3D Facial Features in Neuro Fuzzy Model for Predictive Grading of Childhood Autism

Abstract

Autism Spectrum Disorder (ASD) is a clinically heterogeneous neurological developmental disorder. It is called a spectrum disorder because of its range of symptoms. Early diagnosis and proper intervention is required for the effective treatment of autism. Diagnosis is based on the quantitative and qualitative analysis made by the clinician. The expertise of the clinician is so important in the proper diagnosis and classification of autism. This paper proposes an Expert system that act as a support system to the clinician. Major clinical attributes of autism along with facial features are used as input to the expert system. The main highlight is the use of feautures from 3D facial imagery for autism classification. The expert system operates in two modes, diagnosis mode and grading mode. Naïve Bayes classifier is initially used for diagnosis mode where as overall system is implemented using a Neuro-Fuzzy approach. In the diagnosis mode 100% accuracy and in classification mode 98.8% accuracy is obtained.