Derivative shape estimations with resolved overlapped peaks and reduced noise for time signals encoded by NMR spectroscopy with and without water suppression

Abstract

AbstractDerivative estimation in magnetic resonance spectroscopy (MRS) possesses several attractive features. It has the ability to enhance the inaccessible spectral details when time signals encoded by MRS are analyzed by nonderivative shape estimators. These unfolded subtle spectral features can be diagnostically relevant in differentiating between healthy and diseased tissues. Within the realm of shape estimators, the prerequisite for the success of MRS in the clinic is reliance upon accurate derivative signal processing. However, derivative processing of encoded time signals can be very challenging. The reason is that such spectra may suffer from severe numerical instabilities since even small perturbations (noise) in the input data could produce large errors in the predicted output data. Nevertheless, it is presently demonstrated that this obstacle can be surmounted by an adaptive optimization. The benefit is simultaneously increased resolution and reduced noise in quantitatively interpretable lineshapes. The illustrative spectra are reconstructed from time signals encoded by proton MRS with and without water suppression.