News

"The research involved in the FaciLe project is very innovative and risky"

The project’s name, FaciLe, is an ironic wink for a research program that represents a true scientific challenge and brings together a multitude of skills. (The French word “facile” means easy!)

 

Presentation of the project team and name

 

 

Within the FaciLe project, anthropologists, forensic doctors and maxillofacial surgery specialists play a key role with their anthropometric and anatomic knowledge. That is why the first senior chair-holder was an anthropologist.

Mathematicians play just as strategic a role, however, since facial reconstruction is achieved using mathematical modeling.

As soon as the project was launched in the fall of 2014, Sorbonne University chose a mathematician as junior chair-holder: Maya de Buhan, CNRS research associate at MAP5, the applied mathematics laboratory at University Paris-Descartes.

Since October 2015, the senior chair-holder has also been a mathematician: Peter Deuflhard. Professor at the Free University of Berlin, this German mathematician founded and ran the Zuse Institute Berlin (ZIB), a center for applied mathematics and scientific calculation, for 25 years. Officially retired since 2012, he could have leaned back and relaxed, but apparently that didn’t appeal to him! When Sorbonne University contacted him, he did not hesitate: “The research involved in the FaciLe project is very innovative and risky, since we have no idea whether or not it will succeed. This is the sort of challenge that motivates me.”

 

How can we reduce the range of potential solutions?

 

On a positive note, Deuflhard has already worked on a subject relatively close to facial reconstruction: at the ZIB, he designed a digital tool that predicts the results of maxillofacial surgery, for example in an individual suffering from a malformation like prognathism. The patient’s face is virtually reconstructed in 3D, which lets the surgeon expose the lower and upper jaw of the virtual patient, simulate how to reposition them to correct the malformation, and then replace the soft tissue and skin to view the results in aesthetic terms.

This tool has been used experimentally on several patients and has proven to be effective, since the prediction and the actual result were an almost perfect match. “The problem is that the issue posed by FaciLe is a great deal more complex,” explained Peter Deuflhard.We have the cranium, and the aim is to find the face that goes with it. And there are any number of faces that are potentially compatible with a skull! So our work involves reducing the range of potential solutions.”

The FaciLe team does not yet have an answer for this. The only way to make progress is to test the hypotheses that are emerging from collaborations between mathematicians, anthropologists, forensic doctors and maxillofacial surgery specialists. To validate these hypotheses, the chair has a bank of 3D digital skulls and faces based on scans taken from volunteers. This method helps reconstruct a face blindly based on a cranium, with the user then checking whether or not it resembles the actual face.

This validation is tricky, however, since the resemblance between two individuals is subjective and difficult to quantify. Some people may find that two faces look alike, while others see no connection between them. This raises several fundamental issues in terms of identification: what is it that makes us recognize someone? Do we need to reproduce the exact measurements of each part of the face, or above all stay true to the overall proportions?

 

A first hurdle overcome

 

Despite these many obstacles, the team has already validated an initial hypothesis, that of the relevance of mathematical modeling in facial reconstruction. Chiara Nardoni, a young mathematician who is writing her thesis under this chair, has developed a morphing tool that calculates the elastic deformation between a reference skull for which the person’s face is known and the skull of an unknown individual that needs to be identified. The idea is to then apply the same deformation to the reference face to reconstruct the face of the unknown individual.

Tested on craniums from the database for which the faces of two individuals are known, the algorithm produced very promising results. The bony parts of the reconstructed face (the forehead, the top of the nose) are very close to that of the actual face. However, the rest of the nose, the cheeks and the chin are a lot less similar, especially if the individuals are plump.

 

The next challenges: the facial muscles and nose

 

To achieve greater accuracy in reconstructing cheeks and chins, the team expanded its investigations to include one of the chewing muscles – the masseter muscle – which clearly plays an important role in overall facial physiognomy. Since October 2015, a second PhD student, Lydie Uro, who is both a mathematician and a medical student, has been exploring the mutual influences among the configuration of the skull, the masseter muscle and facial characteristics. If this research reveals correlations among these various elements, the team may manage to deduce the shape and size of the masseter muscle based on the skull, and thus reconstruct the person’s cheeks more accurately, but for now this remains a hypothesis.

As for the nose, the researchers are thinking of a different method. “This facial feature is essentially made up of cartilage and soft tissue. The skull does not have much influence on its shape,” Peter Deuflhard tells us. “But for anthropologists, the anterior nasal spine – the tiny bone located at the base of the nose, above the upper jaw – may provide valuable clues. If we can establish correlations between the morphology of this bone and that of the nose, we could create rules from this to calculate the shape of an individual’s nose.” Here again, these rules are yet to be formalized, validated and computerized.

 

The next step: the face’s secondary characteristics

 

Reconstructing the general structure of the face, however, is only the first step; other characteristics are determining factors in identifying a person. In this area, progress in genomics has already provided a partial answer: a DNA analysis can now be performed to find out whether a person has light or dark eyes.

Facial expressions are also very important. To try to reproduce them, the bio-mechanics and bio-engineering laboratory of the UTC, which is a partner with the chair, studies the facial movements when a person smiles and changes his or her expression. The long-term goal is to be able to animate the reconstructed faces, to make them expressive, and to achieve as real a result as possible.

It’s clear that the field of research opened up by FaciLe is vast – and far from being settled!

Find out more about this team and its work

Read more

A recent study on primate conservation confirms what primatologists have known for a long time: the planet of the apes is on the verge of extinction.

Cranberries, the little red berries from North America, are not effective for curing urinary tract infections. This piece of information is bound to...

visuel

“Let’s support the Albert Einsteins and Edwin Hubbles of the 21st Century” This is the fundraising campaign slogan of Lagrange Institute Paris (ILP),...