Reproducibility of activation in four motor paradigms. An fMRI study

Most functional imaging data are collected in single session experiments; little is known about the reproducibility or test-retest reliability of the activation patterns found in these experiments. In our study, 15 healthy volunteers performed four simple motor-paradigms ("Hand", "Foot", "Mouth" and "Tongue") for functional magnetic resonance imaging (fMRI) in 3 sessions on different days. Reproducibility of activations in four anatomical regions (pre- and postcentral gyri, paracentral lobule and the supplementary motor area) was measured in terms of voxels active in all sessions (common voxels) relative to voxels active in single sessions, giving reliability coefficients from 0 to 1. Two significance levels were used to identify active voxels. Reproducibility of activations was highest for foot and hand movements in the primary motorsensory areas; reliability coefficients were in the range of 0.62 to 0.78. Activations for mouth movements showed a very poor reproducibility. Application of the more stringent statistical threshold always led to a reduction of reproducible voxels. Reliability of fMRI data is not only a theoretical issue, but is of special practical importance in clinical settings such as integration of fMRI into neuronavigation for neurosurgical planning. Much care has to be taken if only single session data are available for interpretation.