Introduction
Identification of human remains has been a major problem for the medicolegal system. Thousands of people every year are buried with their identity and that of their killer unknown. Mass murders have been committed in many parts of the world, e.g. in Europe, Africa and South America, by dictatorial regimes or as a result of ‘ethnic cleansing’. In many countries there is a lack of identification records, such as dental charts or DNA databases, or even ordinary identification cards. In affluent countries, not every individual has a dental record or has been reported to the police as missing. The only approach to identification may be facial reconstruction if the remains are entirely decomposed.
In general, facial identification techniques depend upon the available evidence and therefore may be classified as follows:
• Video or photographic comparison Comparison of a video image or photograph with the actual face, video or photograph of a suspect.
• Skull-photo superimposition Overlaying of a facial image (portrait, photograph or video footage) on a skull.
• Facial restoration When sufficient soft tissues persist on the head, the face may be restored to its original appearance.
• Facial reconstruction This technique is applied to the skull that has no persistent soft tissue.
The facial reconstruction technique has emerged as an important tool in the identification of unknown human remains. It differs from facial restoration, which deals with a face that is partly present but altered as a result of decomposition or trauma. The ultimate goal is to help identification of the unknown person. The reconstruction can be two-dimensional or three-dimensional, both of which can be performed artistically or done on a computer.
Obviously the relationship between every point of the bony face and the corresponding point on the soft tissues cannot be known with precision. Unfortunately the skull cannot give every clue needed to permit a perfect reconstruction. Although soft tissue thickness and bony variation and dimensions in many parts of the face are known for many populations, subtle details of the face will probably never be learned from the bony frame. Therefore, if the reconstructed face leads the investigator to the right person, or stimulates the public eye to recognize the picture, the mission may be considered accomplished. Despite these limitations, this form of identification is becoming an important tool for medicolegal investigation and is attracting research interest. Although there is no systematic survey or scientific test, the identification success rate is estimated to be about 50% or better.
In spite of many efforts to reconstruct the face from the skull, there still are very few research publications with scientific validation.
The aim of this article is to discuss the methods used in forensic facial reconstruction and the approaches taken to recreate the visual appearance of the unknown person. Over the years, faces and busts of historical, well-known people have been reconstructed: for exhibitions, for authentication (if a given skull belongs to a well-known personality), and to satisfy curiosity about the physical appearance of ancient people. The first attempts were made by German anatomists at the end of the nineteenth century in order to identify historically significant individuals. More recently, Russian anthropologists developed a manual for the three-dimensional technique and used it on ancient ethnic skulls as well as in forensic cases. In the last 25 years the two-dimensional technique has also gained in popularity.
Basis of Facial Reconstruction
Whichever technique is chosen, facial reconstruction requires a detailed analysis of the skull. The analysis should include the following:
• Osteological examination Skull morphology is assessed by observing osteological variation in terms of size and shape. In addition, anthropo-metric measurements are taken to discern the proportional relationship between the skull and the reconstructed face. Both morphological and metric examination also yield information about the age, sex, race and body size of the victim. The goal is to discern race, sex and age, as well as peculiarities of the skull and face such as asymmetry, antemortem health, pathology, trauma, cultural modification of the head and face, and individual habits. These can later be incorporated in the final reconstructed face. Standard university-based osteological courses are the best source for gaining an understanding of human variation of this nature.
• Cephalometric/radiographic analysis Although radiographs give some limited information about the biological characteristics mentioned above, it is also important for discovering peculiarities that are not otherwise visible to the naked eye. Both osteo-logical and radiographic examinations should give enough information to create an image of the face which would ‘individualize’ the skull of a specific person.
• Soft tissue thickness Average soft tissue depth at specified skull landmarks is well known for many populations. Its variation has also been investigated for individuals who are thin, of average weight or obese.
Soft Tissue Thickness
One of the main issues in facial reconstruction is to understand the variation of soft tissue thickness in one’s face, as well as within and between populations. The earliest research on facial tissue thickness goes back to the 1880s. It was carried out by inserting a thin knife blade at selected landmarks on cadavers. Later in the same century, measurements were taken with a needle. This needle technique was later adopted by the Japanese in the 1940s and the Americans in the 1980s.
