Standardized smartphone photographs revealed measurable facial differences between children and young adults with sickle cell disease and matched controls in an exploratory study published in PLOS Digital Health. The researchers say the results offer a way to study disease-associated facial patterns, but the model has not been tested in an independent population and is not a diagnostic tool.
The analysis used 308 front-facing photographs from 154 participants with electrophoresis-confirmed sickle cell disease (SCD) and 154 age- and sex-matched controls. Participants were aged 5 months to 19 years and recruited in Kinshasa, Democratic Republic of the Congo. The control group was assessed by local clinicians but did not receive confirmatory electrophoresis, a limitation that could affect how distinct the groups appeared.
To quantify facial features, the researchers marked 33 landmarks on each image and calculated measurements of distances and angles, alongside descriptors of local image texture. Their analysis found 14 of 73 prespecified features differed between the groups after statistical adjustment: eight geometric measurements and six texture descriptors. The geometric differences involved areas around the eyes, nose and mouth. The final classification model selected six geometric features; none of the texture descriptors made that final set.
The model’s area under the receiver operating characteristic curve was 0.843, and its accuracy was 0.795. These results came from leave-one-out cross-validation, in which each participant was held out in turn while the model was trained on the rest of this same study sample. They show that the selected measurements carried information that separated the groups within this dataset. They do not establish how well the model would perform on people from other places or in a future clinical setting.
The study also found that several differences between the SCD and control groups were smaller among younger participants and more pronounced in some older age groups. Because the comparisons were cross-sectional—different people were compared at different ages—the results cannot show that facial features changed over time in any individual. The authors note that age patterns could reflect disease exposure, treatment history, differences between cohorts or other factors not measured in the study.
The work’s significance is exploratory: it shows that measurements from standardized photographs can be used to investigate facial morphology associated with SCD in a population underrepresented in digital-health research. Smartphone photography may offer an accessible way to collect such measurements for research. The authors describe the approach as a framework for further study alongside standard clinical procedures, not a replacement for established testing or a standalone clinical tool.
Several questions remain before the findings can be applied more broadly. The study was conducted at one site, and the researchers lacked clinical details such as hemoglobin genotype, anemia severity, transfusion history and other complications. The model also has not undergone external validation. The authors call for larger studies across multiple sites, with repeated imaging over time and more detailed clinical information, to test whether the patterns are reproducible and what they might mean.