Modular data management and image processing approaches to facilitate medical research

Digitalisation of medical research and diagnosis is becoming more important because it allows us to offer the best possible treatment to patients exploiting technological advances. New medical imaging methods and digital technologies, such as microfocus X-ray computed tomography (µCT) scanning, are opening up a new world for medical researchers. Due to continuous improvements in imaging techniques, the produced images have grown in size in proportion to improvements in computation. This means that, in order to use cutting-edge imaging techniques, equivalent cutting-edge computing hardware and software are required. Such hardware is expensive and the required software is not always available or incomplete with regards to features needed. This thesis will address the problem of extra-large images in three of the main areas of image-based research: management, processing and visualisation of images. The problems involved with each of these fields, arising because of the large image files, are identified and tackled at the example of a medical research project. The management of images needs to provide an established and easy-to-use interface for both users and independent software utilized by the users. We suggest the use of a network file store for data access. We also suggest a folksonomy for metadata management including illustrative visual tools to search through data, metadata and dataset relations. Processing of the stored data requires a lot of random access memory (RAM). We show that the splitting of image data into smaller blocks helps in the application of sophisticated image processing algorithms along two or three dimensions at the example of lung biopsy scans. We also analyse methods to visualise three-dimensional (3D) objects to medical researchers. We provide a tool for fast visualisation, ideal for previewing of data, and also explore new technologies to enable presentation of 3D features including the use of 3D printing for visualisation. Finally, these approaches are shown as part of an end-to-end workflow that handles the management, processing and visualisation of images. We present an exemplary scenario showing the system in use. The workflow is highly modular and uses standard protocols and programming interfaces, which makes it easier to extend than existing workflows. Medical researchers require new approaches for handling their image workflow to further their studies. Adaptation of existing management techniques from other domains, targeted modification of the data structure, and the use of modern manufacturing and visualisation techniques can enable this without requiring new or modified software