FIGURE 2 Nested cross-validation protocol used in the study. The inner loop (light blue) is used for the model selection, while in the outer loop selected models are trained based on the tuned hyperparameters. The procedure is repeated N times based on the outer loop CV method, resulting in the output (green) of estimated performance. For the outer loop, MCCV was used with pre-defined age groups classification and K-fold CV in the age threshold tests. The inner loop model selection included feature selection hyperparameters selection for the MRMR algorithm used by SVM learners.