camera feed to be visualised through the image (see Figure 1). Two hundred manual craniofacial superimpositions were conducted utilising this method for the ten skull models with each skull model undergoing a separate craniofacial superimposition for each face in the face pool.
M
2. The Computer Method was a newly developed and utilised a combination of specialist computer hardware (haptic device) and 3D modelling software to orientate a digital skull model in virtual space for image capture.
Ac ce p
te
d
The computer method also utilised the same procedure, except that it did not require a physical skull model as the entire process was conducted in a virtual space. The laser scan models were used (in an .obj format) and were imported into a 3D modelling software package (Geomagic’s Freeform Modelling®) [68] where the mandible was fixed to the cranium in a relaxed resting position (in the same position as for manual). This modelling package allows a 3D model to be orientated in a virtual space when used in combination with a piece of computer hardware called a Phantom® Desktop™ Haptic Device [69], which enables the user to rotate and orientate a 3D model using an arm-like tool. The computer method allows the user to position a 3D skull model file in virtual space (see Figure 2) and then position an AM image (on a virtual plane) anterior to the skull model. The opacity of the ante-mortem image can be altered to allow better visibility of the 3D skull model. The user can then use the haptic device to orientate the skull model to match in scale and position to the face in the AM image (See Figure 2). A screen shot can then be captured and imported into Adobe Photoshop © CS3 [70] for final analysis and demonstration. Two hundred computer craniofacial superimpositions were conducted utilising this method for the ten skull models with each skull model undergoing a separate craniofacial superimposition for each face in the face pool.
RESULTS
The CFS methods correctly matched the target to the skull at stage 3 in 40% manually and 50% by computer. It was observed that the overall correct match rate at stage 3 was 45% (see Tables 1 & 2). The false match rate at stage 3 was 15% manually and 8% by computer, giving an overall false match rate of 12%. The manual method matched 17% and the computer method matched 11% of the total faces to the skull at stage 3. Ten (50%) of the targets were falsely rejected at stage 1 and a further one (5%) falsely rejected at stage 2. The manual method rejected more targets (60%) at stage 1 than the computer
Page 7 of 19
8 method (40%) and the computer method rejected more targets (10%) at stage 2 than the manual method (0%) (see Tables 1 & 2).
cr
ip t
Thirty-six per cent of foil images were of low, 47% medium and 18% high quality. Fifty per cent were colour and 50% were black & white images, with 68% in frontal, 19% three-quarter and 14% profile views. The majority (76%) of images displayed a smiling face with 40% open mouth smile and 34% closed mouth smile. Twenty-three per cent of images displayed an individual not smiling with a closed mouth and relatively few individuals displayed no smile and an open mouth (5%). The majority of foil images displayed no facial distractions (eg. glasses, hats, facial hair) of any kind (57%), however 18% of individuals were observed to wear spectacles, 18% of individuals showed facial hair and relatively few individuals displayed a mixture of spectacles, facial hair and clothing. Visible photographic distortion was recorded in a relatively small number of images (11%).
an
us
Forty per cent of target images were of low, 30% medium and 30% high quality. Sixty per cent were colour and 40% black & white images, with 50% in frontal, 25% three-quarter and 25% profile views. The majority (70%) displayed a smiling target with 45% of images displaying an open mouth smile and 25% displaying a closed mouth smile. Twenty-five per cent of images displayed the target not smiling with a closed mouth and only 5% displayed the target not smiling with an open mouth. The majority of targets displayed no facial distractions of any kind (80%), however 5% were observed to wear spectacles, 10% wore facial hair and 5% wore both spectacles and facial hair. Visible photographic distortion was only recorded in 5% of target images.
d
M
Most of the targets (67%) that reached stage 3 had non-frontal view images, whereas most of the targets (60%) that were eliminated at stage 1 had frontal view images (see Figure 3). This would suggest that CFS was more successful where the target was seen in profile or ¾ views. These views enable observation of the profile of the face and the jaw line, which can result in optimal assessment due to the relatively thin soft tissue thicknesses over the gonial, nasal and orbital areas.
te
Eighty-nine per cent of the targets that reached stage 3 did not have any facial distractions, and 30% of the targets eliminated at stage 1 displayed facial distractions (see Figure 4). This would suggest that facial distractions had a small effect on CFS success.
Ac ce p
Forty-four per cent of the targets that reached stage 3 displayed an open mouth expression and 60% of the targets that were eliminated at stage 1 displayed an open mouth expression (see Figure 5). This would suggest that CFS was less successful when the AM image depicted an open mouth expression. There was no observable difference between colour and black and white target images in relation to successful match rates, and there was no significant difference in match rates in relation to visible photographic distortion or hair thickness.
DISCUSSION
These results are similar to previous accuracy tests. Austin-Smith and Maples [48] recorded a 9.5% false match rate, which is similar to the 12% false match rate in this study. Yoshino and colleagues (47) also found that the profile and three-quarter views were the most valuable for craniofacial superimposition. However, the poor identification rate achieved using images in frontal view suggests that the MPUB Interpol database would not be optimal for disaster victim identification, and passport images do not provide enough distinguishing facial detail. Gordon and Steyn [71] carried out a craniofacial superimposition study using forty frontal face images of cadavers, ten skulls for comparison with each face (400 comparisons in total) and morphological and landmark superimposition techniques. They also found that frontal images have limited use for identification and recommended corroborative evidence.
Page 8 of 19
9 This study suggests that to improve the chance of a successful CFS in a DVI scenario the following standardised protocols should be followed: 1. The ante-mortem images should be of a high quality and without facial distractions or visible distortion. 2. Ideally at least two images should be utilised; neutral facial expression with a closed mouth and slightly open mouth showing the upper teeth.
ip t
3. Multiple views including profile and ¾ views are ideal. Frontal views alone are not optimal.
cr
There was no significant difference in success between the manual and computer methods, and this suggests that digital technology is as reliable as traditional methods. However, neither method in this study utilised the physical remains, and the detail provided by the laser scans was consistent with the stereolithographic replicas.
an
us
This study concludes that craniofacial superimposition could be a useful tool in disaster victim identification for narrowing the possible identifications of individuals in a large scale closed disaster (such as a commercial aeroplane crash where the manifest of passengers is known) as long as the suggested guidelines are followed. When used in combination with other disaster victim identification techniques, such as DNA and dental analysis, craniofacial superimposition can result in a fairly reliable and accurate outcome for the identification of individuals.
1. 2. 3. 4. 5. 6. 7. 8. 9.
Ac ce p
REFERENCES
te
d
M
Following the success of the computer method, further studies should focus on decreasing the amount of expensive equipment and computer software and the development of a bespoke craniofacial superimposition software system. The researcher noted that although the haptic device was a useful tool for the orientation of the skull, it was the least portable aspect of the CFS equipment and also the most expensive piece of equipment used. Any bespoke computer software would require an ability to position three-dimensional models in virtual space in combination with a two-dimensional plane to support an AM image. This software should also include a sliding scale option for perspective views enabling the user to adjust the scale in response to any photographic distortion observed within the AM image.
