Corroboration of the shape by parameter estimations in the derivative fast Padé transform with applications to ovarian MRS

Abstract

The non-parametric and parametric derivative fast Padé transform (dFPT) were applied to time signals encoded by in vitro magnetic resonance spectroscopy (MRS) from benign and malignant ovarian cyst fluid. Several chemical shift regions in which diagnostically-important metabolites resonate were scrutinized. Consecutive derivative orders were inspected until convergence was attained, with concordance between the two versions of the dFPT. In most of the examined chemical shift regions, this occurred with the third derivative order, which was sufficient for the resonances to appear as tall, thin peaks, emerging from an entirely flat baseline with clearly identifiable integration limits. Thereby, the prominent lactate doublet resonating at \(\sim \)1.41 parts per million (ppm), as well as many other, much less abundant, surrounding metabolites could be unambiguously identified and quantified. This was likewise the case for numerous other recognized and potential cancer biomarkers in the other examined aliphatic chemical shift regions. In the extremely dense spectral region below \(\sim\)1.0 ppm, heretofore presenting major assessment difficulties in reported in vitro MRS studies, convergence was attained at \((m=4)\). Besides the demonstrated concordance of the two completely different computational algorithms, the present results are fully amenable to clinical interpretation. This is an essential precondition to justify further application of the dFPT to urgent clinical problems such as timely and accurate ovarian cancer detection via MRS.