Table 6.
Multiple performance results of our best CNN, in percentage.
| Model | Dataset | Split | Accuracy | TPF | AUC | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| CN | MCI | AD | All | CN | MCI | AD | ||||
| ADNet | Dataset 1 | Train. | 60.6 | 89.6 | 36.7 | 86.8 | 87.9 | 90.3 | 80.6 | 88.8 |
| Val. | 44.1 | 71.1 | 22.4 | 62.4 | 68.9 | 72.2 | 56.9 | 72.5 | ||
| Test | 43.6 | 67.3 | 21.1 | 64.7 | 68.0 | 73.9 | 57.0 | 68.9 | ||
| ADNet | CADD | Train. | 76.7 | 83.3 | 55.6 | 88.9 | 90.3 | 92.1 | 83.1 | 96.3 |
| Test | 51.4 | 77.5 | 27.9 | 46.6 | 68.5 | 70.5 | 61.2 | 73.6 | ||
| AD-DNAet | CADD | Train.* | 76.7 | 75.0 | 55.6 | 100.0 | 88.5 | 90.7 | 79.4 | 95.8 |
| Train. | 90.0 | 83.3 | 88.9 | 100.0 | 98.0 | 95.8 | 97.9 | 100.0 | ||
| Test | 52.3 | 68.2 | 37.7 | 49.5 | 70.9 | 72.8 | 60.5 | 79.0 | ||
AD, Alzheimer's disease; AUC, area under the receiver operating characteristic curve; CN, cognitively normal; CADD, CADDementia; CNN, convolutional neural network; MCI, mild cognitive impairment; Train., training; TPF, true-positive fraction; Val., validation; Train.*, leave-one-out cross-validation results.