Fig. 2.

ML approaches exemplified by supervised learning in the form of an ANN. The network is trained to predict outputs from a set of input data by changing the weights of the network depending on differences between model output and data. Input data typically consists of patient characteristics, such as images and medications. Output data often consists of outcome, e.g. whether the patient has suffered a stroke or not within a specific time period. If the ANN is used for imputation, the output data instead consists of a patient variable.