Table 5 Ablation study results of the Pitt corpus dataset. Values presented are the mean ± standard deviation. Results are averaged over four runs.

From: Alzheimer’s disease recognition using graph neural network by leveraging image-text similarity from vision language model

Architecture

Pr (%)

Rc (%)

F1 (%)

Sp (%)

Ac (%)

Shuffling edge weights

83.99 ± 2.37

83.61 ± 3.28

83.77 ± 2.10

80.10 ± 3.49

82.05 ± 2.27

Independent embedding

81.99 ± 3.15

86.89 ± 2.68

84.31 ± 1.37

76.02 ± 5.86

82.05 ± 1.87

Shuffling & Ind. emb.

81.90 ± 2.21

84.84 ± 2.06

83.30 ± 0.56

76.53 ± 3.91

81.14 ± 0.87

GCN (full)

84.45 ± 2.74

84.02 ± 2.80

84.19 ± 1.67

80.61 ±4.25

82.50 ± 1.87

Max pooling

86.61 ± 4.10

85.66 ± 3.10

86.03 ± 1.11

83.16 ± 6.74

84.55 ± 1.66

Proposed model

86.64 ±2.59

86.89 ±2.32

86.71 ±0.60

83.16 ± 4.21

85.23 ± 0.87

  1. Ac accuracy, F1 F1-score, Pr precision, Rc recal, Sp specificity
Back to article page