Your address will show here +12 34 56 78
Big Data for Med Tech

Medical Imaging

August 2017 - July 2020

The recent advances in computing capabilities and the availability of representative datasets allowed (deep) machine learning approaches to reach impressive performance, competing or outperforming state-of-the art tools in many fields of medical imaging. In our lab, we work more specifically on image segmentation and Radiomics with deep learning.

In adaptive radiotherapy, sequences of 3D medical images need to be automatically segmented for better treatment follow-up. Classical algorithms (i.e. registration-based algorithms) rely on the availability of an initial manual (CT) segmentation in order to propagate the contours to a subsequent (Cone Beam CT) image. However, such methods fail in case of large changes in the patient morphology along the treatment. The availability of large inter-patient labelled datasets raises hope for deep learning segmentation approaches, which will be more robust to such deformations. To this end, our research focuses on multi-organ and tumour image segmentation with fully convolutional neural networks on CT and Cone Beam CT images.

In oncology, Radiomics combines features extracted from medical images (CT, MRI …) with those extracted from other modalities (clinical, genomic, proteomic) allows to set up a personalized medicine that could be useful in the management of cancer. Being able to follow the evolution of Radiomics features during the treatment could lead to an adaptive treatment for each patients. In our work, we analyze the interest of features extracted from Cone Beam CT using convolutional neural network.

Region Wallonne funds the research relative to Big Data in Medical Technologies in the frame of the Biowin pole of competitivity.

People