CWI closely collaborates with the Academic Medical Center (AMC) in Amsterdam to work on innovations in the medical domain along the entire spectrum from algorithmic foundations to clinical integration. Currently, CWI and AMC have a joint project for which we seek multiple talented PhD students to work on novel combinations of multi-objective evolutionary algorithms, machine learning, and biomechanical modelling for deformable image registration.
Deformable image registration is the process of aligning two images by determining the transformation that maps points in one image to their corresponding points in the other image. Deformable image registration has potentially huge added value to many applications in healthcare since images form a cornerstone for many diagnoses, treatments, and follow-up in modern medicine. Key reasons why clinical implementation has proven to be extremely challenging are lack of robustness, limited success for registration problems involving large deformations or content mismatches, and a lack of insightful tuning possibilities.
In this project, we aim to go beyond the proof-of-concept stage of a novel, multi-objective, approach to deformable image registration for which a pilot implementation was previously made by the Academic Medical Center (AMC) and Centrum Wiskunde & Informatica (CWI; the Dutch national research institute for mathematics and computer science). We will focus on the utilization of this approach within a key medical area that stands to benefit from accurate deformable image registration: radiation therapy. With this project we wish to truly bridge the gap between theoretically powerful deformable image registration software and real-world practically (i.e., clinically) useful software tools to realize the true potential of deformable image registration for radiation therapy and ultimately improve the treatment (i.e., make radiation therapy more effective thereby reducing mortality rate and adverse side effects). We aim to arrive at a stage of development where commercial parties will be interested to integrate our technology within their software.
We currently specifically seek candidates for three intertwined subprojects:
- Building an evolutionary multi-objective optimization engine that is especially well-suited for deformable image registration, based on a state-of-the-art evolutionary algorithms research line of CWI (the GOMEA research line).
- Building biomechanical deformation models that
- can cope with large deformations (including e.g., sliding tissues) and content mismatches, and
- can be efficiently optimized over by the evolutionary algorithms, and
- can be instantiated by machine-learning techniques.
- Development and validation of the use of machine learning (e.g., deep learning) to distinguish tissue types and automatically identify landmarks to enrich deformation models.
This project will ultimately consist of four full-time research positions and a part-time radiation therapy technologist position. The fourth full-time researcher will be added to the team after one year.