Lee, W.J., Mackenzie, S. and Wilkinson, C.M. (2010) Facial Identification of the Dead. In: Forensic Anthropology 2000-2010, Ferguson, E. and Black, S.M. (Eds) CRC Press. Black, S. and Hackman, L. (2010) Disaster Victim Identification. Wiley Online Library. Glaister, J. and Brash, J.C. (1937) Medico-legal aspects of the Ruxton case. Vol. 33. Livingstone. Simpson, K.. (1943) Rex vs. Dobkin: the Baptist Church Cellar Murder'. Medico-Legal Review, II, p. 132-45. Gordon, I. and Drennan, M. (1948) Medico-legal aspects of the Wolkersdorfer case. South African Medical Journal, 22: p. 543-549. Prinsloo, I. (1953) The identification of skeletal remains in Regina versus K and another: The Howick Falls murder case. J. Forensic Med, 1: p. 11-7. Helmer, R.P. (1987) Identification of the cadaver remains of Josef Mengele. Journal of Forensic Sciences, 32(6): p. 1622. Ubelaker, D.H., Bubniak, E. and O'Donnell, G. (1992) Computer-assisted photographic superimposition. Journal of Forensic Sciences, 37(3): p. 750-762. Maat, G. (1989) The positioning and magnification of faces and skulls for photographic superimposition. Forensic Science International, 41(3): p. 225-235.
Page 9 of 19
10
16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31.
ip t
cr
us
15.
an
14.
M
13.
d
12.
te
11.
Gerasimov, M. (1955) The reconstruction of the face from the basic structure of the skull. Moscow: Nauka. Wilkinson, C., (2010) Facial reconstruction–anatomical art or artistic anatomy? Journal of Anatomy, 216(2): p. 235-250. Delfino, P. V., Colonna, M., Vacca, E., Potente, F., & Introna Jr, F. (1986). Computer-aided skull/face superimposition. The American Journal of Forensic Medicine and Pathology, 7(3), 201 Shahrom, A. W., Vanezis, P., Chapman, R. C., Gonzales, A., Blenkinsop, C., & Rossi, M. L. (1996). Techniques in facial identification: computer-aided facial reconstruction using a laser scanner and video superimposition. International Journal of Legal Medicine, 108(4), 194-200. Al-Amad, S., McCullough, M., Graham, J., Clement, J., & Hill, A. (2006) Craniofacial identification by computer-mediated superimposition. The Journal of Forensic Odontostomatology, 24(2): p. 47. Ibáñez, O., Cordón, O., Damas, S., & Santamaría, J. (2008). Craniofacial superimposition based on genetic algorithms and fuzzy location of cephalometric landmarks. In: Hybrid artificial intelligence systems (pp. 599-607). Springer Berlin Heidelberg. Glaister, J. and Brash, J.C. (1953) The Ruxton Case. Medical Jurisprudence and Toxicology, 1953: p. 90-101. Yoshino, M. (2012) Craniofacial superimposition. In: Wilkinson, C.M. and Rynn, C. (Eds) Craniofacial Identification. Cambridge University Press; p. 238. Teixeira, W.R.G., (1985) The Mengele report. The American Journal of Forensic Medicine and Pathology, 6(4): p. 279-283. Whittaker, D., Richards, B. and M. Jones, M. (1998) Orthodontic reconstruction in a victim of murder. Journal of Orthodontics, 25(1): p. 11-14. Brown, K. (1983) Developments in cranio-facial superimposition for identification. The Journal of Forensic Odonto-stomatology, 1(2): p. 57. Brocklebank, L. and Holmgren, C. (1989) Development of equipment for the standardization of skull photographs in personal identifications by photographic superimposition. Journal of Forensic Sciences, 34(5): p. 1214. Lan, Y. (1992) Development and current status of skull-image superimpositionMethodology and instrumentation. Forensic Science Review, 4: p. 125-125. Sekharan, P.C. (1993) Positioning the skull for superimposition. In: Iscan, M.Y. and Helmer, R.P. (Eds) Forensic Analysis of the Skull. Wiley-Liss, New York. p. 105-118. Cai, D. and Y. Lan, Y. (1993) Standards for skull-to-photo superimposition. In: Iscan, M.Y. and Helmer, R.P. (Eds) From the Bare Bone to the Full Face. Wiley-Liss, New York, p. 171198. Aulsebrook, W. A., İşcan, M. Y., Slabbert, J. H., & Becker, P. (1995). Superimposition and reconstruction in forensic facial identification: a survey. Forensic Science international, 75(2), 101-120. Taylor, J. and Brown, K. (1998) Superimposition techniques. In: Craniofacial identification in Forensic Medicine. London: Hodder Arnold, p. 151-64. Glassman, D. Methods of superimposition. In: Taylor, K.T. (2001) Forensic Art and Illustration. CRC Press. p. 477-498. Sen, N. (1962) Identification by superimposed photographs. International Criminal Police Review, 162: p. 284-286. Basauri, C. (1967) A body identified by forensic odontology and superimposed photographs. International Criminal Police Review, 204: p. 37-43. Sekharan, P.C. (1971) Revised Superimposition Technique for Identification of the Individual from the Skull and Photograph, A. J. Crim. L. Criminology & Police Sci., 62: p. 107. Reddy, K. (1973) Identification of dismembered parts: the medicolegal aspects of the Nagaraju case. Forensic science, 2: p. 351-374.
Ac ce p
10.
Page 10 of 19
11
39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52.
ip t
cr
38.
us
37.
an
36.
M
35.
d
34.
te
33.
Gejvall, N. (1974) Superimposition plus SEM-comparison of hair cuticle for identification purpose. International J. Skeletal Research, 1(974): p. 99-103. Sivaram, S. and Wadhera, C. (1975) Identity from skeleton—A case study. Forensic Science, 5(2): p. 166. Brown, K.A. (1982) The identification of Linda Agostini: The significance of dental evidence in the Albury Pyjama Girl case. The American Journal of Forensic Medicine and Pathology, 3(2): p. 131-142. McKenna, J., N.G. Jablonski, and R. Fearnhead, A method of matching skulls with photographic portraits using landmarks and measurements of the dentition. J Forensic Sci, 1984. 29(3): p. 787-97. Bastiaan, R.J., Dalitz, G.D. and Woodward, C. (1986) Video superimposition of skulls and photographic portraits—a new aid to identification. J Forensic Sci, 31(4): p. 1373-9. Webster, W. P., Murray, W. K., Brinkhous, W., & Hudson, P. (1986). Identification of human remains using photographic reconstruction. In: Forensic Osteology; advances in the identification of human remains. Springfield: Charles C Thomas, 256-89. Yoshino, M. and Seta, S. (1989) Personal identification of the human skull: superimposition and radiographic techniques. Forensic Sci. Rev, 1: p. 23-42. Solla, H.E. and Işcan, M.Y. (2001) Skeletal remains of Dr. Eugenio Antonio Berrı́os Sagredo. Forensic Science International, 2001. 116(2): p. 201-211. Bilge, Y., Kedici, P. S., Alakoç, Y. D., Ülküer, K. Ü., & İlkyaz, Y. Y. (2003). The identification of a dismembered human body: a multidisciplinary approach. Forensic Science International, 137(2), 141-146. Ghosh, A. and Sinha, P. (2005) An unusual case of cranial image recognition. Forensic Science International, 148(2): p. 93-100. Fenton, T.W., Heard, A.N. and Sauer, N.J. (2008) Skull‐Photo Superimposition and Border Deaths: Identification Through Exclusion and the Failure to Exclude. Journal of Forensic Sciences, 2008. 53(1): p. 34-40. Helmer, R.P. and Grüner, O. (1977) Vereinfachte Schädelidentifizierung nach dem Superprojektionsverfahren mit Hilfe einer Video-Anlage. International Journal of Legal Medicine, 80(3): p. 183-187. Iten, P. (1987) Identification of skulls by video superimposition. J Forensic Sci, 32(1): p. 173-88. Koelmeyer, T.D. (1982) Videocamera superimposition and facial reconstruction as an aid to identification. Am J Forensic Med Pathol, 3(1): p. 45-48. Austin, D. (1999) Video superimposition at the CA Pound Laboratory 1987 to 1992. J Forensic Sci, 44(4): p. 695-9. Yoshino, M., Imaizumi, K., Miyasaka, S., & Seta, S. (1995). Evaluation of anatomical consistency in craniofacial superimposition images. Forensic Science International, 74(1), 125-134. Austin-Smith, D. and Maples, W.R. (1994) The reliability of skull/photograph superimposition in individual identification. Journal of Forensic Sciences, 39(2): p. 446. De Greef, S. and Willems, G. (2005) Three-dimensional cranio-facial reconstruction in forensic identification: latest progress and new tendencies in the 21st century. Journal of Forensic Sciences, 50(1): p. 12. Gwilliam, J.R., Cunningham, S.J.and Hutton, T. (2006) Reproducibility of soft tissue landmarks on three-dimensional facial scans. The European Journal of Orthodontics, 28(5): p. 408-415. Bush, K. and Antonyshyn, O. (1996) Three-dimensional facial anthropometry using a laser surface scanner: validation of the technique. Plastic and Reconstructive Surgery; 98(2): p. 226-235. Aung, S., Ngim, R. and Lee, S. (1995) Evaluation of the laser scanner as a surface measuring tool and its accuracy compared with direct facial anthropometric measurements. British Journal of Plastic Surgery. 48(8): p. 551-558.
