13C MRS provides a well-validated method of measuring invivo glycogen levels [1] offering significant advantages over invasive techniques. However, sensitivity is low due to low natural abundance of 13C (~1%) and gyromagnetic ratio (~1/4 1H). This is overcome by acquiring from a large volume using surface coils with many averages [2, 3].

1H decoupling increases sensitivity by collapsing the coupled peak[3]. However, an inhomogeneous field (B₂) may leads to partial decoupling with a residual coupling constant J_r = JΔω/γB₂ [4]. Furthermore, increased power leads to greater repetition times (TR) and reduced signal averaging. The aim of this study was to explore these effects to optimize acquisition protocol for invivo 13C glycogen MRS.

A Philips Acheiva 3T system was used with a Pulseteq 13C surface coil (glucose experiment – 12cm single loop, glycogen experiments – 8cm single loop) with 1H quadrature decoupling. All spectra were phase corrected and line broadened. Phantom data was fitted using AMARES in jMRUI and Invivo spectra were analysed using lsqcurvefit in Matlab with coupled peaks defined as two Lorentzians of equal linewidth and amplitude, and decoupled peaks as one Lorentzian.

Investigating J_r = JΔω/γB₂: A 13C labelled glucose vial was used to provide large SNR and uniform and 13C MRS acquired, first varying the decoupling frequency (Δν) at constant power (fig 1), and then varying with broadband decoupling (1024datapoints, 7kHz bandwidth, TR ~3500ms, 48averages).

Decoupling efficiency: A 1.4l natural abundance 200 mM glycogen phantom was used to calculate T₁ from TR varying MRS (256 datapoints, bandwidth = 8kHz, 2000 averages, TR = 56, 100, 200, 300 and 500ms). Fully recovered MRS was then acquired non-decoupled (TR=300 ms, 2048datapoints, 2000averages, time = 9:55) and under 3 decoupling regimes for comparison: (1) minimum decoupling, same averages as non-decoupled (TR=400 ms, 1024 datapoints, 3 μT WALTZ decoupling, 2000 averages, time =13:20) , (3) maximum decoupling, same scan time as non-decoupled (TR=7083 ms, 1024 datapoints, 20μT WALTZ decoupling, 84 averages, time =10:09), and (4) a compromise between min and max decoupling, same scan time as non-decoupled (TR=600ms, 1024 datapoints, 5μT narrow band decoupling centred at glycogen, 1000 averages, time = 10:01).

Invivo scanning: In order to test the feasibility of fitting invivo coupled spectra, 3 consented subjects were scanned with no decoupling at short TRs (300ms) over a 20 minute time frame (4000 averages). Decoupled spectra were also acquired for: Subject 1 – maximum broadband WALTZ decoupling, = 20 μT (TR = 5000ms, 248 averages); Subject 2 – medium broadband WALTZ decoupling, = 9 μT (TR = 2000ms, 600 averages); Subject 3 – low narrow band decoupling, = 5 μT (TR = 505ms, 2400 averages).

Investigating J_r: Figure 1 and 2 shows residual J coupling at varying frequency and confirming the linear relationhsip , but showing that spectra at varying were too complicated to fit. In larger samples containing inhomogeneities the net spectra would consist of a sum of these spectra.

Decoupling efficiency: Glycogen T₁ was 45 ms, allowing full recovery at TR>225 ms. Figure 2 shows the glycogen fitting for each decoupling regime. In the case of no decoupling, a coupling constant = 176 Hz was measured and the total SNR on the fit was 3.23 (fig 2a). No peak was found for maximum decoupling (fig 2b). For minimum decoupling multiple overlapping peaks were seen (fig 2c) and best fitting was achieved using 6 peaks giving an overall SNR of 1.3. Finally, medium decoupling gave a single peak (fig 2d)with SNR = 2.61

Invivo scanning: Non-decoupled spectra gave well resolved glycogen peaks with average linewidth = 143 ± 7 Hz and = 168 ± 3 Hz. The peak areas were 1.5 a.u, 1.1 a.u and 0.7 a.u. for subject 1, 2 and 3 respectively (fig 3) with an SNR of 2.2, 1.6 and 1.1. All decoupling methods produced spectra in the same acquisition time where the glycogen peak could not be fitted due to low overall SNR.