Tinnitus is a heterogeneous disorder with respect to its etiology and phenotype. Thus, the identification of sub-types implicates high relevance for treatment recommendations. For this aim, we used cluster analysis of patients for which clinical data, positron-emission tomography (PET) data and voxel-based morphometry (VBM) data were available. 44 patients with chronic tinnitus were included in this analysis. On a phenotypical level, we used tinnitus distress, duration, and laterality for clustering. To correct PET and VBM data for age, gender, and hearing, we built up a design matrix including these variables as regressors and extracted the residuals. We applied Ward's clustering method and forced cluster analysis to divide the data into two groups for both imaging and phenotypical data. On a phenotypical level the clustered groups differed only in tinnitus laterality (uni- vs. bilateral tinnitus), but not in tinnitus duration, distress, age, gender, and hearing. For grey matter volume, groups differed mainly in frontal, cingulate, temporal, and thalamic areas. For glucose metabolism, groups differed in temporal and parietal areas. The correspondence of classification was near chance level for the interrelationship of all three data set clusters. Thus, we showed that clustering according to imaging data is feasible and might depict a new approach for identifying tinnitus sub-types. However, it remains an open question to what extent the phenotypical and imaging levels may be interrelated. This article is part of a Special Issue entitled: Tinnitus Neuroscience.
Copyright © 2012 Elsevier B.V. All rights reserved.