Table 1.
Prediction performance of four regression models on all samples
| Ridge | SVR | GPR | DNN | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| R squared | MAE | R squared | MAE | R squared | MAE | R squared | MAE | |||||||||
| GMV | 0.678 | (0.069) | 1.662 | (0.156) | 0.640 | (0.102) | 1.699 | (0.188) | 0.673 | (0.069) | 1.649 | (0.153) | 0.581 | (0.096) | 1.837 | (0.199) |
| ReHo | 0.574 | (0.075) | 1.929 | (0.207) | 0.565 | (0.069) | 1.931 | (0.197) | 0.579 | (0.073) | 1.924 | (0.211) | 0.569 | (0.068) | 1.934 | (0.163) |
| ALFF | 0.606 | (0.046) | 1.788 | (0.135) | 0.608 | (0.053) | 1.803 | (0.150) | 0.621 | (0.044) | 1.770 | (0.128) | 0.592 | (0.065) | 1.811 | (0.096) |
| FA | 0.524 | (0.103) | 1.959 | (0.171) | 0.510 | (0.082) | 2.003 | (0.164) | 0.533 | (0.097) | 1.965 | (0.159) | 0.399 | (0.107) | 2.392 | (0.145) |
| MD | 0.608 | (0.048) | 1.829 | (0.150) | 0.587 | (0.079) | 1.949 | (0.113) | 0.621 | (0.053) | 1.845 | (0.158) | 0.474 | (0.071) | 2.177 | (0.241) |
| GMV&DTI | 0.772 | (0.042) | 1.396 | (0.129) | 0.747 | (0.052) | 1.440 | (0.106) | 0.768 | (0.046) | 1.395 | (0.129) | 0.737 | (0.052) | 1.436 | (0.120) |
| GMV&rsfMRI | 0.726 | (0.049) | 1.473 | (0.132) | 0.722 | (0.070) | 1.519 | (0.123) | 0.732 | (0.051) | 1.463 | (0.143) | 0.713 | (0.045) | 1.465 | (0.129) |
| DTI&rsfMRI | 0.715 | (0.052) | 1.550 | (0.157) | 0.706 | (0.061) | 1.520 | (0.144) | 0.722 | (0.053) | 1.512 | (0.147) | 0.694 | (0.047) | 1.520 | (0.157) |
| Multimodal | 0.766 | (0.044) | 1.414 | (0.143) | 0.756 | (0.060) | 1.426 | (0.135) | 0.774 | (0.043) | 1.384 | (0.129) | 0.753 | (0.053) | 1.381 | (0.119) |
Note: Prediction accuracy of age prediction is defined as the r squared of the quadratic fitting for prediction age (brain age) with chronological age and gender. Values in brackets are standard deviation across 10‐fold CV. MAE is the absolute value of the difference between the predicted age and chronological age.