Improving the diagnostic yield of magnetic resonance spectroscopy for pediatric brain tumors through mathematical optimization

Abstract

Brain tumors are the leading cause of cancer-related deaths among children. Increasing attention in pediatric neuro-oncology has been given to magnetic resonance spectroscopy (MRS). Notwithstanding the important achievements, the potential of MRS for pediatric neuro-oncology is yet to be realized. This is largely due to reliance upon inadequate signal processing methods based upon the fast Fourier transform (FFT) plus fitting. Herein, we applied an advanced signal processor, the fast Pade transform (FPT) to MRS time signals encoded in vivo from a glioma in a pediatric patient, using a 1.5T scanner. Three echo times (TE) were used: 22, 136 and 272 ms. Compared to those from the FFT, the total shape spectra from the FPT were better resolved. The most striking advantages of the FPT lie in its parametric capabilities from which component spectra were generated. At the shortest TE, for which spectral density is greatest, the FPT resolved the numerous overlapping resonances, delineating myoinositol and other short-lived metabolites. The FPT resolved components of diagnostically-important peaks centered at chemical shifts near 2.0, 3.0 and 3.2 parts per million. The latter includes not only free choline, but also the cancer biomarker, phosphocholine. An information-preserving procedure for suppression of residual water is introduced and validated, via windowing using a step function. This investigation demonstrates that mathematical optimization through the FPT can be successfully applied to MRS time signals encoded in vivo from pediatric brain tumors using standard clinical scanners at 1.5T. Improved diagnostic yield within pediatric neuro-oncology is anticipated thereby.