Explainable-Spatio-Temporal-Transformers-for-Schizophrenia-Classification-from-Functional-MRI

Read full text
Authors: Tamilarasi Sarveswaran, Vijayarajan Rajangam
Cited by: 0

Abstract

# Explainable-Spatio-Temporal-Transformers-for-Schizophrenia-Classification-from-Functional-MRIThis repository provides Python scripts for classifying schizophrenia patients using functional MRI (fMRI) data with deep learning. The proposed framework combines spatial and temporal modeling to capture both structural abnormalities and dynamic brain activity patterns associated with schizophrenia. A 3D Vision Transformer (3D ViT) is used to learn spatial representations from each fMRI volume, while a temporal Transformer models long-range dependencies across the full fMRI time series to identify functional alterations and temporal dysconnectivity. To improve interpretability, the framework includes Explainable AI (XAI) techniques such as Grad-CAM, attention rollout, and temporal attention visualization, highlighting the brain regions and time points driving classification decisions. Group-level analysis reveals symmetric activation for schizophrenia subjects, consistent with known disruptions in brain networks related to the disorder. ## Data Availability The datasets are from the following open repositories: - **UCLA Consortium for Neuropsychiatric Phenomics LA5c Study database**: [https://openfmri.org/dataset/ds000030/](https://openfmri.org/dataset/ds000030/)- **COBRE dataset**: [https://fcon_1000.projects.nitrc.org/indi/retro/cobre.html](https://fcon_1000.projects.nitrc.org/indi/retro/cobre.html) ## How to Run:The data were preprocessed before further analysis, using standard neuroimaging preprocessing pipelines (SPM).1. **First**,run fMRI classification.py to train the 3D ViT+Temporal transformer based classification model.2. **Next**, run Group level analysis.py to generates voxel importance attention and gradcam maps. ## Author Information: Tamilarasi Sarveswaran* and Vijayarajan Rajangam** *School of Electronics Engineering, Vellore Institute of Technology, Chennai, India ** School of Electronics Engineering, Vellore Institute of Technology, Chennai, India

Fields of Study & Subjects

Functional Brain Connectivity Studies Schizophrenia research and treatment Advanced MRI Techniques and Applications

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
2 February 2026