Novel TSK Modeling Method with Joint Group Sparse Learning for Autism Aided Diagnosis

Abstract

Autism is a neurodevelopmental disorder with great uncertainty in its diagnosis. However, the existing modeling methods for autism diagnosis have not been effectively studied for the uncertainty of the diagnosis process so far. In this paper, based on TSK (Takagi-Sugeno-Kang) fuzzy system and combining the association information between functional connections, a new sparse modeling method JGSL-TSK (joint-group-sparse-learning Takagi-Sugeno-Kang) for uncertain joint group is proposed and applied to the auxiliary diagnosis of autism. Firstly, the original rs-fMRI (resting-state functional magnetic resonance imaging) data are preprocessed and extracted to obtain the reduced dimension feature matrix. Secondly, based on the TSK fuzzy system framework, the joint sparse regulari-zation term is introduced to the consequent parameter learning process from the correlation between features, so as to guide the joint selection of features within the same rule and between rules. Finally, the alternating optimization method is used to solve the model. Compared with the existing methods, this method has the advantages of strong interpretability and good classification performance. Experimental results show that this method is conducive to the auxiliary diagnosis of autism.

Keywords

• autism spectrum disorder (asd)
• resting-state functional magnetic resonance imaging (rs-fmri)
• tsk fuzzy system
• joint group sparse