In this section
@article{Yang:2014:10.1002/mrm.25447,
author = {Yang, G and Raschke, F and Barrick, TR and Howe, FA},
doi = {10.1002/mrm.25447},
journal = {Magnetic Resonance in Medicine},
pages = {868--878},
title = {Manifold Learning in MR spectroscopy using nonlinear dimensionality reduction and unsupervised clustering},
url = {http://dx.doi.org/10.1002/mrm.25447},
volume = {74},
year = {2014}
}
TY - JOUR
AB - PurposeTo investigate whether nonlinear dimensionality reduction improves unsupervised classification of 1H MRS brain tumor data compared with a linear method.MethodsIn vivo single-voxel 1H magnetic resonance spectroscopy (55 patients) and 1H magnetic resonance spectroscopy imaging (MRSI) (29 patients) data were acquired from histopathologically diagnosed gliomas. Data reduction using Laplacian eigenmaps (LE) or independent component analysis (ICA) was followed by k-means clustering or agglomerative hierarchical clustering (AHC) for unsupervised learning to assess tumor grade and for tissue type segmentation of MRSI data.ResultsAn accuracy of 93% in classification of glioma grade II and grade IV, with 100% accuracy in distinguishing tumor and normal spectra, was obtained by LE with unsupervised clustering, but not with the combination of k-means and ICA. With 1H MRSI data, LE provided a more linear distribution of data for cluster analysis and better cluster stability than ICA. LE combined with k-means or AHC provided 91% accuracy for classifying tumor grade and 100% accuracy for identifying normal tissue voxels. Color-coded visualization of normal brain, tumor core, and infiltration regions was achieved with LE combined with AHC.ConclusionPurposeTo investigate whether nonlinear dimensionality reduction improves unsupervised classification of 1H MRS brain tumor data compared with a linear method.MethodsIn vivo single-voxel 1H magnetic resonance spectroscopy (55 patients) and 1H magnetic resonance spectroscopy imaging (MRSI) (29 patients) data were acquired from histopathologically diagnosed gliomas. Data reduction using Laplacian eigenmaps (LE) or independent component analysis (ICA) was followed by k-means clustering or agglomerative hierarchical clustering (AHC) for unsupervised learning to assess tumor grade and for tissue type segmentation of MRSI data.ResultsAn accuracy of 93% in classification of glioma grade II and grade IV, with 100% accuracy in distinguishi
AU - Yang,G
AU - Raschke,F
AU - Barrick,TR
AU - Howe,FA
DO - 10.1002/mrm.25447
EP - 878
PY - 2014///
SN - 0740-3194
SP - 868
TI - Manifold Learning in MR spectroscopy using nonlinear dimensionality reduction and unsupervised clustering
T2 - Magnetic Resonance in Medicine
UR - http://dx.doi.org/10.1002/mrm.25447
VL - 74
ER -
For enquiries about the MRI Physics Collective, please contact:
Mary Finnegan
Senior MR Physicist at the Imperial College Healthcare NHS Trust
Pete Lally
Assistant Professor in Magnetic Resonance (MR) Physics at Imperial College
Jan Sedlacik
MR Physicist at the Robert Steiner MR Unit, Hammersmith Hospital Campus