Ablation Experiment Results for MDD Diagnosis via fMRI Graph Neural Networks

Results from an ablation study for a graph contrastive learning framework (EC-GCL) applied to resting-state fMRI data from 1,160 participants, including 597 patients with major depressive disorder and 563 healthy controls. The dataset, authored by Dan Long and last updated on 2026-05-05, contains results that achieved a diagnostic classification AUC of 71.2%. The study identified key brain regions linked to MDD pathophysiology, such as the dorsolateral superior frontal gyrus, thalamus, and insula.

Use Cases

• Benchmarking graph neural network models for MDD diagnosis based on functional connectivity matrices.
• Analyzing the impact of specific brain regions on model performance based on the identified key regions like the dorsolateral superior frontal gyrus.
• Investigating the utility of edge convolution and contrastive learning for preserving spatial specificity in neuroimaging data.
• Reproducing or extending the ablation experiments for the proposed EC-GCL framework.

Strengths

• Includes data from a substantial cohort of 1,160 participants (597 MDD patients, 563 controls).
• Model performance is quantified with a reported AUC of 71.2%.
• Results are linked to specific, interpretable brain regions consistent with prior neuroimaging studies.

Limitations

• Dataset is very small at 5.5 KB, indicating limited scope, likely containing only summary results.
• Row count and column-level documentation are unknown; field semantics must be inferred after download.
• Description metadata is limited; actual data quality and completeness require manual inspection.

Provenance

Source

figshare, author Dan Long.

Collection Method

Derived from resting-state fMRI data analyzed with a proposed Graph Contrastive Learning based on Edge Convolution (EC-GCL) framework.

Freshness

Last updated 2026-05-05 17:43:05; freshness should be verified.