Time-domain fitting software widely used for multi-nuclear (X-nuclei) and short-TE proton Magnetic Resonance Spectroscopy.
jMRUI (Java-based Magnetic Resonance User Interface) is a comprehensive software package designed specifically for processing, analyzing, and quantifying MRS data. While many tools operate in the frequency domain (fitting peaks in a spectrum), jMRUI fundamentally operates in the time domain (fitting the Free Induction Decay, or FID, directly).
Why Time Domain? Fitting the FID directly is often preferred for X-nuclei (like 31P or 23Na) or very short echo-time 1H scans. These scans often feature very fast signal decay (short T2*), leading to broad lineshapes and overlapping peaks in the frequency domain. Furthermore, early time points in the FID often contain artifacts from the RF pulse or hardware. Time domain fitting allows for easily truncating these first few points without causing complex baseline distortions in the frequency domain.
AMARES (Advanced Method for Accurate, Robust and Efficient Spectral fitting) is the flagship fitting algorithm within jMRUI. It models the FID as a sum of exponentially damped sinusoids.
AMARES excels because it allows the user to inject robust Prior Knowledge into the fitting process. If you know certain physical properties of the molecules you are scanning, you can constrain the algorithm to produce biologically realistic results, preventing it from fitting noise.
Load the raw spectrometer data. Perform necessary preprocessing steps such as zero-filling, apodization (e.g., applying a Gaussian or Exponential filter to improve SNR at the expense of resolution), and phase correction (zero-order and first-order) so the real part of the spectrum is purely absorptive.
For 1H MRS, the water signal is often massive and overlaps with metabolites. HLSVD (Hankel Lanczos Singular Value Decomposition) is a non-interactive time-domain method used specifically to model and subtract dominant nuisance signals (like water or fat) before the main fitting occurs.
Switch to the frequency-domain view to visually identify the peaks of interest. Manually select starting estimates for the frequency and amplitude of the metabolites you wish to quantify.
Open the AMARES dialogue. Group peaks together (e.g., grouping the three ATP resonances) and apply constraints. Define the lineshape model (usually Lorentzian, Gaussian, or Voigt) based on the expected decay profile.
Run the AMARES algorithm. Evaluate the results by examining the Residual (Original Signal minus the Fitted Model). A good fit will leave a residual that looks like pure random noise without any structured peaks remaining.
Analyzing Carbon-13 data in jMRUI presents unique challenges, primarily due to low natural abundance and complex J-coupling patterns. Here is a specialized approach for 13C:
If the acquisition used 1H decoupling (e.g., WALTZ-16), ensure that the Nuclear Overhauser Effect (NOE) enhancement is accounted for during final quantification, as it artificially increases the 13C signal.
13C spectra, especially muscle/liver glycogen, often suffer from low Signal-to-Noise Ratio (SNR) and broad lineshapes. Strong apodization (e.g., a heavy exponential filter) is often required during pre-processing to stabilize the AMARES fit, even though it sacrifices spectral resolution.
Because 13C nuclei couple with adjacent protons (if undecoupled) or other 13C nuclei (in enriched studies), peaks often split into multiplets (doublets, triplets). In AMARES, you must build robust prior knowledge: