From Data4lifesciences to Health-RI

In the past four years, the Data4lifesciences community has strived to facilitate research data sharing between the Dutch UMCs. In 2020, Data4lifesciences will broaden its horizon by merging with the national health research infrastructure initiative ‘Health-RI’. Jan-Willem Boiten and Wiro Niessen explain the why and how of the merger.

Dr Jan-Willem Boiten has acted as the programme manager of Data4lifesciences since its start in 2015. He also is a member of the managing board of Health-RI. “While Data4lifesciences has worked on an interconnected research data infrastructure at the Dutch UMCs, Health-RI aims to achieve this at a larger scale. Health-RI’s true north is a state-of-the-art national infrastructure for data, samples and images to facilitate personalised medicine and health research. The initiative was founded by the NFU, BBMRI-NL, DTL, EATRIS-NL, ELIXIR-NL, Health~Holland, and Lygature, and is now supported by a large number of organisations in the health domain. Health-RI is connecting diverse collections of health and biomedical data, empowering researchers to develop personalised medicine and health solutions,” says Boiten. 

Machine learning 

Sharing data at a national level will allow biomedical researchers as well as medical professionals to reap the benefits of the latest technological developments, all to the benefit of patients and healthy citizens. One important example of a technological development is the progress in machine learning. This is the expertise of Wiro Niessen, a professor in biomedical image analysis at Erasmus MC Rotterdam and Delft University of Technology. He also is Chief Technology Officer at Health-RI. Niessen: “Machine learning is a branch of artificial intelligence where computers (i.e., machines) are trained to perform certain tasks (i.e., learning) based on example data. This technology is already applied extensively in sectors such as the travel business, banking world, and logistics. One of the domains where it will have a major impact is in personalised medicine and health research. Machine learning based on health data holds great promises to facilitate prevention and to improve diagnostics and prognostics in healthcare.” 

Prostate cancer 

“An example from my own field is prostate cancer. Many prostate cancers are slow growing and do not require immediate treatment. To discern these innocuous cancers from clinically significant cancers, medical specialists may use magnetic resonance imaging (MRI). Based on MRI scans, they decide if a prostate biopsy – which is an invasive procedure - needs to be taken for further examination. Unfortunately, it is very challenging to interpret MRI data and misclassifications are common. Therefore, scientists are exploring machine learning as an approach to improve clinical decision making. By retrospectively assembling a large number of prostate images with known associated clinical outcomes and other patient characteristics, an algorithm is created to predict which patients need biopsies based on image characteristics. Preliminary studies suggest that such machine-assisted image analysis can support more accurate clinical decision making. However, we need high-quality data from a very large number of patients to create a strong algorithm that produces reliable predictions and can be applied widely. This is why we want to combine data from as many sources as possible,” says Niessen. 

Hurdles 

Boiten continues: “Of course, health data are not exclusively produced within the walls of the UMCs. Researchers at science faculties, independent biomedical research institutes, and companies are also producing valuable research data; non-academic hospitals are producing clinical data; and individual citizens are creating health data with wearables or smartphones. If we could combine all these data sources, we could make immense progress in personalising healthcare. However, it is quite a challenge to reuse and combine these data sources. In addition to concerns about people’s privacy, consent and data de-identification, there are many practical hurdles.” 

FAIR and federated 

Experts from within and outside the Data4lifesciences community have developed techniques to overcome these hurdles. They have created technical solutions to harmonise datasets and to integrally analyse data in a secure and sophisticated way. Crucial to these approaches are the FAIR Principles, which state that research data should be Findable, Accessible, Interoperable, and Reusable for both humans and computers. Another crucial element is federated data analysis, also referred to as ‘distributed data analysis’ or the ‘Personal Health Train’. In this approach, technical solutions allow researchers to perform an integral analysis on multiple datasets without actually transferring the data. Hence, the data remain at the location where they were originally produced. 

Boiten: “Several expert groups - from both within and outside the Data4lifesciences community - have developed solutions for FAIRification and federated data analysis. In Health-RI, we aim to ensure that these groups work synergistically and that their solutions are easily accessible to researchers. We will do this by organising stakeholder meetings, launching new projects, and organising joint software development sessions between related projects.” 

Community 

Boiten: “Building an ambitious community has been one of the most highly valued achievements of Data4lifesciences. This community will definitely continue its efforts within Health-RI, connecting with additional stakeholders in the health domain. We will strive to create easy access to existing services and resources that assist researchers in sharing and analysing research data, samples, procedures, and facilities. To this end, we are building a Health-RI portal, a first version of which has recently been released. In addition to Data4lifesciences, the Parelsnoer Institute and TraIT (Translational Research IT) will also merge into Health-RI. This will reduce the number of initiatives that are striving to improve the Dutch health research data infrastructure, creating more transparency and efficiency in this field.” 

Collective voice 

Health-RI also serves as a collective voice to advocate the need for a national health research infrastructure. “If we want the Netherlands to be at the forefront in personalised medicine and health, and exploit the enormous potential of artificial intelligence techniques like machine learning, we need coordinated action and investments now. This will easily pay off because it will give an immense boost to Dutch life science research and substantially improve the efficiency and quality of our healthcare,” concludes Niessen.  

Share this page…