TY - GEN
T1 - Multi-class image segmentation in fluorescence microscopy using polytrees
AU - Fehri, Hamid
AU - Gooya, Ali
AU - Johnston, Simon A.
AU - Frangi, Alejandro F.
N1 - Funding Information:
This work was supported by MRC fellowship (MR/J009156/1) and the Krebs Institute fellowship. Alejandro F. Frangi is partially funded by BBSRC through grant BB/M01021X/1.
Publisher Copyright:
© Springer International Publishing AG 2017.
PY - 2017
Y1 - 2017
N2 - Multi-class segmentation is a crucial step in cell image analysis. This process becomes challenging when little information is available for recognising cells from the background, due to their poor discriminative features. To alleviate this, directed acyclic graphs such as trees have been proposed to model top-down statistical dependencies as a prior for improved image segmentation. However, using trees, modelling the relations between labels of multiple classes becomes difficult. To overcome this limitation, we propose a polytree graphical model that captures label proximity relations more naturally compared to tree based approaches. A novel recursive mechanism based on two-pass message passing is developed to efficiently calculate closed form posteriors of graph nodes on the polytree. The algorithm is evaluated using simulated data, synthetic images and real fluorescence microscopy images. Our method achieves Dice scores of 94.5% and 98% on macrophage and seed classes, respectively, outperforming GMM based classifiers.
AB - Multi-class segmentation is a crucial step in cell image analysis. This process becomes challenging when little information is available for recognising cells from the background, due to their poor discriminative features. To alleviate this, directed acyclic graphs such as trees have been proposed to model top-down statistical dependencies as a prior for improved image segmentation. However, using trees, modelling the relations between labels of multiple classes becomes difficult. To overcome this limitation, we propose a polytree graphical model that captures label proximity relations more naturally compared to tree based approaches. A novel recursive mechanism based on two-pass message passing is developed to efficiently calculate closed form posteriors of graph nodes on the polytree. The algorithm is evaluated using simulated data, synthetic images and real fluorescence microscopy images. Our method achieves Dice scores of 94.5% and 98% on macrophage and seed classes, respectively, outperforming GMM based classifiers.
UR - https://www.scopus.com/pages/publications/85020503731
U2 - 10.1007/978-3-319-59050-9_41
DO - 10.1007/978-3-319-59050-9_41
M3 - Conference contribution
AN - SCOPUS:85020503731
SN - 9783319590493
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 517
EP - 528
BT - Information Processing in Medical Imaging - 25th International Conference, IPMI 2017, Proceedings
A2 - Zhu, Hongtu
A2 - Niethammer, Marc
A2 - Styner, Martin
A2 - Zhu, Hongtu
A2 - Shen, Dinggang
A2 - Yap, Pew-Thian
A2 - Aylward, Stephen
A2 - Oguz, Ipek
PB - Springer-Verlag Italia
T2 - 25th International Conference on Information Processing in Medical Imaging, IPMI 2017
Y2 - 25 June 2017 through 30 June 2017
ER -