Ac ce p
32.
Page 11 of 19
12
59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70.
ip t
cr
us
58.
an
57.
M
56.
d
55.
te
54.
Moss, J. P., Linney, A. D., Grindrod, S. R., Arridge, S. R., & Clifton, J. S. (1987). Threedimensional visualization of the face and skull using computerized tomography and laser scanning techniques. The European Journal of Orthodontics, 9(4), 247-253. McDonagh, S., Moss, J. P., Goodwin, P., & Lee, R. T. (2001). A prospective optical surface scanning and cephalometric assessment of the effect of functional appliances on the soft tissues. The European Journal of Orthodontics, 23(2), 115-126. Tuncay, O.C. (2001) Three-dimensional imaging andmotion animation. In: Seminars in Orthodontics. Elsevier. Kau, C. H., Richmond, S., Savio, C., & Mallorie, C. (2006). Measuring adult facial morphology in three dimensions. The Angle Orthodontist, 76(5), 773-778. Kau, C. H., Zhurov, A., Richmond, S., Cronin, A., Savio, C., & Mallorie, C. (2006). Facial templates: a new perspective in three dimensions. Orthodontics & Craniofacial Research, 9(1), 10-17. Kau, C. H., Cronin, A., Durning, P., Zhurov, A. I., Sandham, A., & Richmond, S. (2006). A new method for the 3D measurement of postoperative swelling following orthognathic surgery. Orthodontics & Craniofacial Research, 9(1), 31-37. Ismail, S. and J. Moss, (2002) The three-dimensional effects of orthodontic treatment on the facial soft tissues–a preliminary study. British Dental Journal; 192(2): p. 104-108. Kusnoto, B. and C.A. Evans, (2002) Reliability of a 3D surface laser scanner for orthodontic applications. American Journal of Orthodontics and Dentofacial Orthopedics, 122(4): p. 342. Kau, C. H., Zhurov, A., Bibb, R., Hunter, L., & Richmond, S. (2005). The investigation of the changing facial appearance of identical twins employing a three‐dimensional laser imaging system. Orthodontics & Craniofacial Research, 8(2), 85-90. Park, S. H., Yu, H. S., Kim, K. D., Lee, K. J., and Baik, H. S. (2006). A proposal for a new analysis of craniofacial morphology by 3-dimensional computed tomography. American Journal of Orthodontics and Dentofacial Orthopedics, 129(5), p. 600. Polhemus (2007). FastSCAN™ User Manual. Vermont, U.S.A. http://www.polhemus.com/polhemus_editor/assets/FastSCAN%20COBRA%20C1%20 &%20SCORPION%20Brochure.pdf Accessed last at 22.11 on 02/02/2014 Seta, S. and Yoshino, M. (1993) A combined apparatus for photographic and video superimposition. In: Iscan, M.Y. and Helmer, R.P. (Eds) From the bare bone to the full face. New York: Wiley-Liss; p. 161-9. http://pdronline.info/ Accessed last at 22.10 on 02/02/2014 Taylor, K.T. (2001) Forensic Art and Illustration. CRC Press
Ac ce p
53.
http://chime.tv/products/glass2k.shtml Accessed last at 22.10 on 02/02/2014
Geomagic Freeform Plus ® http://geomagic.com/en/products/freeform/overview/ Accessed last at 22.14 on 02/02/2014 Geomagic Phantom ® http://geomagic.com/en/products/phantom-premium/overview Accessed last at 22.15 on 02/02/2014 Adobe Photoshop © CS8.0 (Adobe) Systems, San Jose, CA; http://www.adobe.com/ Accessed last at 22.15 on 02/02/2014
71. Gordon, G. M., and Steyn, M. (2012). An investigation into the accuracy and reliability of skull-photo superimposition in a South African sample. Forensic Science International, 216(1), 198-e1.
Page 12 of 19
13
ACKNOWLEDGEMENTS
Ac ce p
te
d
M
an
us
cr
ip t
The authors would like to thank the University of Tennessee for the use of the William Bass Skeletal Collection and the University of Dundee for funding this research through the CAHID Greenhouse PhD Scholarship Fund.
