Solving Data Integration Problems in Medical Imaging System

Abstract

This research is dedicated to creating a data transformation and unification layer between a number of medical systems (which are essentially lab equipment such as CT, MRI, X-rays scanners and so on) and centralized data management system called PACS (picture archiving and communication system). Source data are present in different (even proprietary) formats and can't be fed to PACS as is.Another challenge is that the data provided by source medical information systems contain personal user information, so it needs to be anonymized prior to analysis. Authors of this paper created a methodology and a set of services to depersonalize, process, store and retrieve medical user data. The system tracks global user identifiers in depersonalized form, so it has the ability to group different samples related to the same user.The system retrieves the data from heterogeneous sources (namely, computer tomography storage, magnetic resonance imaging storage, X-rays storage, angiography storage and ultrasonic investigation storage), processes it using a predefined set of rules and transfers the data to the PACS. PACS allows computer scientists and engineers to perform research on depersonalised data and to generate a set of results. These results are stored in both depersonalized and personalized forms. They are fed to a special service which enriches them with personal user information and stores them in the medical system for a doctor to observe. Also, they are stored in the depersonalized form and can be presented to a general audience. A doctor validates the results of the research.Our research group is in charge of creating the new PACS service which can act as a proxy between the aforementioned medical systems and the old PACS. We are also developing our own DICOM storage (DICOM is an industrial standard of medical imaging information representation) as a backend to our PACS.