Nevertheless, measurements on cadavers have been criticized for various reasons. The anthropological landmarks are not always easy to locate with certainty by palpating the face of cadavers, nor are the landmarks close enough to each other. Soft tissue deformation may occur during this crude process. The cadaveric alterations resulting from dehydration, rigidity and gravity may alter the soft tissue depth, probably even within a few hours after death. The bloating stage of the decomposition process increases the errors further. The correlation between the soft tissue depths measured on cadavers and living subjects is not known.
Recent investigators recommended that tissue depths should be measured in vivo. Commonly used equipment includes the ultrasonic echo detector, computerized tomography (CT) and magnetic resonance imaging. Furthermore, considering the diversity in human variation even in the same region, a large sample size is needed to assess the effect of such factors as age, sex and body weight. The dental occlusion pattern also plays a role in obtaining the right measurements. Using lateral cranial radiographs, it has been shown that the soft tissue depths showed age and sex differences as anticipated.
Two-dimensional Methods
Facial reconstruction rendering a two-dimensional (2D) view of a face can be accomplished by sketching on paper as well as on a computer. The easiest approach is to take a picture (full and/or profile view) of the skull at the Frankfort horizontal plane and enlarge it to its original size. Facial tissue thickness can be marked at known landmarks and these marked dots can be connected with each other to make a general outline of the face. Forensic artists are able to draw the face from this image, if properly guided. Therefore they usually work under the supervision of a forensic anthropologist who has already analyzed the remains and its identifying skeletal characteristics, i.e. age, sex, race, body size, anomalies, trauma, pathological lesions, antemortem health status, and other unique features. The anthropologists should be able to describe where the eyes, ears, nose and mouth should be drawn in relation to the bony face. This requires a systematic understanding of human facial features. Based on these characteristics a face can be reconstructed, step by step. There are already several publications showing results of these studies. A similar attempt can be made using a lateral radiograph. Average soft tissue thicknesses are marked on the film or its copy and connected with each other.
Obviously certain aspects of the face must be modified, based on sex, race and age. The nose, for example, changes throughout life. Other features, such as the shape of the mouth and thickness of the lips, also change with age and their relationship with the facial skeleton. The 2D technique (sketches) has been used by some anthropologists to save time and reduce the cost of an artist.
Three-dimensional Methods
Facial reconstruction producing a three-dimensional (3D) result is the most desirable. It can be viewed in all aspects, rather than just a full face or profile view. The most traditional 3D approach is the manual build-up of the skull with a clay-like substance. Recently, however, computers have been used to give the 3D appearance.
Manual (plastic, sculptural) approach
The manual approach is one of the most popular 3D methods. Material (e.g. clay) is applied directly onto the skull, using the relationship between bone and soft tissues. This method was developed by German anatomists in the 1880s, and then adopted and modified by Russians, eastern Europeans and eventually the rest of the world. Its forensic use, however, goes back to the late 1940s. Today, practically every forensic artist, pathologist and dentist has experimented with the technique. There are even university courses designed to teach this subject.
When developing this 3D method, it is also necessary to assess skull morphology and metric characteristics, as was done for the two-dimensional analysis. In starting the reconstruction, marks are placed on precise anthropological points where the average tissue depths are known for the specific age and sex of that population. The space between these points is filled with clay or a similar substance (Fig. 1). Bit by bit the whole face is reconstructed (Fig. 2). Areas such as the ears, eyes, nose, mouth and lips are difficult to place because they do not have clear bony indicators. The literature varies as to the exact location of these facial structures.
Figure 1 Classical manual 3D reconstruction: beginning of the reconstruction.
Figure 2 Classical manual 3D reconstruction: end of the reconstruction.
There are rarely experimental comparisons between the reconstructed face and the actual face of the deceased (Fig. 3). Similarly, there has not been any systematic comparison of the same face reconstructed in two and three dimensions. In one experiment, two anthropologists produced sketches (full and profile views) and clay images. When compared, it was noted that the similarities were slight. Yet the two images showed similar proportions in the orbit height, nose length and total, upper and lower face dimensions in the profile view.
