Extracting evolving pathologies via spectral clustering

Elena Bernardis, Kilian M. Pohl, Christos Davatzikos

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

A bottleneck in the analysis of longitudinal MR scans with white matter brain lesions is the temporally consistent segmentation of the pathology. We identify pathologies in 3D+t(ime) within a spectral graph clustering framework. Our clustering approach simultaneously segments and tracks the evolving lesions by identifying characteristic image patterns at each time-point and voxel correspondences across time-points. For each 3D image, our method constructs a graph where weights between nodes capture the likeliness of two voxels belonging to the same region. Based on these weights, we then establish rough correspondences between graph nodes at different time-points along estimated pathology evolution directions. We combine the graphs by aligning the weights to a reference time-point, thus integrating temporal information across the 3D images, and formulate the 3D+t segmentation problem as a binary partitioning of this graph. The resulting segmentation is very robust to local intensity fluctuations and yields better results than segmentations generated for each time-point.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages680-691
Number of pages12
Volume7917 LNCS
DOIs
StatePublished - 2013
Externally publishedYes
Event23rd International Conference on Information Processing in Medical Imaging, IPMI 2013 - Asilomar, CA, United States
Duration: Jun 28 2013Jul 3 2013

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7917 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other23rd International Conference on Information Processing in Medical Imaging, IPMI 2013
CountryUnited States
CityAsilomar, CA
Period6/28/137/3/13

Fingerprint

Spectral Clustering
Pathology
Segmentation
Voxel
3D Image
Graph in graph theory
Correspondence
Brain
Graph Clustering
Vertex of a graph
Rough
Partitioning
Clustering
Fluctuations
Binary

Keywords

  • 4D segmentation
  • longitudinal tracking
  • MRI white matter lesion

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Bernardis, E., Pohl, K. M., & Davatzikos, C. (2013). Extracting evolving pathologies via spectral clustering. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7917 LNCS, pp. 680-691). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7917 LNCS). https://doi.org/10.1007/978-3-642-38868-2_57

Extracting evolving pathologies via spectral clustering. / Bernardis, Elena; Pohl, Kilian M.; Davatzikos, Christos.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7917 LNCS 2013. p. 680-691 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7917 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bernardis, E, Pohl, KM & Davatzikos, C 2013, Extracting evolving pathologies via spectral clustering. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 7917 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7917 LNCS, pp. 680-691, 23rd International Conference on Information Processing in Medical Imaging, IPMI 2013, Asilomar, CA, United States, 6/28/13. https://doi.org/10.1007/978-3-642-38868-2_57
Bernardis E, Pohl KM, Davatzikos C. Extracting evolving pathologies via spectral clustering. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7917 LNCS. 2013. p. 680-691. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-38868-2_57
Bernardis, Elena ; Pohl, Kilian M. ; Davatzikos, Christos. / Extracting evolving pathologies via spectral clustering. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7917 LNCS 2013. pp. 680-691 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{277cb14d38a54886bf3e24316e8aeaef,
title = "Extracting evolving pathologies via spectral clustering",
abstract = "A bottleneck in the analysis of longitudinal MR scans with white matter brain lesions is the temporally consistent segmentation of the pathology. We identify pathologies in 3D+t(ime) within a spectral graph clustering framework. Our clustering approach simultaneously segments and tracks the evolving lesions by identifying characteristic image patterns at each time-point and voxel correspondences across time-points. For each 3D image, our method constructs a graph where weights between nodes capture the likeliness of two voxels belonging to the same region. Based on these weights, we then establish rough correspondences between graph nodes at different time-points along estimated pathology evolution directions. We combine the graphs by aligning the weights to a reference time-point, thus integrating temporal information across the 3D images, and formulate the 3D+t segmentation problem as a binary partitioning of this graph. The resulting segmentation is very robust to local intensity fluctuations and yields better results than segmentations generated for each time-point.",
keywords = "4D segmentation, longitudinal tracking, MRI white matter lesion",
author = "Elena Bernardis and Pohl, {Kilian M.} and Christos Davatzikos",
year = "2013",
doi = "10.1007/978-3-642-38868-2_57",
language = "English (US)",
isbn = "9783642388675",
volume = "7917 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "680--691",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Extracting evolving pathologies via spectral clustering

AU - Bernardis, Elena

AU - Pohl, Kilian M.

AU - Davatzikos, Christos

PY - 2013

Y1 - 2013

N2 - A bottleneck in the analysis of longitudinal MR scans with white matter brain lesions is the temporally consistent segmentation of the pathology. We identify pathologies in 3D+t(ime) within a spectral graph clustering framework. Our clustering approach simultaneously segments and tracks the evolving lesions by identifying characteristic image patterns at each time-point and voxel correspondences across time-points. For each 3D image, our method constructs a graph where weights between nodes capture the likeliness of two voxels belonging to the same region. Based on these weights, we then establish rough correspondences between graph nodes at different time-points along estimated pathology evolution directions. We combine the graphs by aligning the weights to a reference time-point, thus integrating temporal information across the 3D images, and formulate the 3D+t segmentation problem as a binary partitioning of this graph. The resulting segmentation is very robust to local intensity fluctuations and yields better results than segmentations generated for each time-point.

AB - A bottleneck in the analysis of longitudinal MR scans with white matter brain lesions is the temporally consistent segmentation of the pathology. We identify pathologies in 3D+t(ime) within a spectral graph clustering framework. Our clustering approach simultaneously segments and tracks the evolving lesions by identifying characteristic image patterns at each time-point and voxel correspondences across time-points. For each 3D image, our method constructs a graph where weights between nodes capture the likeliness of two voxels belonging to the same region. Based on these weights, we then establish rough correspondences between graph nodes at different time-points along estimated pathology evolution directions. We combine the graphs by aligning the weights to a reference time-point, thus integrating temporal information across the 3D images, and formulate the 3D+t segmentation problem as a binary partitioning of this graph. The resulting segmentation is very robust to local intensity fluctuations and yields better results than segmentations generated for each time-point.

KW - 4D segmentation

KW - longitudinal tracking

KW - MRI white matter lesion

UR - http://www.scopus.com/inward/record.url?scp=84879854193&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84879854193&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-38868-2_57

DO - 10.1007/978-3-642-38868-2_57

M3 - Conference contribution

SN - 9783642388675

VL - 7917 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 680

EP - 691

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -