Figure 1.

Data flow diagram for the study: model training iteration. The model was trained and tuned after variable selection on a development set originating from patients of mutiple specialized centers. Firstly the model was internaly evaluated using cross-validation. Then, a test set of couples originating from a different institution was used for an external validation. The model predictions were matched with the expected clinical output. Model training iteration (i.e. introduced externally) process was eventually performed to build a refined model with the selected features.