Enhancing the Detection of BOLD Signal in fMRI by Reducing the Partial Volume Effect

Purpose. To investigate the advantages of reducing the partial volume effect (PVE) to enhance the detection of the BOLD signal in fMRI. Methods. A linear phase term was added in k-space to obtain half-voxel shifting of 64 × 64 T(2)*-weighted echo-planar images. Three sets of image data shifted in the x, y, and diagonal direction, respectively, are combined with the original 64 × 64 data to form the 128 × 128 voxel-shifted interpolated data. Results. A simulation of a synthetic fMRI dataset shows that the voxel-shifted interpolation (VSI) can increase the t-score up to 50% in single-voxel activations. An fMRI study (n = 7) demonstrates that 20.4% of the interpolated voxels have higher t-scores than their nearest neighboring voxels in the original maps. The average increase of the t-score in these interpolated voxels is 13.3%. Conclusion. VSI yields increased sensitivity in detecting voxel-size BOLD activations, improved spatial accuracy of activated regions, and improved detection of the peak BOLD signal of an activated region. VSI can potentially be used as an alternative to the high-resolution fMRI studies in which reduction in SNR and increase in imaging time become prohibitive.