Table 8.
Results of previous deep learning based approaches for predicting MCI-to-AD conversion.
| Study | Number of MCIc/MCInc | Data | Conversion time | Accuracy | AUC |
|---|---|---|---|---|---|
| Li et al., 2015 | 99/56 | MRI + PET | 18 months | 57.4% | – |
| Singh et al., 2017 | 158/178 | PET | – | 72.47% | – |
| Ortiz et al., 2016 | 39/64 | MRI + PET | 24 months | 78% | 82% |
| Suk et al., 2014 | 76/128 | MRI + PET | – | 75.92% | 74.66% |
| Shi et al., 2018 | 99/56 | MRI + PET | 18 months | 78.88% | 80.1% |
| Lu et al., 2018a | 217/409 | MRI + PET | 36 months | 82.93% | – |
| Lu et al., 2018a | 217/409 | MRI | 36 months | 75.44% | – |
| Lu et al., 2018b | 112/409 | PET | – | 82.51% | – |
| This study | 164/100 | MRI | 36 months | 81.4% | 87.8% |
MCIc means MCI converters. MCInc means MCI non-converters. Different subjects and modalities of data are used in these approaches. All the criteria are copied from the original literatures. Bold values indicate the best performance in each column.