Page 13 of 19
14
Figure 1: The manual craniofacial superimposition set up with webcam, lap top and skull orientation machine
ip t
Figure 2: An example of craniofacial superimposition (left) using the computer method (right)
us
cr
Table 1: Results of the manual craniofacial superimposition test (F = female; M = male)
an
Table 2: Results of the computer-based craniofacial superimposition test (F = female; M = male)
M
Figure 3: Survival graph in relation to the angle of view for the target images
d
Figure 4: Survival graph in relation to the facial distractions for the target images
Ac ce p
te
Figure 5: Survival graph in relation to the expression for the target images
Page 14 of 19
15
N = 20
Skull F4
F6
F8
F10
M2
M4
M6
M8
M10
Total
Number of matches
2
1
3
3
5
4
4
3
4
4
17%
False matches
2
0
3
2
5
3
4
3
4
3
15%
Correct match
x
yes
x
yes
x
yes
x
x
x
yes
40%
Stage target eliminated
S1
x
S1
x
S1
x
S1
cr
ip t
F2
S1
x
us
S1
N = 20
M
an
Table 1: Results of the manual craniofacial superimposition test (F = female; M = male)
Skull
F5
Number of matches
1
2
1
False matches
0
1
Correct match
yes
yes
Stage target eliminated
x
x
F7
F9
M1
M3
M5
M7
M9
Total
1
4
1
2
2
4
11%
d
F3
3
te
F1
3
1
3
1
2
2
3
8%
yes
x
x
yes
x
x
x
yes
50%
x
S1
S1
x
S1
S2
S1
x
Ac ce p
0
Table 2: Results of the computer-based craniofacial superimposition test (F = female; M = male)
Page 15 of 19
us
cr
ip t
16
Ac ce p
te
d
M
an
Figure 3: Survival graph in relation to the angle of view for the target images
Figure 4: Survival graph in relation to the facial distractions for the target images
Page 16 of 19
us
cr
ip t
17
Ac ce p
te
d
M
an
Figure 5: Survival graph in relation to the expression for the target images
Page 17 of 19
Ac
ce
pt
ed
M
an
us
cr
i
Figure
Page 18 of 19
Ac
ce
pt
ed
M
an
us
cr
i
Figure
Page 19 of 19
S0379-0738(15)00129-2 http://dx.doi.org/doi:10.1016/j.forsciint.2015.03.023 FSI 7948
To appear in:
FSI
Received date: Revised date: Accepted date:
24-6-2013 21-10-2014 24-3-2015
Please cite this article as: C. Wilkinson, A. Lofthouse, The use of craniofacial superimposition for disaster victim identification, Forensic Science International (2015), http://dx.doi.org/10.1016/j.forsciint.2015.03.023 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1
THE USE OF CRANIOFACIAL SUPERIMPOSITION FOR DISASTER VICTIM IDENTIFICATION
ip t
Comparison of manual and computerised craniofacial superimposition for DVI Twenty skulls compared to face pools
cr
Profile and three-quarter views were the most valuable facial images
Poor identification rate suggests MPUB database images would not be optimal
Ac ce p
te
d
M
an
us
Multiple ante-mortem images with a variety of facial expression should be utilised
Page 1 of 19
2
THE USE OF CRANIOFACIAL SUPERIMPOSITION FOR DISASTER VICTIM IDENTIFICATION Caroline Wilkinson1 and Amy Lofthouse1
cr
ip t
Centre for Anatomy & Human Identification MSI Building Dow Street University of Dundee UK DD1 5EH [email protected] [email protected]
Ac ce p
te
d
M
an
Corresponding author: Prof Caroline Wilkinson Centre for Anatomy & Human Identification MSI Building Dow Street University of Dundee UK DD1 5EH Tel: 01382386324 [email protected]
us
1
Page 2 of 19
3
THE USE OF CRANIOFACIAL SUPERIMPOSITION FOR DISASTER VICTIM IDENTIFICATION
ip t
ABSTRACT
Ac ce p
KEYWORDS
te
d
M
an
us
cr
Skull-to-face comparison is utilised for human identification where there is a suspected identity and the usual methods of identification, such as DNA or dental comparison, are not possible or practical. This research aimed to compare the reliability of manual and computerised craniofacial superimposition techniques and to establish the application of these techniques for disaster victim identification, where there may be a large database of passport-style images, such as the MPUB Interpol database. Twenty skulls (10 female; 10 male) were utilised from the William Bass Skeletal Collection at the University of Tennessee and compared to face pools of 20 face photographs of similar sex, age and ethnic group. A traditional manual photographic method and a new 3D computer-based method were used. The results suggested that profile and three-quarter views of the ante-mortem face were the most valuable for craniofacial superimposition. However, the poor identification rate achieved using images in frontal view suggests that the MPUB Interpol database would not be optimal for disaster victim identification, and passport-style images do not provide enough distinguishing facial detail. This suggests that multiple ante-mortem images with a variety of facial expression should be utilised for identification purposes. There was no significant difference in success between the manual and computer methods.
Craniofacial superimposition, identification, disaster victim
INTRODUCTION
Facial anthropology involves biological analysis and interpretation of the skull and/or the face and can be applied to identification of the living (biometrics, image analysis) and the dead (facial depiction, craniofacial superimposition, osteology, anatomy)[1]. Craniofacial analysis of the dead can be utilised for single unidentified human remains, multiple victims of disasters and mass graves. In the majority of forensic investigations there will be a suspect in relation to identity; in single forensic cases there may be a missing persons list, in mass disasters there may be a closed list (such as a passenger list for a transport system) or an open list (reported missing by families and/or employers) and for mass graves there may be whole missing populations. In these circumstances there may be ante-mortem data available relating to the suspects and this might include biological profiles (age, sex, stature, ethnic group), personal information (body modifications, identifying marks, clothing, jewellery, hair style etc.),
Page 3 of 19
4 photographs (ID cards, passport images, family albums or snapshots), hair samples (collected from a hairbrush etc.), dental records, clinical images and/or medical records [2].
ip t
However, much of this ante-mortem data may not be available, for example where the individual did not have a history of dental or medical treatment, or the available data may not be useful, for example where multiple members of the same family are missing (such as in a mass grave or mass disaster) and DNA analysis cannot separate family members. This may be significant in areas/countries with high levels of poverty, low socio-economic status and poor medical/dental practise.
us
cr
Where ante-mortem images are present, craniofacial superimposition may be effective as it does not require expensive or invasive techniques and is cost and time efficient. Craniofacial superimposition is the process where ante-mortem images are aligned and matched to the skull in order to assess the relationship between the hard and soft tissues of the face. This analysis may allow positive identification, especially where multiple ante-mortem images are available and this has been accepted in international courts as a method of identification [3-8].
an
Traditionally craniofacial superimposition has been carried out by forensic anthropologists or anatomists [5, 7, 9] and the techniques incorporate similar anatomical principles [10] and anthropological standards [11] as utilised in facial depiction. Numerous computerised systems have been developed for skull-to-face alignment [8, 12-15].
Ac ce p
te
d
M
The first documented use of craniofacial superimposition for identification in a medico-legal investigation was in 1935 [3], with a particular case involving a Lancastrian GP and the mysterious disappearance of his wife. Dr Ruxton claimed that his wife had left him for another man, but two weeks later two dismembered bodies were found in Glasgow. Police recovered two human heads and over seventy body parts wrapped in newspaper. The newspaper was from a special edition that was distributed only in the area where the Ruxtons lived. Dr Buck Ruxton had killed his wife and her maid and removed the eyes, noses, lips, skin and teeth to avoid identification of the bodies. The police suspected that the maid and the wife were the victims and a craniofacial superimposition was carried out using ante-mortem images of the two women and photographs of the skulls [16]. The method used in this case was photographic; employing enlargement, measurable objects, anatomical landmarks and craniophore orientation [17]. Known objects in the photographs (e.g. a tiara and a picket fence) were used to enlarge the faces to life size in order to identify Mary Rogers and Isabella Ruxton. There have been other significant forensic cases where craniofacial superimposition has been utilised for identification of human remains, including the identification of the remains of Josef Mengele [7, 18] and the identification of the victims of the serial killers, Fred and Rosemary West, in the UK [19]. Other significant cases are the Dobkin case [4], the Worlkersdorfer case [5] and the Howick Falls murder case [6]. Detailed reviews of the techniques can be found in the literature [17, 20-27] and many case studies have also been published [8, 13, 28-42]. Craniofacial superimposition development has passed through three technological phases: photographic, video and computer-assisted [17, 25]. The photographic technique was pioneered in the 1930s [3-5], the video technique was developed in the 1970s [43-45] and the computerassisted technique was introduced in the 1980s [12, 36, 46]. It is professionally agreed that craniofacial superimposition is of greater value for exclusion than positive identification [17] and that where single facial images are utilised, more than one person may appear consistent with the skull in question. Therefore, forensic practitioners must be well trained in anatomy and anthropology for the effective utilization of craniofacial superimposition and multiple ante-mortem images of the suspect should be analysed. When evaluating anatomical consistency special attention is paid to the cranial outline, the soft tissue thickness at various anthropometric landmarks on the skull and feature relationships between the skull and the face [47].
