Table 5 Contribution of Parallel Learning Elements in MedFusionNet

From: Hybrid deep learning framework MedFusionNet assists multilabel biomedical risk stratification from imaging and tabular data

Model Variant

AUC

F1

Recall (Minority)

Δ vs Full

Full MedFusionNet

0.945

0.91

0.89

-

CNN only

0.882

0.80

0.72

−0.063

ViT only

0.901

0.83

0.76

−0.044

TCN only

0.855

0.78

0.70

−0.090

CNN + ViT

0.932

0.88

0.85

−0.013

CNN + TCN

0.910

0.85

0.82

−0.035

ViT + TCN

0.918

0.86

0.83

−0.027

Back to article page