blog

Jana Hutter is an open data convert and a member of the Early Career Researcher Advisory Board for Wellcome Open Research. Despite all the barriers Jana had to overcome, she now sees open data as the best way forward for research. Here’s her story.

When the first researcher asked at a conference ‘Can you share some of your new datasets? little did I know about the journey ahead – unexpected barriers, scrutiny, crowdsourcing of ideas, surprises and amazing opportunities!

As part of my Wellcome fellowship, working on new Magnetic Resonance Imaging (MRI) techniques to study the human placenta in-vivo, I developed a technique, called ZEBRA, which combines two MRI techniques, Diffusion MRI and Relaxometry MRI, in a more flexible way. Data acquired with this technique started to attract some interest after presentations at conferences in my field.

Luckily, one of the possible constraints to sharing, in terms of unwillingness to share, was never the issue as everyone involved was very supportive. As a physicist working on new acquisition techniques myself, the idea of having other people working on new analysis techniques for my datasets seemed like a real jackpot!

Main barriers and requirements

After initial naive thoughts of “just acquiring some ZEBRA brain data and putting it on dropbox”, I quickly learnt that there are quite substantial challenges and hurdles to overcome first in order to share my data openly.

Funding

Sharing the data openly leads to different requirements, compared with just acquiring the data for my own, focused experiments. It means optimizing the data parameters,  especially for supporting data (for MRI for example this is high resolution anatomical data to allow registration and supporting calibration data). In my case, this required a number of hours on the MRI scanners with phantoms and volunteers, which could not be achieved without additional funding. Luckily, as Wellcome-funded researcher, I was able to apply for, and secure, Wellcome Enrichment funding for “Open Science” which covered the scans, travel costs to invite researchers in the unique position of being able to advise and help me and to contribute towards ways to advertise the data and thus its use. Going forward, I will always plan in these costs in future projects. 

Ethics and approvals

Once I had the funding in place, another barrier – the biggest and most frustrating – became apparent. My ‘data’ consists of MRI images of volunteers and patients, acquired in a close collaboration between my academic institution and an NHS trust. Protecting the volunteers and patients is the top priority and in line with the FAIR principles “research data should be as open as possible, but as closed as necessary”.

The ethical approvals in place for my project until then did not allow me to share any of the datasets, as the informed consent process did not seek explicit consent from the participants. This is the part where time and a support network of experts is key! In my case, our clinical trial coordinator and governance expert were instrumental in getting the required ethics amendment. 

This involves  general multiple steps – new, transparent consent documents, contracts and transfer agreements between the NHS trust and the academic institution regarding data ownership, licenses, complete data descriptions and safety measures, open access agreement between the academic institution and the future ‘data users’, flow charts illustrating the data flow and anonymization processes to minimise the risk of exposing confidential information – each also requiring communication with experts on all sides. The real work acquiring data can only start after these are all in-place and approvals from all sides are obtained – in our case this process took about 6 months. 

‘Open’ data in the context of medical imaging data can still not be fully ‘open’. In our case, we have a two-step process in place with interested researchers required  to fill in a form declaring their academic institution, intended use and to read and sign an open data access document (detailing restrictions such as no further sharing allowed, no attempt made to de-anonymize the data) which we then check to release the data in the second step.

Having had this experience, I would never apply for new ethics involving patients/volunteers without considering these numerous requirements – it is such a shame that the data acquired previously cannot be shared!

Advertisement and evaluation

A proper plan in place to advertise the data to relevant researchers and then to evaluate its use is very important. I found this out the hard way by the initial rejection of my ‘Open Science’ proposal and by being forced to think about it had a deep and positive influence and made me define my objectives, targets and evaluation aims in structured form. 

Realizing the need to target research communities outside of my expertise, I proactively looked for opportunities and was able to join the organization committee for a MICCAI workshop and the associated MUDI challenge – a perfect fit! For the evaluation, recording all information about the data users and generated outcomes for future evaluation is crucial from the beginning on.

Time requirement

The mentioned additional requirements and need to make the data suitable for a range of analysis techniques, led to a long ‘optimization’ phase with 30 phantom and 20 healthy volunteer MRI scans. Then, thanks to the help through the Open Science award funding and the challenge, experts for each step of the pre-processing got involved constructing a pipeline between two London sites, EPFL in Switzerland and Harvard Medical School. In parallel, the first participants promised the datasets for the challenge rightly started getting nervous – a very stressful time! Increasing interest resulted in a growing time commitment both for the outlined two-step process requiring to manually check and archive each potential ‘data user’ form and for actual support for the data users. All in all, I spent around four months full-time working on getting the data ready and communicating with the participants for this challenge.

Scrutiny and availability

We decided to share the data at multiple steps along the processing pipeline, from very early to fully pre-processed. But the level of scrutiny associated with this process was something I neither expected or was prepared for! Having the data out there in this transparent way allows obviously much deeper insights into the data and acquisition technique then a publication ever would!

Benefits

But this high level of scrutiny also came with amazing opportunities. Allowing for such comprehensive insights meant that a number of areas for improvement (and some errors in the metadata) were spotted, amazing ideas brought forward and suggestions sent to me – even in the form of programming code to directly improve a particular step in the sampling! Together with the achieved data processing pipeline as described above, I noticed a clear and positive impact on my own research – crowdsourcing truly at its best!

The actual workshop in October 2019 in Shenzhen where the challenge results were presented, was an amazing opportunity – so cool to see what people were able to do with the data provided! My work being picked up and analysed with methods I barely understand felt very empowering. Other resulting benefits are co-authorships, three invitations to contribute to special editions, speaker invitations and new concrete collaborations. Until today more than 40 people applied and got access to the data all over the world.

Take home messages

After these great experiences – and lessons learnt regarding ethical approval – I am committed and convinced more than ever, that openly sharing my data is the best possible way forward for my research. Planning ahead, defining targets, seeking opportunities and help as early as possible, especially for the required ethical approval, are essential to make data sharing in medical imaging in this way a success.

Do you have a data story? We interested to hear your experiences. So, share your tales with us via email info@wellcomeopenresearch.org or on Twitter, by using the Hashtag #ECROpenData and tag @WellcomeOpenRes. By sharing it with the community allows us all to learn from it and figure out how we can best progress. Find out more about the campaign viahttps://bit.ly/GrowingDataSharingCulture