Table 5 Ablation study based on various deep learning models vs proposed model using MCCV method.

From: A novel interpreted deep network for Alzheimer’s disease prediction based on inverted self attention and vision transformer

Model

Parameters

Size (MB)

MCCV accuracy

MCCV F1-score

InceptionV3

23.9 M

23.9

95.0% ± 1.2%

91.2% ± 1.4%

ResNet50

25.6 M

96

92.1% ± 0.9%

91.8% ± 1.0%

MobileNetV2

3.5 M

13

89.3% ± 1.7%

88.9% ± 1.8%

VGG19

143.6 M

535

90.7% ± 1.0%

90.2% ± 1.1%

ViT-larg16

307 M

924

93.4% ± 0.8%

93.0% ± 0.9%

Proposed ViT

5.7 M

22.6

98.9% ± 0.5%

98.6% ± 0.6%

Proposed IRBwSA

1.3 M

28.4

96.1% ± 0.7%

95.9% ± 0.8%

Back to article page