Face classification technology as a new diagnostic tool for Cushing's syndrome – preliminary results

Introduction:

Cushing's syndrome (CS) is a rare disease characterized by a quantity of signs and symptoms that show a morphological similarity to the metabolic syndrome. Currently the diagnosis is established with a delay of 4-6 years.

Pilot studies regarding face classification technology in acromegaly and CS have shown promising results. We evaluate the performance of this tool with a larger cohort and controls with morphological similarity.

Methods:

57 (16 m., 41 f.) study subjects and 58 (30 m., 28 f.) controls were included so far. For the control group we screened patients with typical clinical signs (e.g. hypertension, diabetes, metabolic syndrome) but biochemically excluded CS.

We acquired standardized frontal and side-view facial photographs, taken with a regular digital camera. 52 Nodes were then placed on disease-relevant structure of the face and analyzed with the software tool FIDA (facial image diagnostic aid), using a combination of Gabor wavelet transformation and geometry analysis and a maximum likelihood classifier.

Results:

70,4% of female patients and 74% of female controls, 68,8% of male patients and 75% of male controls were correctly classified with CS.

Conclusion:

We found a satisfying classification accuracy of CS in this preliminary analysis. Further studies need to include a larger patient cohort and a modification of the analysis methods.