Supplementary file 1_Integrating multi-atlas neuroimaging data for robust biomarker identi

Lijun Zhou published a supplementary PDF on figshare on 2026-03-19. The file describes a multi-atlas graph neural network framework validated on the public ABIDE-I dataset for autism spectrum disorder and a private PTSD dataset with 138 subjects. The model achieved classification AUCs of 82.9% and 89.96% on the respective datasets.

Use Cases

• Training graph neural networks for ASD classification based on multi-atlas functional connectivity data.
• Validating new neuroimaging analysis methods against the ABIDE-I dataset mentioned in the description.
• Identifying neurobiological biomarkers for PTSD based on the interpretability analysis of abnormal brain regions.
• Developing cross-atlas feature fusion techniques for neuroimaging data.

Strengths

• Model performance is quantified with specific metrics: 82.9% AUC on ABIDE-I and 89.96% AUC on a PTSD dataset.
• Validation was performed on a public benchmark (ABIDE-I) and a private dataset with a defined sample size (n=138).
• The file is openly licensed under CC-BY-4.0.

Limitations

• The actual dataset files (e.g., connectivity matrices) are not provided; the PDF is a supplementary description.
• Row count and column-level documentation for the underlying data are absent.
• The private PTSD dataset's full provenance and accessibility are unknown.

Provenance

Source

Lijun Zhou via figshare.

Collection Method

Likely derived from neuroimaging studies, including the public ABIDE-I dataset.

Freshness

Last updated 2026-03-19 06:48:13.