Fig. 1

Comprehensive illustration of the methodologies employed for the training of the Brain Age models and the generation of brain-predicted ages. Model Creation shows the steps taken to train the Brain Age model on the Model Creation Dataset and choose the final model applied on the Application Dataset: a Image processing includes Fastsurfer for brain segmentation and extraction of intensity and morphological features, thus building three feature sets: the Morphological Feature Set, Intensity Feature Set and the Combined Feature Set. For each of these feature sets, a feature selection procedure is performed in a 10-fold cross-validation scheme creating feature sets of 20, 30 and 40 features to feed the machine learning models (SVR, RF and MLP) for each fold. b Validation is performed to select the best combination of feature set size and machine learning technique. c Test on the Model Creation Dataset to assess the performance of the Brain Age prediction model. Model Application depicts the use of the chosen model on the patient and healthy groups. Brain Age Gap is calculated as the difference between the predicted and the actual age. Differences in Brain Age Gap are then analyzed