TY - CHAP
T1 - The Role of MRI Physics in Brain Segmentation CNNs
T2 - 6th International Workshop on Simulation and Synthesis in Medical Imaging, SASHIMI 2021, held in conjunction with the 24th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2021
AU - Borges, Pedro
AU - Shaw, Richard
AU - Varsavsky, Thomas
AU - Klaser, Kerstin
AU - Thomas, David
AU - Drobnjak, Ivana
AU - Ourselin, Sebastien
AU - Jorge Cardoso, M.
N1 - Funding Information:
Acknowledgements. This project was funded by the Wellcome Flagship Programme (WT213038/Z/18/Z) and Wellcome EPSRC CME (WT203148/Z/16/Z).
Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
Copyright:
Copyright 2021 Elsevier B.V., All rights reserved.
PY - 2021
Y1 - 2021
N2 - Being able to adequately process and combine data arising from different sites is crucial in neuroimaging, but is difficult, owing to site, sequence and acquisition-parameter dependent biases. It is important therefore to design algorithms that are not only robust to images of differing contrasts, but also be able to generalise well to unseen ones, with a quantifiable measure of uncertainty. In this paper we demonstrate the efficacy of a physics-informed, uncertainty-aware, segmentation network that employs augmentation-time MR simulations and homogeneous batch feature stratification to achieve acquisition invariance. We show that the proposed approach also accurately extrapolates to out-of-distribution sequence samples, providing well calibrated volumetric bounds on these. We demonstrate a significant improvement in terms of coefficients of variation, backed by uncertainty based volumetric validation.
AB - Being able to adequately process and combine data arising from different sites is crucial in neuroimaging, but is difficult, owing to site, sequence and acquisition-parameter dependent biases. It is important therefore to design algorithms that are not only robust to images of differing contrasts, but also be able to generalise well to unseen ones, with a quantifiable measure of uncertainty. In this paper we demonstrate the efficacy of a physics-informed, uncertainty-aware, segmentation network that employs augmentation-time MR simulations and homogeneous batch feature stratification to achieve acquisition invariance. We show that the proposed approach also accurately extrapolates to out-of-distribution sequence samples, providing well calibrated volumetric bounds on these. We demonstrate a significant improvement in terms of coefficients of variation, backed by uncertainty based volumetric validation.
UR - http://www.scopus.com/inward/record.url?scp=85115843323&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-87592-3_7
DO - 10.1007/978-3-030-87592-3_7
M3 - Conference paper
AN - SCOPUS:85115843323
SN - 9783030875916
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 67
EP - 76
BT - Simulation and Synthesis in Medical Imaging - 6th International Workshop, SASHIMI 2021, Held in Conjunction with MICCAI 2021, Proceedings
A2 - Svoboda, David
A2 - Burgos, Ninon
A2 - Wolterink, Jelmer M.
A2 - Zhao, Can
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 27 September 2021 through 27 September 2021
ER -