TY - GEN
T1 - Combining Deep Learning and Shape Priors for Bi-Ventricular Segmentation of Volumetric Cardiac Magnetic Resonance Images
AU - Duan, Jinming
AU - Schlemper, Jo
AU - Bai, Wenjia
AU - Dawes, Timothy J.W.
AU - Bello, Ghalib
AU - Biffi, Carlo
AU - Doumou, Georgia
AU - De Marvao, Antonio
AU - O’Regan, Declan P.
AU - Rueckert, Daniel
N1 - Publisher Copyright:
© 2018, Springer Nature Switzerland AG.
PY - 2018
Y1 - 2018
N2 - In this paper, we combine a network-based method with image registration to develop a shape-based bi-ventricular segmentation tool for short-axis cardiac magnetic resonance (CMR) volumetric images. The method first employs a fully convolutional network (FCN) to learn the segmentation task from manually labelled ground truth CMR volumes. However, due to the presence of image artefacts in the training dataset, the resulting FCN segmentation results are often imperfect. As such, we propose a second step to refine the FCN segmentation. This step involves performing a non-rigid registration with multiple high-resolution bi-ventricular atlases, allowing the explicit shape priors to be inferred. We validate the proposed approach on 1831 healthy subjects and 200 subjects with pulmonary hypertension. Numerical experiments on the two datasets demonstrate that our approach is capable of producing accurate, high-resolution and anatomically smooth bi-ventricular models, despite the artefacts in the input CMR volumes.
AB - In this paper, we combine a network-based method with image registration to develop a shape-based bi-ventricular segmentation tool for short-axis cardiac magnetic resonance (CMR) volumetric images. The method first employs a fully convolutional network (FCN) to learn the segmentation task from manually labelled ground truth CMR volumes. However, due to the presence of image artefacts in the training dataset, the resulting FCN segmentation results are often imperfect. As such, we propose a second step to refine the FCN segmentation. This step involves performing a non-rigid registration with multiple high-resolution bi-ventricular atlases, allowing the explicit shape priors to be inferred. We validate the proposed approach on 1831 healthy subjects and 200 subjects with pulmonary hypertension. Numerical experiments on the two datasets demonstrate that our approach is capable of producing accurate, high-resolution and anatomically smooth bi-ventricular models, despite the artefacts in the input CMR volumes.
UR - http://www.scopus.com/inward/record.url?scp=85057387021&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-04747-4_24
DO - 10.1007/978-3-030-04747-4_24
M3 - Conference contribution
AN - SCOPUS:85057387021
SN - 9783030047467
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 258
EP - 267
BT - Shape in Medical Imaging - International Workshop, ShapeMI 2018, Held in Conjunction with MICCAI 2018, Proceedings
A2 - Lombaert, Hervé
A2 - Paniagua, Beatriz
A2 - Egger, Bernhard
A2 - Lüthi, Marcel
A2 - Reuter, Martin
A2 - Wachinger, Christian
PB - Springer Verlag
T2 - International Workshop on Shape in Medical Imaging, ShapeMI 2018 held in conjunction with 21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018
Y2 - 20 September 2018 through 20 September 2018
ER -