Table 1 Comparison of A-GHN model with previous state-of-the-art models. Comparison is done by computing the Pearson correlation between the ground-truth FC and predicted FC of test subjects. Overall, the A-GHN model displays a higher correlation value of 0.788, better than previous models. Significant values are in bold.

From: Attention-based fusion of multiple graphheat networks for structural to functional brain mapping

Model

Correlation

MKL7

0.645

GCN Encoder Decoder14

0.732

Autoencoder27

0.561

Macroscale SC-FC38

0.501

GNN (GCN + GTN)15

0.715

M-GHN40

0.741

Random A-GHN (Freezing GHNs)

0.557

A-GHN (one-layer)

0.788

A-GHN (two-layer)

0.799

A-GHN (four-layer)

0.760

Back to article page