Computer-aided diagnosis of school-aged children with ASD using full frequency bands and enhanced SAE: A multi-institution study

Autism spectrum disorder (ASD) is a neurodevelopmental and network-level disorder mainly diagnosed in children. The aim of the current study was to develop a computer-aided diagnosis method with high accuracy to distinguish school-aged children (5–12 years) with ASD from those typically developing (TD). The current study used multi-institutional functional magnetic resonance imaging (fMRI) datasets of 198 school-aged participants from the Autism Brain Imaging Data Exchange II database and employed enhanced stacked auto-encoders to distinguish between school-aged children with ASD from those TD. In the current study, the average diagnostic accuracy was 96.26% (average sensitivity=98.03%; average specificity=93.62%); these results of classification were higher than that observed in previous studies using single or two frequency bands. The current study demonstrated that the proposed computer-aided diagnosis method may be used to distinguish between school-aged children with ASD from those TD. Attempts to use full frequency bands, deep learning based algorithm and multi-institutional fMRI datasets to distinguish between school-aged children with ASD from TD may be a key step towards clinical auxiliary diagnosis independent of sex, handedness, intellectual level or scanning parameters of fMRI data.