A new study brings facial recognition software into the clinic.
Fetal alcohol spectrum disorders (FASDs) refers to a range of conditions caused by a mother’s consumption of alcohol during pregnancy.
Alcohol travels through the placenta and is able to damage the growing fetus using a number of different mechanisms. In particular, it affects the development of the baby’s head, face, and brain.
FASDs include fetal alcohol syndrome (FAS) and partial fetal alcohol syndrome (pFAS), as well as alcohol-related neurodevelopmental disorders (ARNDs).
There are well-defined diagnostic criteria for FAS and pFAS. Signs include facial anomalies, a smaller head circumference, growth retardation, and neuropsychological deficits. FAS and pFAS can normally be diagnosed without knowing whether or not the mother consumed alcohol during her pregnancy.
The challenge of diagnosing ARNDs
However, ARNDs have proven more difficult to spot; diagnosing them relies on the doctor knowing whether or not the fetus had been exposed to alcohol.
It does cause some facial abnormalities, but they are much more subtle and indistinct. The primary signs include a variable range of cognitive and behavioral abnormalities. Although certain cognitive tests have been designed to test for ARNDs, they are complex and unreliable.
Because ARNDs often remain undiagnosed for longer, the person is less likely to receive the extra support they need, increasing the risk of problems further down the line, such as trouble at school, alcohol abuse, and mental illness.
Although facial differences in children with ARND are much more subtle than those in FAS, a recent study published in the journal Pediatrics used a novel approach to diagnosis.
Earlier studies showed that computer-aided analysis of facial differences could pick out subclinical features of patients with ARND. However, the systems involved were complex and relied on expensive 3-D cameras that would not be practical in a clinical setting.
The latest study focused on a system that could carry out facial recognition using photos taken with a standard camera.
The study included participants from the Fetal Alcohol Syndrome Epidemiology Research database. Aged 5–9, they came from South Africa, the United States, and Italy and included 36 people with FAS, 31 with pFAS, and 22 with ARND. The study also included a control group of 50 children without FASD.
Each participant was rated by a computer system and two trained dysmorphologists, or experts at recognizing birth defects, who were unaware of the children’s previous diagnoses.
Automated facial analysis was completed by a software tool called Face2Gene, which analyses 2-D photos of faces. This package combines several different techniques to measure a range of angles, lengths, and ratios on faces. These measurements are then statistically analyzed to pull out any dysmorphic features.
How did the software perform?
The computer-aided method was found to be just as accurate as a dysmorphologist at diagnosing FASDs in general. However, the computer performed significantly better than the human clinicians when it came to the more difficult-to-diagnose ARNDs.
The authors conclude, “We found there was an increased diagnostic accuracy for ARND via our computer-aided method.”
“As this category has been historically difficult to diagnose, we believe our experiment demonstrates that facial dysmorphology novel analysis technology can potentially improve ARND diagnosis by introducing a standardized metric for recognizing FASD-associated facial anomalies.”
These findings are important, as FASDs are often undiagnosed or misdiagnosed, with potentially dire ramifications for the child further down the line. As the authors write, “Earlier recognition of these patients will lead to earlier intervention with improved patient outcomes.”
Because the technology under trial involves simple 2-D images rather than 3-D ones, it could be made available to clinicians without special dysmorphology training. This might be of particular importance in developing nations, where relevant experts are few and far between.
Although computer-aided photo analysis cannot diagnose FASDs alone, it may help to improve accuracy and speed of diagnosis. Further trials will be needed, but these initial findings are encouraging.