Deep learning 

Advances in tissue slide digitization and machine learning have propelled computational pathology research. Especially the use of 'deep learning' techniques, trained with large numbers of histopathology images, has been shown to be very powerful.

Today, computer systems approach the level of humans for certain well-defined tasks in pathology. Examples are counting of mitoses for breast cancer grading and detection of lymph node metastases for tumour staging. 

My research focuses on development of such deep learning algorithms. The aims are twofold:

  1. to support the pathologists' work by increasing efficiency and reducing observer bias;
  2. to identify potential new (prognostic and predictive) biomarkers to aid personalized treatment.

To be able to reach these, a number of eminent challenges still exist. An important prerequisite for development of deep learning algorithms is the availability of (both high quality and high quantity) data. A large part of the research is therefore directed at establishing collaborations, acquiring clinical data as well as human tissues, and working with expert pathologists.

Next, research into different deep learning strategies is required to develop the most optimal models.

Lastly, developed models have to be validated in routine clinical practice, to prove safety and usability.

My research aims to focus on all these different aspects, with the final aim of improving cancer diagnostics and prognostics.