Classifying ASD based on time-series fMRI using spatial–temporal transformer

• Published in: Computers in biology and medicine Vol. 151; no. Pt B; p. 106320
• Main Authors: Deng, Xin, Zhang, Jiahao, Liu, Rui, Liu, Ke
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.12.2022; Elsevier Limited
• Subjects: ABIDE; Adversarial Generation Network(GAN); Autism; Autism spectrum disorder (ASD); Autism Spectrum Disorder - diagnostic imaging; Biomarkers; Brain - diagnostic imaging; Datasets; Deep learning; Deep learning(DL); Diagnosis; Endoscopy; Functional magnetic resonance imaging; Functional magnetic resonance imaging (fMRI); Humans; Internal Medicine; Machine learning; Magnetic resonance imaging; Magnetic Resonance Imaging - methods; Medical imaging; Neuroimaging; Other; Spatial discrimination learning; Time Factors; Transformer; Transformers; Deep learning(DL); ABIDE; Transformer; Functional magnetic resonance imaging (fMRI); Autism spectrum disorder (ASD); Adversarial Generation Network(GAN)
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2022.106320

As the prevalence of autism spectrum disorder (ASD) increases globally, more and more patients need to receive timely diagnosis and treatment to alleviate their suffering. However, the current diagnosis method of ASD still adopts the subjective symptom-based criteria through clinical observation, which is time-consuming and costly. In recent years, functional magnetic resonance imaging (fMRI) neuroimaging techniques have emerged to facilitate the identification of potential biomarkers for diagnosing ASD. In this study, we developed a deep learning framework named spatial–temporal Transformer (ST-Transformer) to distinguish ASD subjects from typical controls based on fMRI data. Specifically, a linear spatial–temporal multi-headed attention unit is proposed to obtain the spatial and temporal representation of fMRI data. Moreover, a Gaussian GAN-based data balancing method is introduced to solve the data unbalance problem in real-world ASD datasets for subtype ASD diagnosis. Our proposed ST-Transformer is evaluated on a large cohort of subjects from two independent datasets (ABIDE I and ABIDE II) and achieves robust accuracies of 71.0% and 70.6%, respectively. Compared with state-of-the-art methods, our results demonstrate competitive performance in ASD diagnosis. •LSTMA is designed to learn spatio-temporal features and accelerate model training.•The proposed GGDB addresses the data imbalance problem in ASD subtype diagnosis.•Compared with SOTA methods, our model shows the robust performance in ASD diagnosis.