Thursday, 20 March 2014

Digital facial reconstruction - blind tests and practical application of the protocol

Learning how to model a face on the computer is not an easy task. In the world there are over six billion people and no faces are alike. Even the artist study a lot, he will never be able to model every possible way, he will never be able to master the nuances of complexity that we have in this region of the body so notorious.

Still speaking of these distinct features, I confess I had not realized this thing until I started working with facial reconstruction. In this case the artist or scientist who will rebuild almost always have no idea what will present itself before his eyes until he finish the work. Once he reconstructs and compares the data with the individual's appearance in life, the more he sees that he can not stop studying.

There are professionals in this area are excellent rebuilders of Caucasian, Negroid, or Malays. Hardly an artist/scientist is good at all ancestries. The reason? Simple. Besides the complexity of the face shape, we yet have the details differ one ancestry from another . Could you describe what differs a native Japanese from a Brazilian aboriginal? Besides describing, you could draw? In addition to drawing, you could create in 3D? It really is not easy. Even when we speak of close people, like relatives or friends, it is difficult to mentally reconstruct every detail of their faces. Imagine having to model all this with only a skull as a base!

General Remarks 

Over those two years of studies I had the opportunity to rebuild faces in blind tests, where I received only the skull, estimated data on sex, age and ancestry. With this information in hand I applied the concepts studied and modeled step by step an approximation based on statistical and anatomical data.

Unfortunately I can not show all cases, because some of them I don't have the license of use the images, but overall were successful experiences. Although the result did not get 100% loyalty, the data gathered from the comparison show that meshes volumes created digitally not differ much from those modeled from scientific parameters.

There are many difficulties related to this heterogeneous reality of the shape of the human face. Because of this it is important  always have a help of professionals related with the areas of forensic dentistry, forensic anthropology and medicine. Having access to books and articles on plastic surgery can help a lot, because aside from having a different approach of classical anatomy textbooks, also offers a vast literature on human ancestry.

Still need to have expertise regarding the modeling of some parts of the face. A classic problem involving facial reconstruction is the area above the craniometric points supra and sub M2 (referring to the location of the second molars), or "the cheek".

The nose is another part of difficult approach when we talk of volumetric accuracy. Even so, when crossing the most popular protocols available in publications, at least we has a projection very close to real.

The passing years also influence a reconstruction, because the human skin becomes more lax with the change of collagen fibers over time. In the elderly, for example, there is a tendency of slightly nose-down, influencing the projections based on the scientific literature.

After some studies, I developed a protocol for facial reconstructions using open software. This was done with Dr. Paulo Miamoto and Dr. Rodolfo Melani, both Forensic Dentists from Forensic Odontology and Anthropology Laboratory at Faculty of Dentistry of University of São Paulo (OFLAB-FOUSP). One of the main aspects of the protocol is that once the user undergoes basic training on the softwares that are used, one may take his/her first steps, as the software functions are specified in the manuscript. Another important information is that the protocol features from digitization of the skull from CT scans/digital photographs until the finishing of the reconstruction. This article can be found here:

Studies, partnerships and experience, this is the holy trinity of a good facial reconstruction strongly committed to impartiality.

How work the tests 

The vast majority of studies presented here were carried out in partnership with Dr. Paul Miamoto (FO-USP). 

1) Initially I send him a 3D skull extracted from a CT-scan.

2) Dr. Miamoto analyze the material and inform me, within the limitations of 3D models, he estimate parameters that comprise the biological profile of the individual, such as gender, age and ancestry. In addition, other relevant information, such as ante-mortem injuries, diseases and other dentocraniofacials changes that can cause changes in facial appearance are also pointed out.

3) With the data in hand I start the process of facial reconstruction, carving the major muscles of the face and other structures, as well as the nose and lips from tissues depths tables.

4) As soon as I finish the basic modeling I send the result to Dr. Miamoto. After preliminary analysis to review the anatomical, anthropological and odontoforensic parameters, the job is routed to completion.

5) For comparison with the true face of the individual, the viewer program tomographic images, the individual's skin is exported as a 3D model. This model of the "real" skin and carved skin are open in a mesh comparison program. The results are shown in a color map with a graph that quantitatively represents the difference in depth between the two. The more the color inclines to green, the more the difference approaches 0 mm.


Despite the great satisfaction with the results I and my research partners are cautious. As explained above, perform well modeling a group or an ancestry does not mean that the technique is mastered. Does not mean that success will be our constant companion. We will only have a broader sure if we study more, and even we study for the rest of life, hard, all the days we still have, surely we know very little before all the possibilities that this world facial heterogeneity offer us.


To Dr. Marcos Machado Paulo Salles, for the opportunities related to forensic facial reconstruction studies. To Dr. Adriana Dal'Acqua by research material and valuable information related to facial geometry.
BlogItalia - La directory italiana dei blog Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.