Refining Mitochondria Segmentation in Electron Microscopy Imagery with Active Surfaces

We present an active surface-based method for refining the boundary surfaces of mitochondria segmentation data. We exploit the fact that mitochondria have thick dark membranes, so referencing the image data at the inner membrane can help drive a more accurate delineation of the outer membrane surface. Given the initial boundary prediction from a machine learning-based segmentation algorithm as input, we compare several cost functions used to drive an explicit update scheme to locally refine 3D mesh surfaces, and results are presented on electron microscopy imagery. Our resulting surfaces are seen to fit very accurately to the mitochondria membranes, more accurately even than the available hand-annotations of the data.


Editor(s):
Agapito, Lourdes
Bronstein, Michael M.
Rother, Carsten
Published in:
Computer Vision - ECCV 2014 Workshops, IV, 367-379
Presented at:
European Conference on Computer Vision (ECCV) Workshop on Non-Rigid Shape Analysis and Deformable Image Alignment, Zurich, Switzerland, September 6-12, 2014
Year:
2014
Publisher:
Berlin, Springer
ISBN:
978-3-319-16220-1
978-3-319-16219-5
Laboratories:




 Record created 2014-09-30, last modified 2018-05-07

Postprint:
Download fulltext
PDF

Rate this document:

Rate this document:
1
2
3
 
(Not yet reviewed)