Simple baseline correction for 1H MRSI data of the prostate

Abstract

Proton magnetic resonance spectroscopic imaging of the prostate can detect and localise tumour tissue in patients with prostate cancer. Pattern recognition methods have the potential to discriminate tumour from normal tissue using the spectral patterns of magnetic resonance spectroscopic imaging data thus providing a powerful tool for diagnosis and monitoring treatment of prostate cancer. Prostate magnetic resonance spectroscopic imaging data acquisition has been optimized for a flat baseline, including a long echo time and the use of water and fat suppression pulses. Despite this acquisition method, lipid resonances may still occur particularly at the margins of the prostate. Correcting the baseline by removing lipid artifacts is, therefore, a necessary processing step. We propose a simple baseline correction method for prostate magnetic resonance spectroscopic imaging through subtraction of a single simulated-lorentzian resonance. This method is shown to restore flat baselines to a test set of spectra and compares favorably with a state-space modeling approach.