A platform to better predict Alzheimer’s disease

October 17th, 2019

In 2018 Dr Esther Bron, assistant professor of Neuroimage Analysis and Machine Learning at the Erasmus Medical Center (Erasmus MC), was named the winner of the Young eScientist Award. Bron received this prize, which is annually awarded by the Netherlands eScience Center, for her proposal to build a platform for developing and sharing prediction methods for Alzheimer’s disease.

In this interview, Bron talks about the project and her research and shares her experience working with the eScience Center.

What is your main motivation for doing research? And what is the primary goal of the TADPOLE project?

My primary motivation is curiosity ­– I enjoy learning new things and being challenged. But I also find it important to do work that is useful and will eventually benefit other people. This is why I carry out research in the field of medicine. I like the combination of medical problems and technical solutions. At the moment I am developing a new methodology for the analysis of brain MRI scans, using mostly machine learning techniques.

The main goal of my collaboration with the Netherlands eScience Center is to build a platform for sharing prediction algorithms for Alzheimer’s disease with the scientific community. The platform will enable the further development and improvement of these prediction methods as well as their further validation on other datasets of patients with Alzheimer’s disease. The project is called TADPOLE-SHARE and builds on the TADPOLE challenge that I recently organized with fellow researchers from University College London. In this challenge, we objectively compared 62 prediction models developed by 34 international research teams.

What is are your aims with your research and the project? And what are some of the challenges you face?

My research is mostly focused on developing novel MRI analysis and disease prediction methods for Alzheimer’s disease. Currently, no curative treatment for Alzheimer’s disease is available. It is therefore essential to establish an accurate prediction of the disease progression in the early stage of the disease in individual patients. This would enable clinical trials to include the patients in which the treatment is expected to be most effective. The main challenges for accurate diagnosis and prediction are the complexity of disease, the difference in the disease pattern between different patients and the differences in datasets between different hospitals.

How did the project take shape?

I submitted a project proposal to the eScience Center and eventually won the Young eScientist Award 2018. I was pleased to hear that the Center liked my proposal and were keen to collaborate with me. In January 2019, I met with Dr Adriënne Mendrik (eScience Coordinator, ed.) and the EYRA team to start our collaboration and prepare a project plan.

Did the project differ from the original proposal?

The original proposal was quite short, so we had to make the project more specific and develop a detailed execution plan. The project consists of two main components. First, we collaborate with researchers from different countries who developed a prediction method for the TADPOLE challenge in order to rewrite their code and make it available to and usable by other researchers. Second, we work with SURFsara to explore whether we can use the ‘Research Cloud’ that they are currently developing as the platform for our project.

What role does collaboration play in the way you do research?

An increasingly important one. Collaboration with clinical researchers and clinicians is essential for developing a methodology that can eventually be used in clinical practice. Moreover, I believe it is essential to combine different methodologies as a way to produce a better, more effective methodology. This is also one of the reasons why we originally organized the TADPOLE challenge. The challenge is set up as a competition to find the method that can best predict which patients are developing Alzheimer’s disease. The primary aim is to share knowledge between methodology researchers and to objectively compare prediction methods. With the TADPOLE-SHARE project and the collaboration with the eScience Center, I am taking this a step further and making the developed methods available in a standardized and user-friendly way to other researchers.

What has the collaboration with the eScience Center been like so far?

Excellent. The atmosphere at the eScience Center is very welcoming and friendly, and there is a lot of expert knowledge on software development in the context of research projects. For this project, I collaborate with the EYRA team, whose members use the Scrum method to plan and carry out their work. The work is planned in two-week sprints, with some members fully dedicated to working on TADPOLE-SHARE. During the sprints, I usually work at the eScience Center to ease collaboration.

The EYRA team brings essential skills to the project, not only on a programming level but also with respect to the planning and focus of the project and the collaboration with other initiatives such the SURFsara Research Cloud.

What long-term impact on your domain and society do you hope your project will have?

I hope this project will make a twofold contribution. First, by allowing researchers to improve the comparison between the algorithms from the TADPOLE challenge and ultimately deliver better algorithms. And second, by increasing the sustainability and re-use of algorithms, thereby making them available for future research.

I should also mention that our motivation for using the Research Cloud is to create a sustainable platform and a proof-of-concept that can later be adopted by other challenges.

Read more about TADPOLE-SHARE.

Stay up to date, sign up for our newsletter