Page 4 of 19
5
us
cr
ip t
It is vitally important to practitioners and law enforcement agencies to establish the level of accuracy of these craniofacial techniques, especially in the case of craniofacial superimposition where this may be utilised in court for positive identification. The credibility of craniofacial superimposition was first established in a study using a database of 52 skulls from the Smithsonian collection [8]. An unknown skull was shown to match an ante-mortem facial photograph, while the four most similar skulls from the collection showed distinct differences to the facial photograph. Another blind study [48] used three skulls and compared them to 97 lateral view and 98 frontal view photographs of subjects (including the targets); in total 585 superimpositions. False matches were recorded for 9.6% of the lateral view and 8.5% of the frontal view superimpositions and all 3 targets were correctly matched. The incidence of false matches was reduced to 0.6% when both frontal and lateral view photographs of the same individual were used, suggesting that multiple photographs are optimal to prove or disprove identity by craniofacial superimposition. A further study [17] found that the outline from trichion to gonion in the lateral or oblique view was the optimal portion for personal identification.
M
an
Technological advances in the field of facial imaging over the past 20 years have produced new opportunities for research in facial morphology and facial growth assessment [49]. Surface scanning technologies such as laser scanners, photogrammetry cameras and other threedimensional systems (including three-dimensional digitizers and structured light systems) have advanced craniofacial landmark location [50-52] and facial anthropometry in both adults and juveniles [53-58]. These systems have benefited the field of orthodontic and plastic surgery with the ability to record subtle changes in surface morphology, especially important in the monitoring orthodontic treatment, facial growth and maxillofacial intervention [59, 60].
te
d
Numerous accuracy studies have been conducted on different laser scanners with regard to the accuracy from three-dimensional capture of faces of the resulting scan measurements when compared with anthropometric measurements from the face [52, 60-62]. These studies all found that laser scanners are a reliable and fast means of capturing a three-dimensional reproduction of a facial surface whilst maintaining high levels of accuracy with regard to the difference in resulting measurements between the scan and the subject. However, Aung et al, [52] did indicate that accurate location of landmarks and operator skill are important factors in achieving reliable results.
Ac ce p
The ability of a forensic practitioner to reliably cite the accuracy of a particular method employed in forensic human identification is highly desirable. The reason for this is not simply for legal reasons but also as a moral and ethical obligation to the victims and their families. This will ensure that if the practitioner is required to provide an expert witness report to the court (with regard to the identification they have assigned to the deceased) the evidence provided is valid with the support of scientific research. In order to improve craniofacial analysis techniques further rigorous quantitative accuracy studies need to be conducted especially with regard to the scientific method. Improving the accuracy of these techniques and increasing publication of the findings of accuracy studies will allow the use of craniofacial analysis techniques more widely in the field of forensic human identification, especially with regard to disaster victim identification and also increase the acceptance of these methods in a medico-legal investigation. The aim of this research was to compare the accuracy of manual and computerised craniofacial superimposition techniques and to establish the application of these techniques for disaster victim identification, where there may be a large database of passport-style images, such as the MPUB Interpol database.
Page 5 of 19
6
METHOD This study utilised a three-dimensional laser scanning system (FastSCAN™ Polhemus Scorpion™ handheld laser scanner) [63] for collection of 3D skull models for use in craniofacial analysis.
us
cr
ip t
Data collection was conducted on the William Bass donated skeletal collection from the University of Tennessee, USA, following the approval of the University of Tennessee Ethics Board. This collection consists of skeletal material and related ante-mortem images. Twenty (10 male and 10 female) crania and their associated mandibles were scanned using a glass platform and wooden mounting pole to enable 360 degree scanning ability. The crania were scanned separately from the associated mandibles to ensure as much detail was captured as possible. The mandibles were scanned using a glass platform that enabled scanning from all views. Before scanning of the crania and mandible commenced, photographic images were captured of the skeletal remains from numerous angles to display any features that the laser scanner may not be able to acquire. The related ante-mortem images were collected and employed as the target images.
an
Three male face pools and three female face pools were created using the target images in addition to a large number of other ante-mortem images collected from unrelated cases in the William Bass Collection. This enabled the creation of the face pools with consistent demographic profiles. The following face pools were created: Female Face Pool 1 -FFP1 (20 faces – 3 targets, 17 foils)
Female Face Pool 2 - FFP2 (20 faces – 3 targets, 17 foils)
M
Female Face Pool 3 - FFP3 (20 faces – 4 targets, 16 foils) Male Face Pool 1 - MFP1 (20 faces – 3 targets, 17 foils)
d
Male Face Pool 2 - MFP2 (20 faces – 3 targets, 17 foils) Male Face Pool 3 - MFP3 (20 faces – 4 targets, 16 foils)
Ac ce p
te
Within each face pool there were a variety of images, ranging from candid shots to professional portraits. The image quality ranged from poor quality (i.e. photocopies of original documents and low resolution photographic shots – small pixel count) to high quality (i.e. higher resolution photographic shots – large pixel count) and the angle of the faces in the photographs varied from frontal to profile images. The age of the images also varied (with some from the 1960s) although the age of the subjects in the images was similar. The individuals in the face pools also displayed a variety of facial hair and clothing, some of which caused areas of the face to be distorted or obscured. This variation in the face pools represented a typical DVI scenario where photographs would originate from a variety of sources. The twenty skulls were equally divided between the two methods with 10 skulls per method (with 5 females and 5 males in each group). The craniofacial superimposition guidelines used to determine a successful or unsuccessful CFS were based upon the morphology of the skull and the closeness of the match between the skull and the ante-mortem (AM) image (no measurements were taken). The researcher created a set of 3 stages to aid with determination of a match (for both manual and computerised), these were as follows: Stage 1: Individuals were eliminated if the skull did not fit within the facial outline of the AM image. Stage 2: Individuals were eliminated if skeletal features did not match in position or morphology to the AM image. Stage 3: Additional AM images were requested for further analysis and elimination. Not all individuals had additional photographs to reflect a real life DVI scenario. The remaining
Page 6 of 19
7 individuals were then classified as a match or no match to the skull based on all available material. The two methods of craniofacial superimposition (CFS) included: 1. The Manual method was an updated version of a pre-existing method described by Seta and Yoshino [64] for craniofacial analysis that utilises a remote guided motor driven machine to orientate the skull for image capture.
an
us
cr
ip t
The manual method required physical skulls for use with the skull orientation machine. However the original skulls were not available and replicas of the 10 skulls were created by stereolithography (rapid prototyping) [65]. The mandibles were then temporarily fixed to the crania with modelling putty (based upon guidelines produced in [66]). The skull orientation machine was set up in front of a high resolution digital web camera which recorded the movement of the skull model and transmitted the live feed to a PC (see Figure 1). Current literature states that the optimum objective length for photographic superimposition is at 1 metre [23, 24]. Craniofacial superimposition can be conducted with the aid of free online software called Glass 2K (Chime Software Ltd) [67] which enables the user to change the opacity of computer windows. The user can set up the web camera feed and then overlay an image and change the opacity of the image file window to enable the web camera feed to be visualised through the image (see Figure 1). Two hundred manual craniofacial superimpositions were conducted utilising this method for the ten skull models with each skull model undergoing a separate craniofacial superimposition for each face in the face pool.
M
2. The Computer Method was a newly developed and utilised a combination of specialist computer hardware (haptic device) and 3D modelling software to orientate a digital skull model in virtual space for image capture.
