Classification of brain tumour 1H MR spectra: Extracting features by metabolite quantification or nonlinear manifold learning?

Proton magnetic resonance spectroscopy (1H MRS) provides non-invasive information on brain tumour biochemistry. Many studies have shown that 1H MRS can be used in an objective decision support system, which gives additional diagnosis and prognostic information to the data obtained using conventional radiological modalities. Fully automatic analyses of 1H MRS have been previously applied and can be separated into two types: (i) model dependent signal quantification followed by pattern recognition (PR), or (ii) model independent PR methods. However, there is not yet a consensus as to the best techniques of MRS post-processing or feature extraction to be used for optimum classification. In this study, we analysed the single-voxel MRS acquisitions of 74 patients with histologically diagnosed brain tumours. Our classification results show that the model independent nonlinear manifold learning method can produce superior results to those of using model dependent metabolite quantification.

[1]  Franklyn A Howe,et al.  1H MR spectroscopy of brain tumours and masses , 2003, NMR in biomedicine.

[2]  J R Griffiths,et al.  Linear discriminant analysis of brain tumour 1H MR spectra: a comparison of classification using whole spectra versus metabolite quantification , 2007, NMR in biomedicine.

[3]  Vanhamme,et al.  Improved method for accurate and efficient quantification of MRS data with use of prior knowledge , 1997, Journal of magnetic resonance.

[4]  B. Scheithauer,et al.  The New WHO Classification of Brain Tumours , 1993, Brain pathology.

[5]  E. Fuster-Garcia,et al.  Classification of single‐voxel 1H spectra of brain tumours using LCModel , 2012, NMR in biomedicine.

[6]  Paulo J. G. Lisboa,et al.  Non-negative matrix factorisation methods for the spectral decomposition of MRS data from human brain tumours , 2012, BMC Bioinformatics.

[7]  Kevin P. Murphy,et al.  Machine learning - a probabilistic perspective , 2012, Adaptive computation and machine learning series.

[8]  Mikhail Belkin,et al.  Laplacian Eigenmaps for Dimensionality Reduction and Data Representation , 2003, Neural Computation.

[9]  B. Efron.  Estimating the Error Rate of a Prediction Rule: Improvement on Cross-Validation , 1983 .

[10]  J. Havel,et al.  Artificial Neural Networks for Classification in Metabolomic Studies of Whole Cells Using 1H Nuclear Magnetic Resonance , 2010, Journal of biomedicine & biotechnology.

[11]  J R Griffiths,et al.  Pattern recognition of MRSI data shows regions of glioma growth that agree with DTI markers of brain tumor infiltration , 2009, Magnetic resonance in medicine.

[12]  S. Provencher.  Estimation of metabolite concentrations from localized in vivo proton NMR spectra , 1993, Magnetic resonance in medicine.

[13]  Zhiqiang Weng,et al.  Communication An efficient algorithm for automatic phase correction of NMR spectra based on entropy minimization , 2002 .

[14]  Martin T. Hagan,et al.  Neural network design , 1995 .

[15]  C. Metz.  ROC Methodology in Radiologic Imaging , 1986, Investigative radiology.

[16]  M. Sdika,et al.  Time‐domain semi‐parametric estimation based on a metabolite basis set , 2005, NMR in biomedicine.

[17]  Theodoros N. Arvanitis,et al.  A constrained least‐squares approach to the automated quantitation of in vivo 1H magnetic resonance spectroscopy data , 2011, Magnetic resonance in medicine.