Aasa Feragen og Eike Petersen fra DTU Compute - og  Melanie Ganz fra Københavns Universitet med online. Foto: Hanne Kokkegård

A research project offered new opportunities for students

Thursday 23 Feb 23


Aasa Feragen
DTU Compute
+45 26 22 04 98


Eike Willi Petersen
DTU Compute


International Conference on Medical Image Computing and Computer-Assisted Intervention: MICCAI 2022.

Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 pp 88–98 Cite as Feature Robustness and Sex Differences in Medical Imaging: A Case Study in MRI-Based Alzheimer’s Disease Detection, for the Alzheimer’s Disease Neuroimaging Initiative:

  • Eike Petersen, 
  • Aasa Feragen, 
  • Maria Luise da Costa Zemsch, 
  • Anders Henriksen, 
  • Oskar Eiler Wiese Christensen & 
  • Melanie Ganz 
When DTU Compute urgently needed help sorting a dataset, students stepped in. The work resulted in a scientific article and vital knowledge for the students.

Photo: Aasa Feragen (left), Eike Petersen from Visual Computing at DTU Compute, and Melanie Ganz from University of Copenhagen. The students have 'left the building'.

It is quite unusual for bachelor's and master's students to be involved in specific research projects and even contribute to publishing scientific articles. However, this is exactly what happened to the two DTU students Oskar Eiler Wiese Christensen and Anders Henriksen when they were bachelor's students at the study Artificial Intelligence and Data at DTU Compute and to former Master's student, Maria Luise da Costa Zemsch from Berlin.

Originally, a foreign postdoc should have analyzed a large, complex data set of MRI scans for the detection of Alzheimer's disease, but due to illness in her family, she had to return home after only two months. Since the research section Visual Computing had a project going on, one had to think creatively.

Instead, the three students were linked to the project, which a Master's student from the University of Copenhagen had also worked on. Their work was later taken over by Postdoc Eike Petersen, and later on, Eike presented a common research paper at the world's leading conference for medical imaging, MICCAI in Singapore.

So now the students' names appear side by side with the researchers, and the students believe this has benefited their careers. 

"My participation has clearly been career-promoting. I went straight into the labor market after my bachelor's," says Oskar Eiler Wiese Christensen, who, alongside his master's studies, currently works as a student employee at McKinsey.

Has provided important insight into research life

His fellow student, Anders Henriksen, also works at Novo Nordisk alongside his studies, and at the job interview, the boss was quite impressed by the scientific article.

"It's really nice to be able to write several different areas of competence on my CV so it reflects that I have both deep programming knowledge and academic knowledge dissemination. At the same time, it has given a lot of insight not only into the topic you write about but also into how the entire process within the research world works," says Anders Henriksen.

The two student's work with the dataset crossed paths with Maria Luise when she wrote her master's project at DTU, and she has similar experiences. Today, she is a PhD student – on a so-called standard track at the university hospital Charité in Berlin. But she is applying to a more demanding advanced track of a PhD, where only around 20 students per semester get through the eye of the needle at Charité.

"There, I think it is a big advantage that I have helped to publish a scientific article. And I can document that I have learned to sort data, can immerse myself in data and work independently," says Maria Luise.

"At the same time, the work has also been very educational, because when you hear researchers talk about their work, it sounds very exciting, and you might not think so much about how demanding it is to organize data. So working with the dataset has helped me tremendously. Now I know that data is maybe 80 percent of research. So I can use both the experience and the publication itself in my career,” she says.

Helped a lot

At DTU Compute, postdoc Eike Petersen has been enthusiastic about the way of working.

He checked that the students' processing of data was correct before working further. And the students' work has probably saved him half a year's work.

"The students took care of the heavy work. So without their help, it would have taken me much longer. It has also given me the opportunity to test out new things. And it also seemed attractive to them to do something that could potentially lead to a conference article for MICCAI. So it could be fun to do that again," says Eike Petersen.

Complicated dataset

The specific dataset is the first open, large dataset in medical imaging, and it is used worldwide because the large amount of data makes it possible to use the advanced form of machine learning 'deep learning' for analysis.

According to Professor Aasa Feragen, perhaps half of the research that is presented at all conferences on medical imaging is based on just that dataset. Despite that, it is a messy dataset where it can be difficult to understand which images belong to which patients.

"First you have to understand data from brain images and understand the challenges that are in data and understand what type of software you need to use to process data. But you also have to understand data to check that you have processed it correctly. So the students spent a long time on data sorting, and therefore it was a great advantage that they could take over from each other," explains Aasa Feragen.

"Here, we were forced to do co-creating, and the coordination of their work was quite demanding. But it has also shown that student research can be a great help and speed up the work, at the same time that the students have a relevant project to train on and learn from. And it is great that we - at the same time - have the opportunity to strengthen the students' careers," she says.

Important news for researchers worldwide

The research also led to a surprising and gratifying conclusion, which, according to the researchers, is probably most relevant to the many research groups working with that particular dataset.

When analyzing the brain scan images using a machine learning model, it detected both men and women with Alzheimer's, although either male or female patient groups were overrepresented in the data. Normally, one would expect the model to perform worse (bias) on minority groups in the data, but the model was very stable.

News and filters

Get updated on news that match your filter.