Advanced Autism Spectrum Disorder Detection with Hybrid Image Features and Machine Learning Models

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition that affects communication and social behavior. Early detection and timely intervention are critical for ASD detection. This work proposes an improved approach to ASD detection using a combination of state-of-the-art imaging features and state-of-the-art machine learning models. For feature extraction, manual methods such as SURF, Zernike Moments, Wavelet Transform and deep transfer learning models like EfficientNet and MobileNet are used. The extracted features are combined to form a hybrid feature set, which are then classified using advanced ML models including gradient Boosting, random forest, LightGBM and CatBoost. With the fusion approach, the performance is significantly improved for ASD detection. This work highlights the use of novel image features and ML classifiers in the context of ASD detection.