Ac ce p
te
d
The computer method also utilised the same procedure, except that it did not require a physical skull model as the entire process was conducted in a virtual space. The laser scan models were used (in an .obj format) and were imported into a 3D modelling software package (Geomagic’s Freeform Modelling®) [68] where the mandible was fixed to the cranium in a relaxed resting position (in the same position as for manual). This modelling package allows a 3D model to be orientated in a virtual space when used in combination with a piece of computer hardware called a Phantom® Desktop™ Haptic Device [69], which enables the user to rotate and orientate a 3D model using an arm-like tool. The computer method allows the user to position a 3D skull model file in virtual space (see Figure 2) and then position an AM image (on a virtual plane) anterior to the skull model. The opacity of the ante-mortem image can be altered to allow better visibility of the 3D skull model. The user can then use the haptic device to orientate the skull model to match in scale and position to the face in the AM image (See Figure 2). A screen shot can then be captured and imported into Adobe Photoshop © CS3 [70] for final analysis and demonstration. Two hundred computer craniofacial superimpositions were conducted utilising this method for the ten skull models with each skull model undergoing a separate craniofacial superimposition for each face in the face pool.
RESULTS
The CFS methods correctly matched the target to the skull at stage 3 in 40% manually and 50% by computer. It was observed that the overall correct match rate at stage 3 was 45% (see Tables 1 & 2). The false match rate at stage 3 was 15% manually and 8% by computer, giving an overall false match rate of 12%. The manual method matched 17% and the computer method matched 11% of the total faces to the skull at stage 3. Ten (50%) of the targets were falsely rejected at stage 1 and a further one (5%) falsely rejected at stage 2. The manual method rejected more targets (60%) at stage 1 than the computer
Page 7 of 19
8 method (40%) and the computer method rejected more targets (10%) at stage 2 than the manual method (0%) (see Tables 1 & 2).
cr
ip t
Thirty-six per cent of foil images were of low, 47% medium and 18% high quality. Fifty per cent were colour and 50% were black & white images, with 68% in frontal, 19% three-quarter and 14% profile views. The majority (76%) of images displayed a smiling face with 40% open mouth smile and 34% closed mouth smile. Twenty-three per cent of images displayed an individual not smiling with a closed mouth and relatively few individuals displayed no smile and an open mouth (5%). The majority of foil images displayed no facial distractions (eg. glasses, hats, facial hair) of any kind (57%), however 18% of individuals were observed to wear spectacles, 18% of individuals showed facial hair and relatively few individuals displayed a mixture of spectacles, facial hair and clothing. Visible photographic distortion was recorded in a relatively small number of images (11%).
an
us
Forty per cent of target images were of low, 30% medium and 30% high quality. Sixty per cent were colour and 40% black & white images, with 50% in frontal, 25% three-quarter and 25% profile views. The majority (70%) displayed a smiling target with 45% of images displaying an open mouth smile and 25% displaying a closed mouth smile. Twenty-five per cent of images displayed the target not smiling with a closed mouth and only 5% displayed the target not smiling with an open mouth. The majority of targets displayed no facial distractions of any kind (80%), however 5% were observed to wear spectacles, 10% wore facial hair and 5% wore both spectacles and facial hair. Visible photographic distortion was only recorded in 5% of target images.
d
M
Most of the targets (67%) that reached stage 3 had non-frontal view images, whereas most of the targets (60%) that were eliminated at stage 1 had frontal view images (see Figure 3). This would suggest that CFS was more successful where the target was seen in profile or ¾ views. These views enable observation of the profile of the face and the jaw line, which can result in optimal assessment due to the relatively thin soft tissue thicknesses over the gonial, nasal and orbital areas.
te
Eighty-nine per cent of the targets that reached stage 3 did not have any facial distractions, and 30% of the targets eliminated at stage 1 displayed facial distractions (see Figure 4). This would suggest that facial distractions had a small effect on CFS success.
Ac ce p
Forty-four per cent of the targets that reached stage 3 displayed an open mouth expression and 60% of the targets that were eliminated at stage 1 displayed an open mouth expression (see Figure 5). This would suggest that CFS was less successful when the AM image depicted an open mouth expression. There was no observable difference between colour and black and white target images in relation to successful match rates, and there was no significant difference in match rates in relation to visible photographic distortion or hair thickness.
DISCUSSION
These results are similar to previous accuracy tests. Austin-Smith and Maples [48] recorded a 9.5% false match rate, which is similar to the 12% false match rate in this study. Yoshino and colleagues (47) also found that the profile and three-quarter views were the most valuable for craniofacial superimposition. However, the poor identification rate achieved using images in frontal view suggests that the MPUB Interpol database would not be optimal for disaster victim identification, and passport images do not provide enough distinguishing facial detail. Gordon and Steyn [71] carried out a craniofacial superimposition study using forty frontal face images of cadavers, ten skulls for comparison with each face (400 comparisons in total) and morphological and landmark superimposition techniques. They also found that frontal images have limited use for identification and recommended corroborative evidence.
Page 8 of 19
9 This study suggests that to improve the chance of a successful CFS in a DVI scenario the following standardised protocols should be followed: 1. The ante-mortem images should be of a high quality and without facial distractions or visible distortion. 2. Ideally at least two images should be utilised; neutral facial expression with a closed mouth and slightly open mouth showing the upper teeth.
ip t
3. Multiple views including profile and ¾ views are ideal. Frontal views alone are not optimal.
cr
There was no significant difference in success between the manual and computer methods, and this suggests that digital technology is as reliable as traditional methods. However, neither method in this study utilised the physical remains, and the detail provided by the laser scans was consistent with the stereolithographic replicas.
an
us
This study concludes that craniofacial superimposition could be a useful tool in disaster victim identification for narrowing the possible identifications of individuals in a large scale closed disaster (such as a commercial aeroplane crash where the manifest of passengers is known) as long as the suggested guidelines are followed. When used in combination with other disaster victim identification techniques, such as DNA and dental analysis, craniofacial superimposition can result in a fairly reliable and accurate outcome for the identification of individuals.
1. 2. 3. 4. 5. 6. 7. 8. 9.
Ac ce p
REFERENCES
te
d
M
Following the success of the computer method, further studies should focus on decreasing the amount of expensive equipment and computer software and the development of a bespoke craniofacial superimposition software system. The researcher noted that although the haptic device was a useful tool for the orientation of the skull, it was the least portable aspect of the CFS equipment and also the most expensive piece of equipment used. Any bespoke computer software would require an ability to position three-dimensional models in virtual space in combination with a two-dimensional plane to support an AM image. This software should also include a sliding scale option for perspective views enabling the user to adjust the scale in response to any photographic distortion observed within the AM image.
Lee, W.J., Mackenzie, S. and Wilkinson, C.M. (2010) Facial Identification of the Dead. In: Forensic Anthropology 2000-2010, Ferguson, E. and Black, S.M. (Eds) CRC Press. Black, S. and Hackman, L. (2010) Disaster Victim Identification. Wiley Online Library. Glaister, J. and Brash, J.C. (1937) Medico-legal aspects of the Ruxton case. Vol. 33. Livingstone. Simpson, K.. (1943) Rex vs. Dobkin: the Baptist Church Cellar Murder'. Medico-Legal Review, II, p. 132-45. Gordon, I. and Drennan, M. (1948) Medico-legal aspects of the Wolkersdorfer case. South African Medical Journal, 22: p. 543-549. Prinsloo, I. (1953) The identification of skeletal remains in Regina versus K and another: The Howick Falls murder case. J. Forensic Med, 1: p. 11-7. Helmer, R.P. (1987) Identification of the cadaver remains of Josef Mengele. Journal of Forensic Sciences, 32(6): p. 1622. Ubelaker, D.H., Bubniak, E. and O'Donnell, G. (1992) Computer-assisted photographic superimposition. Journal of Forensic Sciences, 37(3): p. 750-762. Maat, G. (1989) The positioning and magnification of faces and skulls for photographic superimposition. Forensic Science International, 41(3): p. 225-235.
