Quantification by the fast Padé transform of magnetic resonance spectroscopic data encoded at 1.5 T: implications for brain tumor diagnostics

Abstract

Magnetic resonance spectroscopy (MRS) and spectroscopic imaging (MRSI) are widely applied in brain tumor diagnostics. Notwithstanding the important achievements, there are major problems in MRS and MRSI for neuro-oncology due to inadequate signal processing. In this proof-of-concept study, we apply the fast Pade transform (FPT) to MRS time signals of short signal length $$(N = 512)$$ encoded from a standard test phantom head on a 1.5 T MR scanner. The employed phantom head contains a number of major metabolites that are also detectable by way of MRS scans of in vivo human brain. Detailed parametric analysis is performed by which quantification is achieved with full reliability. The effectiveness of the FPT-based signal-noise separation is confirmed in handling these experimentally-encoded MRS time signals. The $$\hbox {FPT}^{(+)}$$ variant of the FPT shows particularly favorable capabilities for distinguishing the genuine from the numerous spurious resonances in the presence of the unavoidable noise, as is typical of data encoded at MR clinical scanners. Exhaustive examination of the convergence process is performed to confirm the stability of the reconstructed spectral parameters. Statistical analysis, as a procedure called “parameter averaging” demonstrates the accuracy and precision of the reconstructed complex-valued fundamental frequencies and amplitudes. This includes the dense regions of the spectrum, containing several very small and/or very closely overlapping resonances. Overall, the obtained results confirm that Pade-optimized MRS and MRSI can be directly used in our fully automated matlab program for in vivo encoded data with complete fidelity within neuro-oncology, with anticipated benefit for brain tumor diagnostics.