Abstract #2529
CFIT: A novel circle-fitting approach to spectral analysis
Ouwerkerk R, Bottomley P, Gabr R
Johns Hopkins University
Spectral analysis of in vivo spectra is problematic due to low signal-to-noise ratio, overlapping peaks and intense baseline artefacts. We present a new frequency-domain technique for MRS analysis, CFIT, wherein the circular trajectories that result when spectral peaks are projected onto the complex plane, are fitted with active circles. This eliminates the need for phasing spectra. The baseline is dealt with using active contours (snakes). CFIT exhibited the lowest failure rate compared to standard time- and frequency-domain methods applied to 98 real human chest and heart 31P MRS data sets. Its accuracy was comparable to our best current method.