Toward fully automatic estimation of in vivo 31P spectra

Fitting a model to an experimental spectrum is a difficult nonlinear estimation problem. The solution presented here is to start an iterative search procedure sufficiently close to the optimal model parameter set. This is achieved by providing tissue-dependent a priori peak information and by a novel correlation method to get good primary estimates of the resonance and phase offset parameters. The resulting estimation procedure is fully automatic and has proven to be robust for 31P data.