A novel eigenvector-based technique for spectral estimation of time-domain data in medical imaging

Abstract

The use of a complex MUSIC (multiple signal classification) algorithm to signal average MR spectroscopic data from a 1-cm/sup 3/ voxel of diseased brain tissue for only five min and obtain diagnostically useful studies is discussed. A complex eigenvector-based method for performing spectral analysis of time-domain data independent of the signal-to-noise ratio is demonstrated. The implementation of the procedure requires no preprocessing of the time-domain data record. The technique is well suited for magnetic resonance spectroscopy and imaging, where the signal available from small regions corresponding to areas of diseased tissue in patients presenting for diagnosis is always dominated by the Johnson noise present in the receiver circuit.<>