Manifold Learning in MR spectroscopy using nonlinear dimensionality reduction and unsupervised clustering
Magnetic Resonance in Medicine
PURPOSE: To investigate whether non-linear dimensionality reduction improves unsupervised classification of 1 H MRS brain tumour data compared to a linear method. METHODS: In-vivo single voxel 1 H MRS (55 patients) and 1 H MRSI (29 patients) data were acquired from histopathologically diagnosed gliomas. Data reduction by Laplacian eigenmaps (LE) or independent component analysis (ICA) was followed by k-means clustering or agglomerative hierarchical clustering (AHC) for unsupervised learning to
... ssess tumour grade and for tissue type segmentation of MRSI data. RESULTS: An accuracy of 93% in classification of glioma Grade-II and Grade-IV, with 100% accuracy in distinguishing tumour and normal spectra, was obtained by LE with unsupervised clustering, but not with the combination of k-means and ICA. With 1 H 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 tumour grade and 100% accuracy for identifying normal tissue voxels. Colour-coded visualisation of normal brain, tumour core and infiltration regions was achieved with LE combined with AHC. CONCLUSION: The LE method is promising for unsupervised clustering to separate brain and tumour tissue with automated colour-coding for visualisation of 1 H MRSI data after cluster analysis.