Table 3.
Summary of the optimal combination of features to use for long-term (1 year) predictions of motor, cognitive, and psychological stroke outcomes
| FEATURE | Motor | Cognitive | Emotional | |
|---|---|---|---|---|
| age | 1 | 1 | 0 | |
| gender | 1 | l | 0 | |
| DWI | 0 | 0 | 0 | |
| T1 | 1 | 0 | 0 | |
| lesions | 1 | 1 | 1 | |
| FLAIR | 0 | 1 | 0 | |
| tracts | 1 | 0 | 0 | |
| atlases | 1 | 0 | 0 | |
| discos | 0 | 0 | 1 | |
| clustering | 1 | 1 | 0 | |
| ANNs | 1 | 0 | 1 | |
| regression | 0 | 1 | 1 | |
| feature selectton | 1 | 1 | 0 | |
| dimension reduction | 0 | 1 | 0 | |
| parcellation | 0 | 0 | 0 | |
| cross validation | 0 | 1 | 0 | |
| bootstrapping | 0 | 0 | 0 |
Note: Resulting combinations of features to use for the motor, cognitive, and psychological (emotional) domains. The first column of the table lists the features (inputs and methods) that were investigated, followed by the columns ‘Motor’, ‘Cognitive’, and ‘Emotional’ that represent the domains of the scores the prediction performance of the teams was evaluated on. Gray cells containing ‘1’ indicate which features belong to the optimal, ‘theoretically best’ combination that leads to the best prediction of a given domain. White cells containing ‘0’ represent the features that do not contribute to the optimal feature combination.