MSc Student: Advancing Collaborative Medical Image Analysis: Federated Learning and Disentangled Representation for Privacy-Preserving Segmentation

In the context of medical imaging, it is often infeasible to collect patient data in a centralized data lake due to privacy regulations. Furthermore, medical data sets are often siloed across many institutes and are highly unbalanced due to the low incidence of pathologies or biased towards local demographics. This situation gains particular significance in the European Union, where data protection regulations are enacted to uphold high levels of trust, security, and safeguarding of personal data. In contrast to the conventional centralized approach, Federated Learning (FL) emerges as a learning paradigm that seeks to grapple with issues of data governance and privacy by collaboratively training algorithms without the exchange of raw data. FL facilitates collaborative insight generation without necessitating the movement of patient data beyond their respective institutional firewalls. Rather, the learning process occurs locally at each participating institution. FL enables multiple parties to collaboratively train a Deep Learning (DL) model without exchanging local data. Nevertheless, the divergence in data characteristics (non-IID) among the distributed clients remains a challenge. To surmount this limitation, our intent is to delve into the integration of Disentangled Representation Learning (DRL) frameworks within the FL context. The DRL model would separate the generating factors of variability in the dataset into representations related to anatomy and those pertaining to the domain/center. Subsequently, solely the anatomy-relevant factors would be shared with the clients, thus augmenting the model's capacity for generalization. As of now, this research domain remains relatively unexplored, presenting ample opportunity for novel approaches.

Key objectives of the research project encompass:

Designing and probing the fusion of DRL segmentation techniques with FL methodologies. Training and conducting comparative evaluations of DRL frameworks within a FL framework using real-world available datasets (e.g. PolypGen).

What we expect from you:

A serious work ethic with a proactive attitude, with willingness to engage in a research project and understand the technical question. Knowledge of deep learning and experience with python and deep learning frameworks, from courses and projects. We don’t require proficiency, you will have plenty of time to learn! We are also looking for interest in publishing the final results in international medical imaging and/or deep learning conferences.

What you can expect from us:

An exciting and relevant AI research topic in one of the best cancer institutes in the world. You will work in the Radiology Department of the Netherlands Cancer Institute, a multidisciplinary environment, where you will collaborate closely with medical professionals. In particular, you will integrate the AI for Multi-center Data research team. You will have the possibility to explore and come up with your own ideas, while being under the direct supervision of a PhD student and the PI of the research line, Dr. Wilson Silva. The research line has its own biweekly group meetings, technical journal club and 1:1 meetings. You will also participate in the department meetings. You will be granted access to the research high performance facility.