Experience made using public cloud infrastructure to analyse clinical patient data

Patient data describing operations and results of treatment in clinics is stored in clinical information systems as part of the clinical process. Since these documents are unstructured free-texts stored as scanned documents or as documents prepared with a word processing system there was no automated way to find patterns in these documents that indicate significant accumulations of e.g., treatments and side effects. As a result, valuable information hidden in this huge amount of data spread across clinics is just neglected. A solution is the automated processing of unstructured data from different sources with advanced text mining technology. However, processing a large amount of scanned documents in general exceeds the computational power available in clinics. Using Cloud resources on a p ay per use basis is a cost-efficient alternative to accomplish this task. We describe the approach of the cloud4health project, its framework for processing anonymized patient data, and its data protection and security developments to turn a Cloud into a trusted Cloud where the data may be processed in compliance with legal requirements.