Table 1 Network performance in different settings (mean ± standard deviation, %)
From: Multimodal diagnosis model of Alzheimer’s disease based on improved Transformer
Methods | ACC | PRE | SPE | SEN | F1S | AUC |
|---|---|---|---|---|---|---|
Proposed method | 98.10 ± 2.46 | 99.09 ± 2.87 | 96.75 ± 5.28 | 95.82 ± 5.03 | 97.81 ± 2.87 | 98.35 ± 2.14 |
Only sMRI | 91.91 ± 5.96 | 91.05 ± 10.49 | 92.72 ± 8.67 | 90.91 ± 9.41 | 90.31 ± 6.16 | 91.33 ± 5.99 |
Only PET | 87.14 ± 7.12 | 85.4 ± 10.31 | 88.62 ± 8.59 | 85.05 ± 15.61 | 83.99 ± 9.80 | 87.43 ± 6.61 |
Typical Transformer | 94.32 ± 4.01 | 93.57 ± 7.76 | 93.33 ± 6.34 | 92.25 ± 7.36 | 93.40 ± 4.25 | 95.05 ± 3.11 |
Without Transformer | 92.86 ± 7.12 | 90.60 ± 10.31 | 92.20 ± 8.59 | 92.03 ± 15.61 | 90.78 ± 9.80 | 93.11 ± 6.62 |
Without 3DCNN | 90.01 ± 3.52 | 83.44 ± 8.77 | 89.46 ± 5.05 | 93.02 ± 7.59 | 87.45 ± 4.22 | 90.06 ± 3.48 |