Page 9 of 19
10
16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31.
ip t
cr
us
15.
an
14.
M
13.
d
12.
te
11.
Gerasimov, M. (1955) The reconstruction of the face from the basic structure of the skull. Moscow: Nauka. Wilkinson, C., (2010) Facial reconstruction–anatomical art or artistic anatomy? Journal of Anatomy, 216(2): p. 235-250. Delfino, P. V., Colonna, M., Vacca, E., Potente, F., & Introna Jr, F. (1986). Computer-aided skull/face superimposition. The American Journal of Forensic Medicine and Pathology, 7(3), 201 Shahrom, A. W., Vanezis, P., Chapman, R. C., Gonzales, A., Blenkinsop, C., & Rossi, M. L. (1996). Techniques in facial identification: computer-aided facial reconstruction using a laser scanner and video superimposition. International Journal of Legal Medicine, 108(4), 194-200. Al-Amad, S., McCullough, M., Graham, J., Clement, J., & Hill, A. (2006) Craniofacial identification by computer-mediated superimposition. The Journal of Forensic Odontostomatology, 24(2): p. 47. Ibáñez, O., Cordón, O., Damas, S., & Santamaría, J. (2008). Craniofacial superimposition based on genetic algorithms and fuzzy location of cephalometric landmarks. In: Hybrid artificial intelligence systems (pp. 599-607). Springer Berlin Heidelberg. Glaister, J. and Brash, J.C. (1953) The Ruxton Case. Medical Jurisprudence and Toxicology, 1953: p. 90-101. Yoshino, M. (2012) Craniofacial superimposition. In: Wilkinson, C.M. and Rynn, C. (Eds) Craniofacial Identification. Cambridge University Press; p. 238. Teixeira, W.R.G., (1985) The Mengele report. The American Journal of Forensic Medicine and Pathology, 6(4): p. 279-283. Whittaker, D., Richards, B. and M. Jones, M. (1998) Orthodontic reconstruction in a victim of murder. Journal of Orthodontics, 25(1): p. 11-14. Brown, K. (1983) Developments in cranio-facial superimposition for identification. The Journal of Forensic Odonto-stomatology, 1(2): p. 57. Brocklebank, L. and Holmgren, C. (1989) Development of equipment for the standardization of skull photographs in personal identifications by photographic superimposition. Journal of Forensic Sciences, 34(5): p. 1214. Lan, Y. (1992) Development and current status of skull-image superimpositionMethodology and instrumentation. Forensic Science Review, 4: p. 125-125. Sekharan, P.C. (1993) Positioning the skull for superimposition. In: Iscan, M.Y. and Helmer, R.P. (Eds) Forensic Analysis of the Skull. Wiley-Liss, New York. p. 105-118. Cai, D. and Y. Lan, Y. (1993) Standards for skull-to-photo superimposition. In: Iscan, M.Y. and Helmer, R.P. (Eds) From the Bare Bone to the Full Face. Wiley-Liss, New York, p. 171198. Aulsebrook, W. A., İşcan, M. Y., Slabbert, J. H., & Becker, P. (1995). Superimposition and reconstruction in forensic facial identification: a survey. Forensic Science international, 75(2), 101-120. Taylor, J. and Brown, K. (1998) Superimposition techniques. In: Craniofacial identification in Forensic Medicine. London: Hodder Arnold, p. 151-64. Glassman, D. Methods of superimposition. In: Taylor, K.T. (2001) Forensic Art and Illustration. CRC Press. p. 477-498. Sen, N. (1962) Identification by superimposed photographs. International Criminal Police Review, 162: p. 284-286. Basauri, C. (1967) A body identified by forensic odontology and superimposed photographs. International Criminal Police Review, 204: p. 37-43. Sekharan, P.C. (1971) Revised Superimposition Technique for Identification of the Individual from the Skull and Photograph, A. J. Crim. L. Criminology & Police Sci., 62: p. 107. Reddy, K. (1973) Identification of dismembered parts: the medicolegal aspects of the Nagaraju case. Forensic science, 2: p. 351-374.
Ac ce p
10.
Page 10 of 19
11
39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52.
ip t
cr
38.
us
37.
an
36.
M
35.
d
34.
te
33.
Gejvall, N. (1974) Superimposition plus SEM-comparison of hair cuticle for identification purpose. International J. Skeletal Research, 1(974): p. 99-103. Sivaram, S. and Wadhera, C. (1975) Identity from skeleton—A case study. Forensic Science, 5(2): p. 166. Brown, K.A. (1982) The identification of Linda Agostini: The significance of dental evidence in the Albury Pyjama Girl case. The American Journal of Forensic Medicine and Pathology, 3(2): p. 131-142. McKenna, J., N.G. Jablonski, and R. Fearnhead, A method of matching skulls with photographic portraits using landmarks and measurements of the dentition. J Forensic Sci, 1984. 29(3): p. 787-97. Bastiaan, R.J., Dalitz, G.D. and Woodward, C. (1986) Video superimposition of skulls and photographic portraits—a new aid to identification. J Forensic Sci, 31(4): p. 1373-9. Webster, W. P., Murray, W. K., Brinkhous, W., & Hudson, P. (1986). Identification of human remains using photographic reconstruction. In: Forensic Osteology; advances in the identification of human remains. Springfield: Charles C Thomas, 256-89. Yoshino, M. and Seta, S. (1989) Personal identification of the human skull: superimposition and radiographic techniques. Forensic Sci. Rev, 1: p. 23-42. Solla, H.E. and Işcan, M.Y. (2001) Skeletal remains of Dr. Eugenio Antonio Berrı́os Sagredo. Forensic Science International, 2001. 116(2): p. 201-211. Bilge, Y., Kedici, P. S., Alakoç, Y. D., Ülküer, K. Ü., & İlkyaz, Y. Y. (2003). The identification of a dismembered human body: a multidisciplinary approach. Forensic Science International, 137(2), 141-146. Ghosh, A. and Sinha, P. (2005) An unusual case of cranial image recognition. Forensic Science International, 148(2): p. 93-100. Fenton, T.W., Heard, A.N. and Sauer, N.J. (2008) Skull‐Photo Superimposition and Border Deaths: Identification Through Exclusion and the Failure to Exclude. Journal of Forensic Sciences, 2008. 53(1): p. 34-40. Helmer, R.P. and Grüner, O. (1977) Vereinfachte Schädelidentifizierung nach dem Superprojektionsverfahren mit Hilfe einer Video-Anlage. International Journal of Legal Medicine, 80(3): p. 183-187. Iten, P. (1987) Identification of skulls by video superimposition. J Forensic Sci, 32(1): p. 173-88. Koelmeyer, T.D. (1982) Videocamera superimposition and facial reconstruction as an aid to identification. Am J Forensic Med Pathol, 3(1): p. 45-48. Austin, D. (1999) Video superimposition at the CA Pound Laboratory 1987 to 1992. J Forensic Sci, 44(4): p. 695-9. Yoshino, M., Imaizumi, K., Miyasaka, S., & Seta, S. (1995). Evaluation of anatomical consistency in craniofacial superimposition images. Forensic Science International, 74(1), 125-134. Austin-Smith, D. and Maples, W.R. (1994) The reliability of skull/photograph superimposition in individual identification. Journal of Forensic Sciences, 39(2): p. 446. De Greef, S. and Willems, G. (2005) Three-dimensional cranio-facial reconstruction in forensic identification: latest progress and new tendencies in the 21st century. Journal of Forensic Sciences, 50(1): p. 12. Gwilliam, J.R., Cunningham, S.J.and Hutton, T. (2006) Reproducibility of soft tissue landmarks on three-dimensional facial scans. The European Journal of Orthodontics, 28(5): p. 408-415. Bush, K. and Antonyshyn, O. (1996) Three-dimensional facial anthropometry using a laser surface scanner: validation of the technique. Plastic and Reconstructive Surgery; 98(2): p. 226-235. Aung, S., Ngim, R. and Lee, S. (1995) Evaluation of the laser scanner as a surface measuring tool and its accuracy compared with direct facial anthropometric measurements. British Journal of Plastic Surgery. 48(8): p. 551-558.
