Table 2 Averaged binary classification accuracy results for the experiments with VGG16, ResNet-152 and DenseNet-121.

From: Automatic dementia screening and scoring by applying deep learning on clock-drawing tests

Index

Data distribution

VGG16

ResNet-152

DenseNet-121

1

randomly

0.9055

0.9405

0.9601

2

t-SNE

0.8577

0.8552

0.9521

3

PCA

0.9102

0.9422

0.9640

4

LLE

0.9359

0.9490

0.9665

Back to article page