The fast Padé transform for noisy magnetic resonance spectroscopic data from the prostate: potential contribution to individualized prostate cancer care

Abstract

Magnetic resonance spectroscopy (MRS) and spectroscopic imaging (MRSI) can enhance prostate cancer diagnostics, but have limitations that are largely due to reliance upon conventional Fourier-based signal processing. MRS of the prostate is exceedingly difficult, due to high spectral density with numerous multiplet resonances. We apply advanced signal processing methods through the fast Pade transform (FPT) to time signals generated according to in vitro MRS data as encoded from normal glandular and stromal prostate as well as from prostate cancer. Random Gauss-distributed zero mean noise is added to the noise-free time signal. The high resolution capabilities are demonstrated: at short partial signal lengths $$(N_{\mathrm{P}})$$ , converged total and component spectra from the prostate are generated by the FPT. In comparison, Fourier-based processing provides only rough, uninformative total shape spectra. Detailed analysis reveals the powerful, complementary features of the two variants $$\hbox {FPT}^{(\pm )}$$ of the FPT in separating the copious spurious content from genuine resonances. At short $$N_{\mathrm{P}}$$ , the FPT resolved all the physical resonances, including multiplets and closely overlapping peaks of different metabolites, exactly reconstructing all the input spectral parameters, from which the metabolite concentrations were precisely computed. Systematic study of noise-corrupted time signals in the controlled setting is a critical step in benchmarking the FPT for clinical applications. We discuss how these results could increase the diagnostic accuracy of MRS and MRSI of the prostate, and how this could contribute to a more individualized care of patients with prostate cancer.