Ac ce p
32.
Page 11 of 19
12
59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70.
ip t
cr
us
58.
an
57.
M
56.
d
55.
te
54.
Moss, J. P., Linney, A. D., Grindrod, S. R., Arridge, S. R., & Clifton, J. S. (1987). Threedimensional visualization of the face and skull using computerized tomography and laser scanning techniques. The European Journal of Orthodontics, 9(4), 247-253. McDonagh, S., Moss, J. P., Goodwin, P., & Lee, R. T. (2001). A prospective optical surface scanning and cephalometric assessment of the effect of functional appliances on the soft tissues. The European Journal of Orthodontics, 23(2), 115-126. Tuncay, O.C. (2001) Three-dimensional imaging andmotion animation. In: Seminars in Orthodontics. Elsevier. Kau, C. H., Richmond, S., Savio, C., & Mallorie, C. (2006). Measuring adult facial morphology in three dimensions. The Angle Orthodontist, 76(5), 773-778. Kau, C. H., Zhurov, A., Richmond, S., Cronin, A., Savio, C., & Mallorie, C. (2006). Facial templates: a new perspective in three dimensions. Orthodontics & Craniofacial Research, 9(1), 10-17. Kau, C. H., Cronin, A., Durning, P., Zhurov, A. I., Sandham, A., & Richmond, S. (2006). A new method for the 3D measurement of postoperative swelling following orthognathic surgery. Orthodontics & Craniofacial Research, 9(1), 31-37. Ismail, S. and J. Moss, (2002) The three-dimensional effects of orthodontic treatment on the facial soft tissues–a preliminary study. British Dental Journal; 192(2): p. 104-108. Kusnoto, B. and C.A. Evans, (2002) Reliability of a 3D surface laser scanner for orthodontic applications. American Journal of Orthodontics and Dentofacial Orthopedics, 122(4): p. 342. Kau, C. H., Zhurov, A., Bibb, R., Hunter, L., & Richmond, S. (2005). The investigation of the changing facial appearance of identical twins employing a three‐dimensional laser imaging system. Orthodontics & Craniofacial Research, 8(2), 85-90. Park, S. H., Yu, H. S., Kim, K. D., Lee, K. J., and Baik, H. S. (2006). A proposal for a new analysis of craniofacial morphology by 3-dimensional computed tomography. American Journal of Orthodontics and Dentofacial Orthopedics, 129(5), p. 600. Polhemus (2007). FastSCAN™ User Manual. Vermont, U.S.A. http://www.polhemus.com/polhemus_editor/assets/FastSCAN%20COBRA%20C1%20 &%20SCORPION%20Brochure.pdf Accessed last at 22.11 on 02/02/2014 Seta, S. and Yoshino, M. (1993) A combined apparatus for photographic and video superimposition. In: Iscan, M.Y. and Helmer, R.P. (Eds) From the bare bone to the full face. New York: Wiley-Liss; p. 161-9. http://pdronline.info/ Accessed last at 22.10 on 02/02/2014 Taylor, K.T. (2001) Forensic Art and Illustration. CRC Press
Ac ce p
53.
http://chime.tv/products/glass2k.shtml Accessed last at 22.10 on 02/02/2014
Geomagic Freeform Plus ® http://geomagic.com/en/products/freeform/overview/ Accessed last at 22.14 on 02/02/2014 Geomagic Phantom ® http://geomagic.com/en/products/phantom-premium/overview Accessed last at 22.15 on 02/02/2014 Adobe Photoshop © CS8.0 (Adobe) Systems, San Jose, CA; http://www.adobe.com/ Accessed last at 22.15 on 02/02/2014
71. Gordon, G. M., and Steyn, M. (2012). An investigation into the accuracy and reliability of skull-photo superimposition in a South African sample. Forensic Science International, 216(1), 198-e1.
Page 12 of 19
13
ACKNOWLEDGEMENTS
Ac ce p
te
d
M
an
us
cr
ip t
The authors would like to thank the University of Tennessee for the use of the William Bass Skeletal Collection and the University of Dundee for funding this research through the CAHID Greenhouse PhD Scholarship Fund.
Page 13 of 19
14
Figure 1: The manual craniofacial superimposition set up with webcam, lap top and skull orientation machine
ip t
Figure 2: An example of craniofacial superimposition (left) using the computer method (right)
us
cr
Table 1: Results of the manual craniofacial superimposition test (F = female; M = male)
an
Table 2: Results of the computer-based craniofacial superimposition test (F = female; M = male)
M
Figure 3: Survival graph in relation to the angle of view for the target images
d
Figure 4: Survival graph in relation to the facial distractions for the target images
Ac ce p
te
Figure 5: Survival graph in relation to the expression for the target images
Page 14 of 19
15
N = 20
Skull F4
F6
F8
F10
M2
M4
M6
M8
M10
Total
Number of matches
2
1
3
3
5
4
4
3
4
4
17%
False matches
2
0
3
2
5
3
4
3
4
3
15%
Correct match
x
yes
x
yes
x
yes
x
x
x
yes
40%
Stage target eliminated
S1
x
S1
x
S1
x
S1
cr
ip t
F2
S1
x
us
S1
N = 20
M
an
Table 1: Results of the manual craniofacial superimposition test (F = female; M = male)
Skull
F5
Number of matches
1
2
1
False matches
0
1
Correct match
yes
yes
Stage target eliminated
x
x
F7
F9
M1
M3
M5
M7
M9
Total
1
4
1
2
2
4
11%
d
F3
3
te
F1
3
1
3
1
2
2
3
8%
yes
x
x
yes
x
x
x
yes
50%
x
S1
S1
x
S1
S2
S1
x
Ac ce p
0
Table 2: Results of the computer-based craniofacial superimposition test (F = female; M = male)
Page 15 of 19
us
cr
ip t
16
Ac ce p
te
d
M
an
Figure 3: Survival graph in relation to the angle of view for the target images
Figure 4: Survival graph in relation to the facial distractions for the target images
Page 16 of 19
us
cr
ip t
17
Ac ce p
te
d
M
an
Figure 5: Survival graph in relation to the expression for the target images
Page 17 of 19
Ac
ce
pt
ed
M
an
us
cr
i
Figure
Page 18 of 19
Ac
ce
pt
ed
M
an
us
cr
i
Figure
Page 19 of 19