Computerized methods
Usually computerized methods allow 2D rendering but some computers can generate a virtual image. Two-dimensional programs operate mostly by importing images or morphing. In one study using a virtual captured digitized image of the skull, the soft tissue contour was drawn using the soft tissue depths. The program provided an average of a series of features from a scanned face database. The average feature was then superimposed, like a mask upon the face. This method may limit the possible variation around the ‘average’ face. Newer programs have become more flexible and the process of reconstruction can be constantly checked by looking at a previous step.
The computerized 3D facial reconstruction has been developed by only a few scientists. Many are at an experimental stage. Successful identification should increase if a person is observed from different directions. Such data can be obtained by CT. Very often the conventional clinical CT protocol uses a plane which is parallel to the Frankfort horizontal. Laser scanning systems can also capture 3D images. In one study, for example, a color laser scanner is used to generate a representation of the scanned object as a 256 x 256 matrix. A wire frame of 256 x 256 radii is reconstructed. The wire frame matrix of the skull is transformed, using the tissue depth distances, through an algorithm which generates a 3D facial reconstruction.
The advantages of computerized methods are numerous because of the speed and the possibility of rapidly editing several versions of the reconstruction. Characteristics such as obese versus emaciated face can be easily altered. Color and textures can be changed quickly. The image can be rotated in three dimensions. The image could cycle through a variety of versions of the reconstruction based on the known range of tissue depth or other feature. There are additional benefits of a computerized reconstruction. Replication is possible when reconstruction is developed from one skull and uses the same computer setup. The image can then be animated to display a moving face. A 2D image can easily be written on a CD-ROM disc or recorded on a videotape. The image can be sent via the Internet or made accessible on the web. Age changes can be made for an older person. The face can be compared with the digitized image databases of missing persons. It is possible to rotate, zoom and alter the original image. The skull and virtual facial reconstruction can be superimposed on each other and the virtual facial reconstruction can be compared with the actual photograph of the missing person.
Figure 3 Comparison of (A) the actual face of the subject, and (B) the polyester resin casting of the reconstructed face.
A review of the literature reveals that there are only a few articles on the subject of facial imaging and computer-assisted reconstruction. Only a few of the computed facial reconstruction programs are true 3D. One of them is based on the distortion of volumes assessed by the process of field morphing, using multiple pairs of disks. In this technique the original skull (S1) is altered into another (S2), and the same algorithm is applied to the original face (Actual Face AF1) to get the unknown face (Reconstructed Face RF2). The disks have to be placed on key levels all around the reference skull to be reconstructed. In another study two sets of models are used. The first set consists of a skull S1 and a known facial cast F1, the reference set. The second set is the skull to be reconstructed (S2), and the ‘unknown’ face (RF2) to be produced. Data acquisition was made by the digi-talization of the two data sets of S1, S2 and F1 by CT. CT of a living subject is impossible in this protocol because very thin slices are needed and the amount of radiation received by the subject could be too much. CT of the head of a cadaver may raise ethical problems. As an alternative, the skull S1 and the mold of the face of the cadaver (F1) are used. CT obtained a set of slices with an interslice distance up to 1 mm. Those slices were stacked into volumetric images, then the skull and facial model were produced by an isosurface extraction algorithm. The method of reconstruction was a global parametric algorithm that transforms the reference skull S1 into S2. Then the same algorithm is applied to the reference head F1, to obtain the image of the unknown face (RF2). The experiment allows researchers to compare the reconstructed face (RF2) with that of the actual face (AF2). It should be noted that the transformation was not based on the traditional landmarks of the skull, but rather only on some salient lines called crest lines. Mathematically these lines are defined by differential invariants and correspond to lines of absolute maxima of the largest principal curvature. On the skull these lines correspond, for example, to the outlines of the mandible, the orbits, the cheekbones or the temples. In fact, when the crest lines are extracted from each skull, the algorithm that matches the set of crest lines of the first skull on to the set of crest lines of the second skull can be obtained. The algorithm is able to generate directly a 3D image of the face from the unknown skull. The image can be easily translated, rotated, zoomed, moved in a continual sequence and altered in different ways, including adding hair or spectacles. In addition, a set of landmarks can be selected and, if necessary, the distance between face and landmarks may be computed.
