Fig. 1

REST platform: the first step consists of the data acquisition, i.e., nasal airflow, chest and abdominal band, pulse, and oxygen saturation signals on paediatric patients (a). Then the signals are preprocessed, i.e., signal autoscaling and windowing (b). Signals are then fed in three monodimensional convolutional neural network for the simultaneous detection of apnoea, desaturations, and artefacts events (c). Lastly, an explanation of the decision process is given, highlighting the portions of the input signals that gave the main contribution to the final outcomes (d).