Redacted Patient Manager

The Redacted Patient Data Manager is the central component receiving and processing the patient data from the Healthcare Information Systems. It handles both the patient data submitted to train the AI predictive models as well as those for which AI predictions and simulations shall be applied. All submitted training data are forwarded to the Edge AI Models Manager, from which it obtains trained AI models to impute missing values in the patient data sets. Furthermore, the component includes the tools to handle extreme or unlikely values, so-called outliers, which can result from erroneous manual input, measurement errors, or real outliers. These outliers are filtered before passing on the data. Finally, as one of the privacy-enhancing technologies (cf. Section 7.4), it includes the differential privacy component, which introduces a small and controlled amount of noise in the data to prevent the identification of patients from trained models. Following that, the patient datasets are also homomorphically encrypted and forwarded in an encrypted format to the ASCAPE Cloud to be included in the dataset of encrypted patients from all ASCAPE edge nodes connected to the ASCAPE Cloud.