The successive projection algorithm as an initialization method for brain tumor segmentation using non-negative matrix factorization

Nicolas Sauwen; Marjan Acou; Halandur N Bharath; Diana M Sima; Jelle Veraart; Frederik Maes; Uwe Himmelreich; Eric Achten; Sabine Van Huffel

doi:10.1371/journal.pone.0180268

The successive projection algorithm as an initialization method for brain tumor segmentation using non-negative matrix factorization

PLoS One. 2017 Aug 28;12(8):e0180268. doi: 10.1371/journal.pone.0180268. eCollection 2017.

Authors

Nicolas Sauwen^{1

2}, Marjan Acou³, Halandur N Bharath^{1

2}, Diana M Sima^{1

2

4}, Jelle Veraart⁵, Frederik Maes⁶, Uwe Himmelreich⁷, Eric Achten³, Sabine Van Huffel^{1

2}

Affiliations

¹ KU Leuven, Department of Electrical Engineering (ESAT), STADIUS Centre for Dynamical Systems, Signal Processing and Data Analytics, Leuven, Belgium.
² imec, Leuven, Belgium.
³ Ghent University Hospital, Department of Radiology, Ghent, Belgium.
⁴ Icometrix, R&D Department, Leuven, Belgium.
⁵ University of Antwerp, iMinds Vision Lab, Department of Physics, Antwerp, Belgium.
⁶ KU Leuven, Department of Electrical Engineering (ESAT), PSI Centre for Processing Speech and Images, Leuven, Belgium.
⁷ KU Leuven, Department of Imaging and Pathology, Biomedical MRI/MoSAIC, Leuven, Belgium.

Abstract

Non-negative matrix factorization (NMF) has become a widely used tool for additive parts-based analysis in a wide range of applications. As NMF is a non-convex problem, the quality of the solution will depend on the initialization of the factor matrices. In this study, the successive projection algorithm (SPA) is proposed as an initialization method for NMF. SPA builds on convex geometry and allocates endmembers based on successive orthogonal subspace projections of the input data. SPA is a fast and reproducible method, and it aligns well with the assumptions made in near-separable NMF analyses. SPA was applied to multi-parametric magnetic resonance imaging (MRI) datasets for brain tumor segmentation using different NMF algorithms. Comparison with common initialization methods shows that SPA achieves similar segmentation quality and it is competitive in terms of convergence rate. Whereas SPA was previously applied as a direct endmember extraction tool, we have shown improved segmentation results when using SPA as an initialization method, as it allows further enhancement of the sources during the NMF iterative procedure.

MeSH terms

Algorithms
Brain / diagnostic imaging*
Brain Neoplasms / diagnostic imaging*
Humans
Image Processing, Computer-Assisted / methods*
Magnetic Resonance Imaging
Pattern Recognition, Automated / methods*

Grants and funding

NS, DMS and MA received funding from the Research Foundation Flanders (FWO), grant number G.0869.12N. JV received funding from the Henri Benedictus Fellowship of the Belgian American Educational Foundation; Interuniversity Attraction Poles Program (P7/11) initiated by the Belgian Science Policy Office. SVH received funding from the government agency for Innovation by Science and Technology (IWT), grant number IM 135005. SVH and HNB received funding from the European Research Council under the European Union’s Seventh Framework Programme, grant number 339804. SVH and DMS received funding from the European Research Council under the European Union’s Seventh Framework Programme, grant number 316679. Icometrix provided support in the form of salary for author DMS, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.