» Diagnosis

New software may aid early diagnosis

A new piece of software may help parents of new born babies find a diagnosis of CdLS a lot earlier – even if the diagnosing doctor has never seen a case before.

Genetic syndromes are often associated with characteristic facial traits that practiced clinicians use to make their diagnosis.

Now, for the less experienced, German researchers have designed a computer programme to do it for them.

Dagmar Wieczorek of the Institute for Human Genetics at the Universitaetsklinikum Essen, in Germany, and colleagues have developed a piece of software that recognises facial patterns by comparing a 2-dimensional frontal photograph of a person with a database of stored faces.

The researchers presented their findings at the European Human Genetics conference in Geneva in May.

They fed in 55 photos of people diagnosed with genetic syndromes including Cornelia de Lange, fragile X and Williams-Beuren syndromes.

The photos were first processed using Gabor wavelets, a filter that eliminates lighting differences and transforms the image data in a similar way to that performed by simple cells in the vertebrate primary visual cortex.

Labelled

The face is then represented in terms of a graph labelled with 48 nodes. By comparing the positions of these nodes and the distances between them with those on faces in the database, the programme makes its recognition decision.

The program was correct in around 60% of cases to begin with. But then the researchers refined the software by noting which nodes were used most in making a correct identification.

These tended, not surprisingly, to be on the eyes, nose, mouth, chin and forehead, as opposed to those around the borders of the face and hair.

Successful recognition increased to 76% when they tweaked the program to pay more attention to these information-rich points.

Wieczorek predicts that improvement will continue as more examples are added to the database and the programme gets better at accommodating small variations while still correctly identifying a syndrome.

“It takes a particular skill to extract a disease-specific pattern from a facial appearance that is also influenced by family background and environment,” she said.

“The idea is to help inexperienced clinicians to make more accurate diagnoses.”

Eyebrows

Children diagnosed with Cornelia de Lange Syndrome tend to have thin eyebrows that often meet in the middle, long eyelashes, and a short, upturned nose. Those with Williams-Beuren, on the other hand, have a wide mouth, full lips, small chin and puffiness around the eyes.

As with all human beings, their facial configuration changes with age, and in Williams-Beuren syndrome, the characteristics tend to become more pronounced.

Unlike Wieczorek’s technique, which is unpublished, previous attempts to identify syndromes by facial features - using anthropometric measurements, for instance - have proved less than reliable because they do not account for the face’s overall configuration.

There was considerable interest by the scientific community in the findings and the CdLS Foundation was asked to make statements to organisations such as the BBC who were covering the conference.