TY - GEN
T1 - Pelvis segmentation using multi-pass U-Net and iterative shape estimation
AU - Wang, Chunliang
AU - Connolly, Bryan
AU - de Oliveira Lopes, Pedro Filipe
AU - Frangi, Alejandro F.
AU - Smedby, Örjan
N1 - Funding Information:
Acknowledgements. This study was supported by the Swedish Heart-lung foundation (grant no. 20160609), Swedish Medtech4Health AIDA research grant, and the Swedish Childhood Cancer Foundation (grant no. MT2016-00166).
Publisher Copyright:
© 2019, Springer Nature Switzerland AG.
PY - 2019
Y1 - 2019
N2 - In this report, an automatic method for segmentation of the pelvis in three-dimensional (3D) computed tomography (CT) images is proposed. The method is based on a 3D U-net which has as input the 3D CT image and estimated volumetric shape models of the targeted structures and which returns the probability maps of each structure. During training, the 3D U-net is initially trained using blank shape context inputs to generate the segmentation masks, i.e. relying only on the image channel of the input. The preliminary segmentation results are used to estimate a new shape model, which is then fed to the same network again, with the input images. With the additional shape context information, the U-net is trained again to generate better segmentation results. During the testing phase, the input image is fed through the same 3D U-net multiple times, first with blank shape context channels and then with iteratively re-estimated shape models. Preliminary results show that the proposed multi-pass U-net with iterative shape estimation outperforms both 2D and 3D conventional U-nets without the shape model.
AB - In this report, an automatic method for segmentation of the pelvis in three-dimensional (3D) computed tomography (CT) images is proposed. The method is based on a 3D U-net which has as input the 3D CT image and estimated volumetric shape models of the targeted structures and which returns the probability maps of each structure. During training, the 3D U-net is initially trained using blank shape context inputs to generate the segmentation masks, i.e. relying only on the image channel of the input. The preliminary segmentation results are used to estimate a new shape model, which is then fed to the same network again, with the input images. With the additional shape context information, the U-net is trained again to generate better segmentation results. During the testing phase, the input image is fed through the same 3D U-net multiple times, first with blank shape context channels and then with iteratively re-estimated shape models. Preliminary results show that the proposed multi-pass U-net with iterative shape estimation outperforms both 2D and 3D conventional U-nets without the shape model.
KW - Deep learning
KW - Multi-pass U-net
KW - Pelvis segmentation
KW - Shape context
KW - Statistic shape model
UR - http://www.scopus.com/inward/record.url?scp=85060256089&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-11166-3_5
DO - 10.1007/978-3-030-11166-3_5
M3 - Conference contribution
AN - SCOPUS:85060256089
SN - 9783030111656
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 49
EP - 57
BT - Computational Methods and Clinical Applications in Musculoskeletal Imaging - 6th International Workshop, MSKI 2018, Held in Conjunction with MICCAI 2018, Revised Selected Papers
A2 - Zheng, Guoyan
A2 - Vrtovec, Tomaž
A2 - Yao, Jianhua
A2 - Pozo, Jose M.
PB - Springer-Verlag Italia
T2 - 6th International Workshop on Computational Methods and Clinical Applications in Musculoskeletal Imaging, MSKI 2018 was held in conjunction with Medical Image Computing and Computer-Assisted Intervention, MICCAI 2018
Y2 - 16 September 2018 through 20 September 2018
ER -