Figure 2.

Predicting progression clusters from cross-sectional data. Performance of machine learning classifiers differentiating between ‘fast’ and ‘slow’ progressors at their time of dementia diagnosis, averaged across 10 repeats. The data set on which the respective performance was evaluated is shown in parenthesis on the x-axis. The application of the common classifier to ADNI represents an external validation. AUC-ROC, area under the receiver operator characteristic curve; AUC-PR, area under the precision-recall curve.