Discussion
Facial reconstruction is a difficult task. Some of the problem areas are the nutritional status and individual aging intensity as reflected in the face. Necessary information needed, such as the details of the nose, eye, ear, lips and chin, is very difficult to guess from the facial skeleton. Because of discrepancies between reconstructed faces and actual faces, some anthropologists have expressed pessimism about the effectiveness of facial reconstruction. However, there may yet be unidentified relationships between bone and soft tissue.
The nose is a good example for illustrating the difficulty in facial reconstruction because it is one of the most important features of the face used in identification. Unfortunately, the bony frame is limited and the tip is varied in shape and its appearance is very difficult to predict from the nasal and other surrounding bones. Furthermore, some parts of the bony frame of the nose (e.g. the nasal spine) are often broken away. The cartilaginous part is often absent, which dramatically increases the difficulties of the reconstruction. There are also population and sex differences in the human face. Facial morphology also changes throughout life. Many individuals may have a longitudinally oriented groove in the nose tip, which represents the overriding of the lateral crus of the alar cartilage over the lateral nasal cartilage, but such a structure is difficult to predict precisely. Another difficulty is how to determine whether a reconstructed face resembles the photograph of the victim. The identification process is very complex. ‘Random items’, like the ponderal status, pilosity and the color of the hair and eyes, are always difficult to guess because of continual changes in these areas as a result of aging, individual preference and disease.
There is a lack of scientific validation of the methods summarized above. One of the main problems raised by 3D reconstruction is the quality of the results and replication. Obviously, replication is necessary for scientific reliability. Most of the reconstructions presented in the literature have been performed quite intuitively and have thus not been tested for their validity. Recognition does not always depend upon the general morphology of the face. In a forensic case, a close relative, when shown the reconstructed face, said it did not look like the victim because the neck looked ‘too long’, and it had a ‘smile’. He stated that in reality the victim ‘rarely smiled’. Also, the victim had ‘dimples’, which were not predicted from the skull morphology. In an experiment using the clay technique, it was found that 19 out of 22 cases resembled the victim’s photograph. The most extensive attempt at validation was carried out by making duplicates of six skulls and letting two experienced artists perform the reconstruction. Then the two reconstructions were compared with the actual faces of the deceased individuals and with each other. The results showed that there was physiognomic similarity between the blind reconstructions, and the actual faces ranged from acceptable to excellent.
There are basically two scientific issues that need to be resolved in attempting to reconstruct a face from the skull of an unknown person. First, reliability refers to accuracy of the results and is an area in which some progress has been made. In several experimental cases, similar faces can be replicated by the same experimenter. In some cases, two different experimenters using the same methodology have obtained similar results. Second, validity refers to the truthfulness of the result, in which one expects to see a resemblance between the reconstruction and the victim. This aspect of facial reconstruction has been randomly attained and is inconsistent even with the same experimenter. In many cases it has been a ‘hit and miss’ result.
While forensic scientists are aware of these difficulties, there is always the question of the practical use of the results, that is, whether the reconstructed image can lead the police to the identity of the victim. There are some suggested solutions to these problems to assist the police in the identification process. One is to provide several shapes of one variant, like nose shape, ear projection, lip thickness, hair color and pilosity, ponderal status, and so on.
In conclusion, facial reconstruction has become an important and scientifically changing field of anthropology. Many techniques have been developed in the last 100 years to predict the facial appearance of an unknown person from his or her skull. In spite of the fact that there are many practitioners of this approach to identification, the relationship between bone and the corresponding soft tissue is not well known and knowledge of soft tissue depth itself is not sufficient to perform accurate facial reconstruction.
The result is therefore only an approximation of the real face. The best one can hope for is to provide a stimulus which may direct the police to the identity of the missing person. This may in turn lead to identification when incorporated with other information collected about the victim. Furthermore, it is recommended that this technique should be carried out only after all others have failed. It is feared that once a ‘wrong’ face is presented to the public, the real identity may never be known. Nevertheless, if the victim’s general characteristics (age, sex, race, height, ante-mortem health status and personal artifacts) are presented to the general public through the mass media, there is a greater chance of establishing the identity of the victim.
