Vacancies: Chest X-ray Analysis with Deep Learning


Project description

We are in the middle of a revolution of Artificial Intelligence, and deep learning in particular, that is rapidly changing the field of medical image analysis. Computers are now able to do things that were not considered possible a few years ago.

This project focuses on the most common radiological exam: the chest x-ray. Hundreds of millions of such images are acquired in hospitals and clinics all over the world, for a large number of indications. In our group, we have developed algorithms for assessing the presence of tuberculosis in these images, and this software is now operational in 24 countries at 150 sites. In this project, we aim to substantially extend our portfolio of algorithms, in order to automatically find and quantify a broad spectrum of diseases and abnormal signs in chest radiographs. We have already collected hundreds of thousands of image exams, and are working with a network of partner sites to extend and annotate our databases. Research questions include:

  • How to best parse text reports, e.g. with LSTM, and images, e.g. with CNNs, to semi-automatically annotate data?
  • How can generative adversarial networks standardize and improve the image quality of the radiographs, and make automated analysis robust to variations in data sources (i.e. images from different types of x-ray equipment)?
  • Can we improve algorithms with unsupervised and semi-supervised deep learning?
  • Can we jointly train models on two chest x-rays (a frontal and lateral 2D projection image) and a corresponding chest CT scan of the same patient (a 3D image), and reconstruct 3D information from the 2D projections?

You will be part of a team of in total three researchers (two vacancies, one position is filled with our TTW Deep Learning in Medical Image Analysis Perspective Programme). We collaborate with various companies (Delft Imaging Systems, Thirona, and Smart Reporting) and we expect that the results of the project will be integrated in products, both aimed at the Western market and developing countries.


We are looking for ambitious deep learning engineers, data scientists, or machine learning experts. You should be a creative, and enthusiastic and have an MSc/Ph.D. degree in Computer Science, Data Science, Physics, Engineering or Biomedical Sciences or similar, with a clear interest in deep learning, image analysis and medical applications. Good communication skills and expertise in software development, preferably in Python/C++, are essential. Experience with machine learning should be evident from the (online) courses you've followed, your publications, GitHub account, etc. Experience with medical image processing is preferred.

Terms of employment

The positions can be filled by either Ph.D. students or postdocs.

  • Ph.D. positions are for four years and have the standard salary and secondary conditions for Ph.D. students in the Netherlands. Your performance will be evaluated after 1 year. If the evaluation is positive, the contract will be extended by 3 years. The research should result in a Ph.D. thesis. See also our page with general information about doing a Ph.D. in our group.
  • Postdoc positions are for three years. Your performance will be evaluated after 1 year. If the evaluation is positive, the contract will be extended by 2 years.


The Diagnostic Image Analysis Group (DIAG) is a research division of the Department of Radiology and Nuclear Medicine of the Radboud University Medical Center Nijmegen. Nijmegen is the oldest Dutch city with a rich history and one of the liveliest city centers in the Netherlands. Radboud University has over 17,000 students. Radboud UMC is a leading academic center for medical science, education, and healthcare with over 8,500 staff and 3,000 students.

The focus of DIAG is the development and validation of novel methods in a broad range of medical imaging applications. Research topics include image analysis, image segmentation, machine learning, and the design of decision support systems. Application areas include neuro, breast, prostate, lung and retinal imaging and computational pathology. Key to the success of the group is close cooperation with clinicians. Currently, the group consists of around 40 researchers. See our deep learning paper overview to get an idea of our work in this field.

We offer excellent research facilities with large data storage facilities, a cluster of 40 high-end GPUs which can be scaled dynamically to include cloud servers, and support from a team of research software engineers, data analysts, and radiologists.


This application has been filled.