Simplifying Diagnosis of Fetal Alcohol Syndrome Using Machine Learning Methods

Area of Science:

• Medical imaging analysis
• Machine learning in diagnostics
• Pediatric dysmorphology

Background:

• Fetal alcohol spectrum disorder (FASD) is a complex condition resulting from prenatal alcohol exposure.
• Fetal alcohol syndrome (FAS), the most severe form of FASD, is characterized by distinct facial features.
• Objective diagnostic tools are needed to improve the identification of FAS.

Purpose of the Study:

• To utilize 3D metric facial data from FAS patients.
• To identify machine learning (ML) methods for objective FAS diagnosis.
• To enhance the diagnostic process for FAS.

Main Methods:

• Analysis of 3D facial scans from 30 children with FAS and 30 controls.
• Evaluation of skeletal, facial, dental, and orthodontic parameters for ML-based diagnosis.
• Testing decision trees, support vector machine, and k-nearest neighbors ML algorithms.

Main Results:

• All tested ML methods achieved a high diagnostic accuracy of 89.5%.
• Top predictive parameters included midfacial length, right palpebral fissure length, and nose breadth at sulcus nasi.
• These facial measurements demonstrated significant value in ML-driven FAS detection.

Conclusions:

• Machine learning, using specific facial parameters, provides an efficient method for objective FAS detection.
• Right palpebral fissure length, midfacial length, and nose breadth at sulcus nasi are key diagnostic indicators.
• Decision trees are recommended as the most practical ML method for clinical application in FAS diagnosis.