Table 10.
AUC score comparison on brain datasets with ICA components by using all, top 5% and bottom 5% time-points only. We train and test a logistic regression (LR) model using the time-points identified by DICE and compare the results when using top and bottom 5% time-points. We see that using only top 5% time-points are enough to almost reach the AUC using all time-points.
Method | FBIRN | OASIS | ABIDE | |
---|---|---|---|---|
| ||||
100% | DICE | 0.86 | 0.752 | 0.722 |
Top 5% | LR | 0.85 | 0.743 | 0.706 |
Bottom 5% | LR | 0.566 | 0.548 | 0.532 |