Feature Selection Framework for White Matter Fiber Clustering Based on Normalized Cuts
Due to its ability to automatically identify spatially and functionally related white matter fiber bundles, fiber clustering has the potential to improve our understanding of white matter anatomy. The normalized cuts (NCut) criterion has proven to be a suitable method for clustering fiber tracts. In this work, we show that the NCut value can be used for unsupervised feature selection as a measure for the quality of clustering. We further present a method how feature selection can be improved by penalizing spatially illogical clustering results, which is achieved by employing the Silhouette index for a fixed set of geometric features.
KeywordsFeature Selection Cluster Result Spectral Cluster Silhouette Index Unsupervised Feature Selection
Unable to display preview. Download preview PDF.
- 1.Brun A, Knutsson H, Park HJ, et al. Clustering fiber traces using normalized cuts. Med Image Comput Assist Interv. 2004; p. 368–75.Google Scholar
- 4.Neher PF, Laun FB, Stieltjes B, et al. Fiberfox: an extensible system for generating realistic white matter software phantoms. Comput Diffusion MRI Brain Connectivity. 2014; p. 105–13.Google Scholar
- 3.Goldberg DE. Genetic Algorithms in Search, Optimization and Machine Learning. 1st ed. Addison-Wesley Longman Publishing Co., Inc.; 1989.Google Scholar
- 6.Ramirez EH, Brena R, Magatti D, et al. Probabilistic metrics for soft-clustering and topic model validation. Proc IEEE WIC ACM Int Conf Web Intell Agent Technol. 2010;1:406–12.Google Scholar