Table 4. Cross validation performance for nQi in the Early-PD and De-novo datasets.

Test dataset	Train dataset	Area Under the ROC curve (AUC)
Early-PD	De-novo	0.92
De-novo	Early-PD	0.77
Combined		0.81

The nQi scores for the combined dataset are generated by combining the output of the cross-validation of Early-PD and De-novo dataset, without any further training. The AUC scores allow a reliable comparison of the performance even when the number of PD and Controls is not fully balanced as in the Early-PD and De-novo datasets evaluated independently.