Strategic Steps for Advanced Molecular Imaging with Magnetic Resonance-Based Diagnostic Modalities

Abstract

With the rapidly-expanding sophistication in our understanding of cancer cell biology, molecular imaging offers a critical bridge to oncology. Molecular imaging through magnetic resonance spectroscopy (MRS) can provide information about many metabolites at the same time. Since MRS entails no ionizing radiation, repeated monitoring, including screening can be performed. However, MRS via the fast Fourier transform (FFT) has poor resolution and signal-to-noise ratio (SNR). Moreover, subjective and non-unique (ambiguous) fittings of FFT spectra cannot provide reliable quantification of clinical usefulness. In sharp contrast, objective and unique (unambiguous) signal processing by the fast Padé transform (FPT) can increase resolution and retrieve the true quantitative metabolic information. To illustrate, we apply the FPT to in vitro MRS data as encoded from malignant ovarian cyst fluid and perform detailed analysis. This problem area is particularly in need of timely diagnostics by more advanced modalities, such as high-resolution MRS, since conventional methods usually detect ovarian cancers at late stages with poor prognosis, whereas at an early stage the prognosis is excellent. The reliability and robustness of the FPT is assessed for time signals contaminated with varying noise levels. In the presence of higher background noise, all physical metabolites were unequivocally identified and their concentrations precisely extracted, using small fractions of the total signal length. Via the "signal-noise separation" concept alongside the "stability test", all non-physical information was binned, such that fully denoised spectra were generated. These results imply that a reformulation of data acquisition is needed, as guided by the FPT in MRS, since a small number of short transient time signals can provide high resolution and good SNR. This would enhance the diagnostic accuracy of MRS and shorten examination times, thereby improving efficiency and cost-effectiveness of this high throughput cancer diagnostic modality. Such advantages could be particularly important for more effective ovarian cancer detection, as well as more broadly for improved diagnostics and treatment within oncology.

Padé optimization of noise-corrupted magnetic resonance spectroscopic time signals from fibroadenoma of the breast

“…By varying the partial signal length and also by adding yet more noise, all the true metabolites, including those with very small amplitudes could be confidently identified by their stability. Spurious resonances were unstable with even a minimal change in partial signal length or noise level, and were binned for discarding, whereas all the stable and thus true metabolic information was kept in a denoised spectrum [34,99]. 1.7 Performance of the FPT for noise-corrupted MRS time signals from breast fibroadenoma: the present study Based upon the results described above, it is justifiable to apply the FPT to noisecorrupted time signals from the breast.…”

“…Moreover, some of the spurious resonances showed near zero amplitudes rather than actual zero amplitudes [34,99]. Since genuine metabolites might conceivably be present at very low concentrations, their peak amplitudes could be extremely small.…”

“…Functional as well as molecular imaging have much to offer in this regard [34][35][36]. In particular, magnetic resonance (MR)-based modalities, including diffusion-weighted MRI could be appropriate for breast cancer screening/surveillance, since, as noted, MR is free from ionizing radiation.…”

“…We use RMS as a quantifier of noise because this parameter reduces both the bias relative to the actual (sought) value as well as the variance of noise. The RMS is a measure of the signal dynamics, and converts signal oscillations to variations of the power of the signal across the given bandwidth [34].…”

“…We proceeded to apply the FPT to MRS time signals associated with ovarian cancer in the presence of added noise [34,98,99]. With the addition of extraneous noise (σ = 0.01156 RMS, where RMS is the root-mean-square of the noisefree time signal, RMS = (N −1 N −1 n=0 |c n | 2 ) 1/2 , convergence was achieved at N /8 = 128(N = 1024), with accurate reconstruction of the spectral parameters for all twelve genuine metabolites.…”