| Summary: | The 38 papers of this Neuroimage special issue on brain parcellation and segmentation provide a snapshot of a vibrant area of neuroimaging research. Parcellation, segmentation, clustering, community detection, etc., are different names for techniques aimed at dividing a collection of examples into subsets with similar statistical properties. Although clustering methods are used to solve seemingly disparate problems in neuroimaging, they all share the common goal of providing a broad understanding of the data, while abstracting away less relevant finer-grained information. So when the time came to write this editorial, we could not resist using a cluster analysis to organize these 38 papers into data-driven categories. We used a bag-of-words approach implemented in scikitlearn (Pedregosa et al., 2011) to measure the pairwise similarity between the abstracts of the papers. Using hierarchical clustering, we subdivided the papers into 7 categories (Fig. 1a) and identified the 20 most relevant words for each category (Fig. 1b).1 We used these categories in the following sections to provide a brief synopsis of the special issue's content.
|
|---|