EnCR_XGBoost: Ensemble Convolutional Recurrent Based XGBoost for Autism Detection Using MRI
Jagadesh Balasubramani, R Surendran
Abstract
The Autism detection is crucial for enhancing the social behavior and interaction of the people because the autism is a complex neurodevelopmental disorder. The deep learning based autism detection is employed for obtaining the accurate detection of disease. Still, inaccurate detection, computation complexity and over-fitting issues limit the performance. Thus, deep learning based autism detection is introduced in this research for solving the issues faced by the existing models. The input obtained from the MRI image is pre-processed using the median filter to enhance the quality of the image. Then, the appropriate features are extracted using proposed Ensemble Convolutional Recurrent (EnCR) model designed with the ensemble of convolutional and recurrent model for extracting both spatial and temporal features for enhancing the detection accuracy. Finally, XGBoost method is utilized for performing the autism detection task using the EnCR based features. The designed EnCR+XGBoost model is evaluated based on Accuracy, Precision, Recall, and F-score and obtained the values of 98.1500,98.4320,98.0320, and 98.2316 respectively.