NH 15 - Large cohorts: imaging biobanks
NH 15 - Large cohorts: imaging biobanksSaturday, March 4, 14:00 - 15:30 Room: Studio 2017 Session Type: New Horizons Session Topics: Management/Leadership, Evidence-Based Imaging Digital Evaluation: Open Digital Evaluation for this Session Moderators: M. H. Fuchsjäger (Graz/AT), M. Pasterk (Graz/AT) Add session to my schedule In your schedule (remove)
1. To understand why biobanks should include imaging data.
2. To learn about the relevance of incidental findings in large cohorts.
3. To consolidate knowledge on integration and analysis of imaging data in biobanks.
This session will offer insights not only into why biobanks should include imaging data, but also into the technological basis required to integrate imaging data and on how to deal with potential incidental findings in these large cohorts. Furthermore, the potential future benefits - specifically for radiology and radiologists - of including imaging data in large biobanks will be discussed.
1. To understand why inclusion of imaging data is the logical next step of phenotypical characterisation in cohorts.
2. To learn about the importance of enriching clinical datasets in biobanks with imaging data.
3. To learn about biobanks in which imaging data have already been included.
Imaging e-infrastructure is based on radiomics, a new omic science which provides a comprehensive quantification of tissue phenotypes by extracting large number of quantitative image features to higher dimensional data and the subsequent mining of these data for improved decision support. Correlating imaging phenotype with genomic information (referred to as radiogenomics), to better understand genetic variability and the ability to predict prognosis or response to therapy, is a new field of research. It has been initiated mainly in oncology and neuroradiology studies, but it is applicable to other diseases and can be performed with several diagnostic imaging modalities.The overlap between image-based tumour phenotype features and genomic characteristics is not currently well established. Image features are extracted from regions or volumes of interest, which can be either entire tumours or defined sub-volumes within tumours, known as habitats. Recently, the interest is increased in clinical research. Radiomics is potentially useful in the diagnosis of many diseases and can be used in decision support of personalised medicine. Development of “omics” analyses in parallel with growing interests towards translational research has also created a need for quality-assured sample sets with available clinical, imaging, epidemiological and treatment data. Imaging biobanks can be defined as organised databases of medical images and associated imaging biomarkers (radiology and beyond) shared among multiple researchers, and linked to other biorepositories. European Society of Radiology contributes to develop imaging e-infrastructures combined with work on digital pathology.
1. To consolidate knowledge on the significance of incidental findings in large cohorts.
2. To learn about how incidental findings may be detected in large cohorts.
3. To understand the management of incidental findings in large cohorts.
The German National Cohort (GNC) is a long-term, multi-centre population-based cohort study in Germany with the goal of investigating the development of common chronic diseases. 30.000 out of 200.000 participants are being examined by whole body MR imaging at 3 Tesla. Incidental findings (IF) are findings deemed beyond the aims of a study and an expected consequence of imaging studies with potential high impact. By the use of national and international ethical guidelines and the current literature, a system was developed within the GNC to classify and report incidental findings that might be detected on whole body MR imaging and possibly are of risk to the participant’s health. This system focuses on guiding radiologists in the decision of reporting or not-reporting a finding in an attempt to balance the risk of over- and underreporting and to minimize false-positive and false-negative findings. In the centre of that process is a (constantly updated and online available) list of defined incidental findings separating them into reportable and not reportable - taking into account study-specific limitations and confounders. In the talk, the necessary steps to develop such a reporting system are explained, particular challenges and ethical dilemmas are summarized. Quality assurance tools to guarantee high quality and consistency for incidental finding reporting in large multicentre studies are presented.
1. To consolidate knowledge on the technological basis necessary to enable analysis and interpretation of big data.
2. To learn about how imaging data can be integrated in large research databases.
3. To understand the importance of imaging on new biomarker development.
Collection of large datasets, often with many thousands of subjects poses specific problems for data analysis. Although specific analyses might be used by individual investigators on selected subsets this does not make optimal use of the availability of datasets, characterised by extensive imaging and other data derived from soluble and tissue biomarkers and genomic analysis. The analysis of by biobank imaging data must be considered from the inception of the imaging protocols. Automated image analysis techniques must be applicable so that quality control of individual imaging datasets across the database is essential. Selection of acquisition protocols for MRI and other imaging techniques should be designed to facilitate automated analysis with generation of quantitative imaging biomarkers. To facilitate information extraction series of preselected quantitative imaging biomarkers should be identified. These might include relatively standard biomarkers such as regional grey matter volume and CSF volume cardiac ejection fraction. The important features of these analytical techniques is that they should produce a reliable and reproducible qualitative imaging biomarker that can be included in combined biomarker databases for analysis using data extraction and informatics methodologies.