(我得说这个模型训练好像鬼打墙）code available at ‣
Sample size: Resilient (n = 547),Vulnerable (n = 1033) Features: rs-fMRI series (cnt = 169), sMRI series (cnt = 213, including area, volume, thickness
Snap shots of distribution between groups, for all the features involved, please check the attached file.
| ROI | vulnerable(mean,sd) | resilient(mean,sd) | p |
|---|---|---|---|
| rsfmri_c_ngd_ad_ngd_ad | 0.31 (0.08) | 0.30 (0.08) | 0.055 |
| rsfmri_c_ngd_ad_ngd_cgc | 0.17 (0.06) | 0.16 (0.06) | <0.001 |
| rsfmri_c_ngd_ad_ngd_ca | -0.06 (0.08) | -0.06 (0.08) | 0.974 |
| rsfmri_c_ngd_ad_ngd_dt | -0.04 (0.05) | -0.05 (0.05) | 0.392 |
| rsfmri_c_ngd_ad_ngd_dla | -0.02 (0.05) | -0.02 (0.05) | 0.773 |
| rsfmri_c_ngd_ad_ngd_fo | -0.04 (0.05) | -0.04 (0.05) | <0.001 |
| rsfmri_c_ngd_ad_ngd_n | 0.00 (0.03) | 0.00 (0.03) | 0.902 |
| rsfmri_c_ngd_ad_ngd_rspltp | -0.04 (0.08) | -0.03 (0.08) | 0.095 |
| rsfmri_c_ngd_ad_ngd_smh | 0.12 (0.07) | 0.13 (0.07) | 0.235 |
| rsfmri_c_ngd_ad_ngd_smm | 0.21 (0.08) | 0.21 (0.08) | 0.211 |
| ROI | vulnerable(mean,sd) | resilient(mean,sd) | p |
|---|---|---|---|
| smri_vol_cdk_banksstslh | 3042.53 (596.87) | 3044.45 (599.37) | 0.952 |
| smri_vol_cdk_cdacatelh | 2124.96 (545.05) | 2112.32 (553.57) | 0.664 |
| smri_vol_cdk_cdmdfrlh | 8162.32 (1420.94) | 8039.41 (1459.45) | 0.108 |
| smri_vol_cdk_cuneuslh | 3731.19 (650.05) | 3657.85 (598.28) | 0.025 |
| smri_vol_cdk_ehinallh | 1881.58 (429.14) | 1837.05 (408.45) | 0.043 |
| smri_vol_cdk_fusiformlh | 11148.18 (1510.13) | 10958.27 (1387.52) | 0.012 |
| smri_vol_cdk_ifpllh | 15620.68 (2342.50) | 15631.85 (2206.47) | 0.925 |
| smri_vol_cdk_iftmlh | 13486.25 (2021.12) | 13335.75 (1937.62) | 0.148 |
| smri_vol_cdk_ihcatelh | 3225.57 (534.45) | 3200.43 (497.13) | 0.352 |
| smri_vol_cdk_locclh | 14318.58 (1999.88) | 14057.23 (1993.61) | 0.013 |
Unbalanced Data, SMOTE
Train: Test, 7:3
Repeated CV, 5-fold cross validation with 3 times of repeat.
Models, Elastic Net (EN), Random Forrest (RF), Support Vector Machine (SVM).
Using RF as an example of model performance on each test-fold. For all 3 models please check the attached file.
| ROC | Sens | Spec | Resample |
|---|---|---|---|
| 0.577586 | 0.407895 | 0.710345 | Fold3.Rep3 |
| 0.610029 | 0.493506 | 0.694444 | Fold4.Rep2 |
| 0.556471 | 0.428571 | 0.717241 | Fold4.Rep3 |
| 0.584684 | 0.376623 | 0.731034 | Fold5.Rep1 |
| 0.500363 | 0.342105 | 0.655172 | Fold5.Rep2 |
| 0.543126 | 0.337662 | 0.703448 | Fold2.Rep2 |
| 0.524407 | 0.363636 | 0.682759 | Fold2.Rep3 |
| 0.537879 | 0.311688 | 0.6875 | Fold1.Rep1 |
| 0.538737 | 0.350649 | 0.682759 | Fold1.Rep2 |
| 0.594572 | 0.381579 | 0.722222 | Fold5.Rep3 |
| 0.523866 | 0.381579 | 0.682759 | Fold3.Rep2 |
| 0.543557 | 0.368421 | 0.717241 | Fold3.Rep1 |
| 0.613249 | 0.447368 | 0.737931 | Fold2.Rep1 |
| 0.562024 | 0.402597 | 0.717241 | Fold1.Rep3 |
| 0.500761 | 0.350649 | 0.641379 | Fold4.Rep1 |
The results from test set.