Table 1. Evaluation of different classification algorithms using the 5-fold cross-validation method at T=-10% (mean ± std).

^*Feature selection configurations and hyperparameters were determined and optimized based on a subset of subjects in the training set of one fold and then used for all experiments.

^**The method uses two percent of the learned features by the CADNet model for the prediction purpose. The final feature set was selected based on the RFE.EN method.