ESR Position Paper on Imaging Biobanks

Imaging biobanks are defined as organised databases of medical images, and associated imaging biomarkers (radiology and beyond), shared among multiple researchers, linked to other biorepositories [10].

Already existing biobanks are designed to give researchers access to large collections of patient/subject samples and data. Biobanks group biological material of healthy subjects (population-based) and/or patients with specific pathologies (disease-oriented), of which the most frequent are cancer-related. Most biobanks focus only on the collection of genotype data, but do not simultaneously come with a system to collect related clinical or phenotype data. In particular, most biobanks do not include or are not linked to any kind of imaging information, neither primary images nor imaging biomarkers. Comprehensive exploitation of biobanks that also include imaging (genotype and phenotype) is an important cornerstone in diagnostics in the era of “personalised medicine” [10–14].

Personalised medicine describes the intent to provide individual patients with state-of-the-art diagnostic tests, tailored interventions and specific treatments, whenever clinically indicated. Personalised medicine proposes the adjustment/customisation of health care from a one-size-fits-all approach to a patient-specific diagnosis and treatment [15]. As such, personalised medicine can be described as an evidence-supported pre-selection and assignment of tests and therapy selections to patients in need. Quantitative medical imaging, potentially resulting in the discovery of imaging biomarkers, is an essential part of personalised medicine providing a priori selection criteria and a posteriori follow-up strategies, tailored to a given patient with a specific clinical need. Obviously, development of such strategies will be greatly enhanced by the availability of large data repositories [16–19].

Biobanks give researchers access to large repositories of biomaterials for a broad spectrum of further and future analysis, e.g., genetic, genomic, epigenetic, mRNA, proteomics and transcriptomics. The large scale and the broad spectrum of data allow the detection and validation of relevant biomarkers for personalised medicine. Moreover, biobanking in European networks will result in harmonisation of health, lifestyle and other exposure data as well as the development and implemention of harmonised definitions of diseases by increased consensus on the criteria for clinical endpoints.

The classical biobanking activities are to be mirrored by a similar network of imaging biobanks. Modern radiology and nuclear medicine can also provide multiple imaging biomarkers of the same patient, using quantitative data derived from all sources of digital imaging, such as CT, MRI, PET, SPECT, US, x-ray, etc. [20, 21]. Imaging biobanks are infrastructures with massive storage and computing capacity. High-performance computing resources are needed to facilitate image processing comparison, standardisation and validation. Integration of resources and services through a platform that manages the information flow and image processing is a step needed in the development of imaging biobanks.

Other types of images can also be collected from endoscopy, microscopy, surgery, etc., also providing measurable personalised data. All this imaging information should be considered as the phenotypic expression of a patient and can be linked to the genotype. Such data should be available to the research community [22, 23].

A European imaging biobanks network would significantly boost European research in the imaging domain by stimulating the design and validation of new imaging biomarkers, as well as improving our understanding of their biological significance.

This requires standardisation, validation and benchmarking of the data in imaging biobanks. This activity will further stimulate the linking and integration of existing (national and regional) image data repositories as well as the link between imaging biobanks and traditional biobanks.

Standards will have to be developed and implemented. Innovative solutions that promote fair access to high-quality data sets with regard to image-based phenotypes and imaging biomarkers will provide support to users for its utilisation.

Finally, the economic and ethical/legal issues for the management of imaging biobanks have to be explored. These will advance insights and yield benefits to enhance collaborative research, utilise limited resources effectively and share data, technology and expertise. Research on image data management and analysis plays a key role in improving the performance of protocols, software-based analysis and further methodologies for imaging biobanks and the development and validation of imaging biomarkers. All these aspects will surely foster high-level multicenter collaboration.

An imaging biobank is a data repository supporting the gathering, querying and dissemination of imaging data for primary or secondary use in research, education and training. It comprises a repository of images and a database to support indexing based on the organ, modality or pathology, etc.

An imaging biobank may be used for different research purposes, for example, defining and validating new imaging biomarkers for diagnosis or prognosis, by comparing them with existing proven (imaging) biomarkers. It may also be used to identify the genetic origin of a disease or factors (including environmental factors) favouring disease development (cohort studies). The images may also be used to build high-quality anatomical templates, tailored to specific populations, e.g., young subjects between 15 and 20 years or patients with Parkinson's disease above 65 years.

General requirements with respect to the data collection are therefore a database facilitating storage of image data and metadata, storage of derived image-based measurements and storage of associated non-imaging data, taking into account the need to deal with longitudinal data and to cope with multiple file formats (DICOM, of course, but also formats used in research and post-processing settings such as NifTI) [24]. With respect to security, functional and technical requirements should be defined with respect to data transformation (including de-identification and encryption), infrastructure (e.g., user identity management, audit log) and data access and movement (including authorisation and transmission protection) [25]. International collaboration will require the advance of both technological and organisational matters (incorporation, procedures, protocols, data sharing, boards and access criteria). This new environment can be considered as a framework on top of a set of computing and data-intensive infrastructures that will provide researchers with tools, protocols, data and expertise to improve medical imaging research and patient health care.

Imaging biobanks should be focussed on collecting data from healthy subjects or—disease-oriented—from patients (oncologic imaging, rare diseases, other). In the case of healthy subjects (as in the UK Biobank), it should be considered how these data can be collected to ensure participant safety and identity protection. In principle, the imaging acquisition should minimise or avoid the use of ionising radiation with greater dependence on US and MRI in healthy subjects. In contrast, disease-related IB can also be collected by x-ray and (ultra) low-dose CT, represented by data derived from screening programmes for breast cancer (x-ray mammography) and colon and lung cancer (CT colonography and low-dose lung CT) [26–30]. Furthermore, developments in CT imaging with ultra-low-dose acquisitions could open the possibility of CT imaging in population screening for subjects above a certain age.

Oncologic imaging seems to be the easier setting for the collection of imaging biomarkers from multiple imaging modalities. Oncology patients routinely undergo multiple imaging studies for staging and follow-up of the disease to evaluate the TNM and the treatment response. The same patients undergo multiple laboratory tests, pathologic specimen analysis (which provides further imaging data), genetic sequencing, etc. In this setting it seems highly relevant to link imaging biomarkers derived from clinical investigation with those collected by traditional biobanking [31].

Clinical research will benefit from the increased production of scientific researchers' contributions from the field of radiology.

Security and legal issues are two of the most challenging tasks in building large research databases in general. This is due to the requirements for the probable long-term aggregation of medical data from different sources linked with personal information on one side and the request for patient privacy and security aspects on the other. Such databases might contain personal data and information, and also digitised collections of specimens such as tissues or blood, genomic and imaging data.

Many of such research-oriented databases are built as legacy systems with proprietary regulations, regulated by different levels of privacy rules, for example [36, 37].

There are some examples of concepts for the legal frameworks for biobanking; one of these is from the German “Telematics Platform of Medical Research Network” (TMF) [38]. One of the relevant requirements is a solution for interdisciplinary and active collaboration between different institutions, for which the property rights and privacy of the “donors” (information data and/or biological material from volunteers and patients) have to be protected, even for scientific and also possible commercial use by third parties.

The complexity in building large databanks for international collaboration may increase in case of different national regulations. The actual development at the European level with the preparation of a General Data Protection Regulation (COM (2012) 0011) will also be relevant for the future of imaging biobanks [40].

There is an interest in unified worldwide legislation regarding privacy issues and ownership, also to require mandatory registration of biobanks.