Detail publikace

Trainable Segmentation Based on Local-level and Segment-level Feature Extraction

Originální název

Trainable Segmentation Based on Local-level and Segment-level Feature Extraction

Anglický název

Trainable Segmentation Based on Local-level and Segment-level Feature Extraction

Jazyk

en

Originální abstrakt

This paper deals with the segmentation of neuronal struc- tures in electron microscope (EM) stacks, which is one of the challenges of the ISBI 2012 conference. The data for the challenge consists of a stack of 30 EM slices for training and 30 EM stacks for testing. The training data was labelled by an expert human neuroanatomist. In this paper a segmentation using local-level and segment-level features and machine learning algorithms was used. The results achieved on the ISBI 2012 challenge test set were: the Rand error: 0.139038440, warping er- ror: 0.002641296 and pixel error: 0.102285508. The main criterion for segmentation evaluation was the Rand error.

Anglický abstrakt

This paper deals with the segmentation of neuronal struc- tures in electron microscope (EM) stacks, which is one of the challenges of the ISBI 2012 conference. The data for the challenge consists of a stack of 30 EM slices for training and 30 EM stacks for testing. The training data was labelled by an expert human neuroanatomist. In this paper a segmentation using local-level and segment-level features and machine learning algorithms was used. The results achieved on the ISBI 2012 challenge test set were: the Rand error: 0.139038440, warping er- ror: 0.002641296 and pixel error: 0.102285508. The main criterion for segmentation evaluation was the Rand error.

BibTex


@inproceedings{BUT94573,
  author="Radim {Burget} and Václav {Uher} and Jan {Mašek}",
  title="Trainable Segmentation Based on Local-level and Segment-level Feature Extraction",
  annote="This paper deals with the segmentation of neuronal struc-
tures in electron microscope (EM) stacks, which is one of the challenges
of the ISBI 2012 conference. The data for the challenge consists of a
stack of 30 EM slices for training and 30 EM stacks for testing. The
training data was labelled by an expert human neuroanatomist. In this
paper a segmentation using local-level and segment-level features and
machine learning algorithms was used. The results achieved on the ISBI
2012 challenge test set were: the Rand error: 0.139038440, warping er-
ror: 0.002641296 and pixel error: 0.102285508. The main criterion for
segmentation evaluation was the Rand error.",
  booktitle="IEEE International Symposium on Biomedical Imaging",
  chapter="94573",
  edition="1",
  howpublished="online",
  year="2012",
  month="june",
  pages="17--24",
  type="conference paper"
}