Image Segmentation
Segmentation, also understood as delineation of one or more tissue or organ of interest is one of the key early steps of imaging analysis whose result can then feed a wide range of further analyses among which volumetry or shape investigation are commonly considered in association with clinical research questions.
As for many imaging analysis tasks, while manual delineation is generally considered as the gold standard, automation is required to allow large scale population analyses.
While different solutions may be required to address the challenges met in different contexts and for different organs, consistency, reproducibility, robustness and accuracy are always required to allow for the output to be used in further investigation.
This technical developments for segmentation purposes are tightly linked to other technical themes developed in the group. For instance, uncertainty modelling and calibration is crucial in light of varying imaging quality or task difficulty (Uncertainty modelling) while the need for generalisation across acquisition protocols, scanner differences request domain adaptation solutions to be integrated in the developed pipelines.
Publications
Physics-Informed Brain MRI Segmentation
Borges, P., Sudre, C., Varsavsky, T., Thomas, D., Drobnjak, I., Ourselin, S. and Cardoso, M.J., 2019, October. In International Workshop on Simulation and Synthesis in Medical Imaging (pp. 100-109). Springer, Cham.
PIMMS: permutation invariant multi-modal segmentation
Varsavsky, T., Eaton-Rosen, Z., Sudre, C.H., Nachev, P. and Cardoso, M.J., 2018. (pp. 201-209). Springer, Cham.
AdaPT: an adaptive preterm segmentation algorithm for neonatal brain MRI
Cardoso, M.J., Melbourne, A., Kendall, G.S., Modat, M., Robertson, N.J., Marlow, N. and Ourselin, S., 2013. NeuroImage, 65, pp.97-108.