Joint eScience and Data Science grant for CWI's Peter Bosman and AMC

Recently, a research proposal that is spearheaded by CWI was granted in the Joint eScience and Data Science programme that was organized jointly by the NWO Domain Science and Netherlands eScience Center (NLeSC). In all projects in this programme, researchers collaborate with eScience engineers of the NLeSC.

Publication date: 21-12-2017

Recently, a research proposal that is spearheaded by CWI was granted in the Joint eScience and Data Science programme that was organized jointly by the NWO Domain Science and Netherlands eScience Center (NLeSC). In all projects in this programme, researchers collaborate with eScience engineers of the NLeSC. The proposal, entitled "FEDMix: Fusible Evolutionary Deep Neural Network Mixture Learning from Distributed Data for Robust Medical Image Analysis" was submitted by Peter Bosman of CWI's Life Sciences & Health group and Tanja Alderliesten of the Department of Radiation Oncology of the Academic Medical Center (AMC) in Amsterdam.

The proposal is focused on establishing innovation in automated Medical Image Analysis (MIA). MIA has the potential to truly innovate clinical practice by offering solutions to routine, yet key tasks, such as segmentation (i.e., delineating organs). With recent advances in machine learning, in particular in Deep Neural Networks (DNNs), there has been an explosive growth of successful MIA studies reported in academic literature. Yet, labor-intensive manual performance of these tasks is still often daily clinical practice.

The researchers aim to make new combinations of modern DNNs with other state-of-the-art computational intelligence techniques, in particular evolutionary algorithms, to overcome 2 key obstacles in moving toward widespread clinical uptake of computationally intelligent MIA techniques: 1) observer variation in the definition of a ground truth, and 2) image quality variation due to different acquisition protocols and scanners at different institutes. In particular, they will design and develop efficient-computing-compatible implementations of mixtures of DNNs, the results of which can be fused with results learned from other data sets (i.e., from different institutes). To maintain sufficient focus while doing so, an elementary, but key MIA task will be considered: segmentation. The newly developed technology will be validated on real-world patient data within the runtime of the proposed project by means of an application in radiotherapy treatment planning, in collaboration with the AMC.

 

Life Sciences & Health group