Multidimensional wavelet analysis of functional magnetic resonance images

Abstract

Analysis of functional magnetic resonance imaging (fMRI) data requires the application of techniques that are able to identify small signal changes against a noisy background. Many of the most commonly used methods cannot deal with responses which change amplitude in a fashion that cannot easily be predicted. One technique that does hold promise in such situations is wavelet analysis, which has been applied extensively to time-frequency analysis of nonstationary signals. Here a method is described for using multidimensional wavelet analysis to detect activations in an experiment involving periodic activation of the visual and auditory cortices. By manipulating the wavelet coefficients in the spatial dimensions, activation maps can be constructed at different levels of spatial smoothing to optimize detection of activations. The results from the current study show that when the responses are at relatively constant amplitude, results compare well with those obtained by established methods. However, the technique can easily be used in situations where many other methods may lose sensitivity.