Meet the Netherlands Cancer Institute

The Netherlands Cancer Institute combines excellence in research with a specialised oncology clinic, accelerating the translation of scientific discoveries into clinical practice.

We spoke with Sara Oliveira, a postdoctoral researcher in computational pathology at the NKI and a former researcher at INESC TEC.

How did the collaboration between NKI and INESC TEC begin? 

In 2016, I joined INESC TEC as a researcher and carried out my PhD thesis at VCMI, under the supervision of researcher Hélder Oliveira and Professor Jaime Cardoso. At the end of my PhD, I decided to take on a challenge abroad and, in January 2023, I moved to Amsterdam, where I am currently a postdoctoral researcher in the Computational Pathology group at The Netherlands Cancer Institute (NKI). Following my PhD work, maintaining collaboration with VCMI – and particularly with Professor Jaime Cardoso – was a very natural step, given the extensive experience already built in this topic and the strong results from previous projects. 

What stands out in this partnership? 

I would highlight the complementarity: NKI provides access to clinical data, specialised expertise in pathology and oncology, and a strong translational perspective, while INESC TEC stands out for technical thoroughness and the capacity for innovation in Artificial Intelligence (AI) algorithms and pipelines. 

Within the scope of the joint supervision of theses and student exchanges, what experiences or learnings do you consider most valuable? 

The most valuable experience is exposing students to different scientific perspectives and giving them the opportunity to work on problems that cut across disciplinary boundaries.
In addition, co-supervised theses help to train highly multidisciplinary and flexible profiles, capable of communicating effectively with engineers, pathologists and biomedical researchers. 

What have been the main challenges when working with image analysis or AI algorithms in digital pathology – and within this collaboration? 

The main challenges relate to data volume, computational requirements and the need to balance performance with interpretability. In computational pathology, we analyse very large images, which requires robust computational infrastructure, as well as efficient strategies for data storage, processing and model training.
Furthermore, there is a constant need to balance the performance of AI models with their interpretability. Although more complex models tend to achieve better results, it is essential that their decisions are explainable and clinically understandable, to ensure trust and adoption by pathologists.   

What do you value most in this experience with INESC TEC? 

The excellence of the VCMI team: they show concrete scientific thoroughness, along with great flexibility and curiosity to experiment with and explore innovative approaches. In addition, the collaborative environment is a major asset for students, as there is always someone available to help. It is great to see that the group’s values remain unchanged. 

What can we expect from this collaboration in the coming years? 

I expect a significant strengthening of joint research, with more projects in computational pathology, but also in multimodal models. I believe the partnership will continue to produce relevant results and to bring research closer to real impact in clinical practice.