TY - GEN
T1 - CFCM
T2 - 21st International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2018
AU - Milletari, Fausto
AU - Rieke, Nicola
AU - Baust, Maximilian
AU - Esposito, Marco
AU - Navab, Nassir
N1 - Publisher Copyright:
© 2018, Springer Nature Switzerland AG.
PY - 2018
Y1 - 2018
N2 - Recent neural-network-based architectures for image segmentation make extensive usage of feature forwarding mechanisms to integrate information from multiple scales. Although yielding good results, even deeper architectures and alternative methods for feature fusion at different resolutions have been scarcely investigated for medical applications. In this work we propose to implement segmentation via an encoder-decoder architecture which differs from any other previously published method since (i) it employs a very deep architecture based on residual learning and (ii) combines features via a convolutional Long Short Term Memory (LSTM), instead of concatenation or summation. The intuition is that the memory mechanism implemented by LSTMs can better integrate features from different scales through a coarse-to-fine strategy; hence the name Coarse-to-Fine Context Memory (CFCM). We demonstrate the remarkable advantages of this approach on two datasets: the Montgomery county lung segmentation dataset, and the EndoVis 2015 challenge dataset for surgical instrument segmentation.
AB - Recent neural-network-based architectures for image segmentation make extensive usage of feature forwarding mechanisms to integrate information from multiple scales. Although yielding good results, even deeper architectures and alternative methods for feature fusion at different resolutions have been scarcely investigated for medical applications. In this work we propose to implement segmentation via an encoder-decoder architecture which differs from any other previously published method since (i) it employs a very deep architecture based on residual learning and (ii) combines features via a convolutional Long Short Term Memory (LSTM), instead of concatenation or summation. The intuition is that the memory mechanism implemented by LSTMs can better integrate features from different scales through a coarse-to-fine strategy; hence the name Coarse-to-Fine Context Memory (CFCM). We demonstrate the remarkable advantages of this approach on two datasets: the Montgomery county lung segmentation dataset, and the EndoVis 2015 challenge dataset for surgical instrument segmentation.
UR - http://www.scopus.com/inward/record.url?scp=85053841373&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-00937-3_76
DO - 10.1007/978-3-030-00937-3_76
M3 - Conference contribution
AN - SCOPUS:85053841373
SN - 9783030009366
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 667
EP - 674
BT - Medical Image Computing and Computer Assisted Intervention – MICCAI 2018 - 21st International Conference, 2018, Proceedings
A2 - Frangi, Alejandro F.
A2 - Fichtinger, Gabor
A2 - Schnabel, Julia A.
A2 - Alberola-López, Carlos
A2 - Davatzikos, Christos
PB - Springer Verlag
Y2 - 16 September 2018 through 20 September 2018
ER -