Project information

Ductal carcinoma in situ (DCIS) may progress to invasive breast cancer (IBC), but about 3 out of 4 never will. We cannot reliably distinguish the minority of high-risk, potentially progressive DCIS lesions from the low-risk, harmless ones in current practice. Consequently, almost all women with DCIS are treated with surgery, often supplemented by radiotherapy, to prevent progression to IBC. Therefore, there is an urgent need to distinguish the high- from low-risk lesions to avoid overtreatment of the latter ones. This is not trivial as DCIS is a widespread finding in women, primarily due to the introduction of population-based breast cancer screening, comprising up to 25% of all newly diagnosed breast 'malignancies.' We have the most extensive well-annotated DCIS datasets in the world available with over 12 years of follow-up. Using this, we will develop state-of-the-art artificial intelligence (AI) algorithm to distinguish high- from low-risk DCIS at the level of histopathology. In addition, we will investigate the morphological and biological features that the AI model will select; we will also integrate AI with genomics data. This project will develop deep learning techniques for quantifying biomarkers in histopathology images (H&E and IHC) in multimodal data. As the input dimensionality can exceed 250 000 x 250 000, we will research novel self-supervised contrastive learning techniques combined with transformers/multiple instance learning models. Eventually, we will combine these with model interpretability techniques to discover new insights into DCIS.

The job description

As a PhD-candidate/postdoc, you will be responsible for developing and evaluating state-of-the-art deep learning techniques in multidisciplinary data. You will validate these algorithms in independent cases to ensure the devised AI-algorithms’ applicability in clinical practice.

You are embedded in the Wesseling Lab and the Teuwen Lab (AI for Oncology) at the Netherlands Cancer. The Wesseling’ lab overarching research focus is to develop new strategies to substantially improve the prognostic and predictive power of breast cancer pathology and to understand the underlying biology. The mission of AI for Oncology Lab is to develop artificial intelligence innovations for the improvement of cancer diagnostics and therapy. You will discuss results with our team, publish your work in artificial intelligence / medical journals, and present it at international conferences.

Your profile

We are looking for a motivated, independent, and proactive PhD candidate or postdoc who is enthusiastic about working in a multidisciplinary setting to improve breast cancer care using AI. A teamplayer’s mentality and excellent verbal and written communication skills are essential for this position. You will develop novel deep learning algorithms and shape the research in collaboration with the project leaders, AI, and medical experts.

Preferably you have a master's degree in artificial intelligence, computer science, physics, mathematics, or equivalent by experience. In any case, you should have experience with deep learning and have excellent programming skills. Affinity with deep learning and the other required techniques should be clear from your materials submitted, including courses followed and your GitHub account.

Interested?

For questions or more information regarding this vacancy or research, you can contact Dr. Jonas Teuwen via j.teuwen@nki.nl or Prof. dr. Jelle Wesseling via j.wesseling@nki.nl.

How to apply

Please send your application in via our website, applications sent directly to e-mail will not be processed. Furthermore we would like to receive a single PDF file with all documents listed below:

  • A motivation letter that explains why you are interested in our vacancy and joining our team;
  • Curriculum vitae;
  • The names and contact addresses of at least two references, please list the references in your letter of interest;
  • A list of projects and publications you have worked on (with brief descriptions of your contributions, max two pages).