Map control
TDFDFit workflow
Applying predefined TDFDFit models to spectra
This page describes how an existing prior-knowledge model can be applied to selected MRS or MRSI spectra in spectrIm-QMRS, and how fitted parameters can be reviewed as spectra, residuals, and maps.
Where the model comes from
A TDFDFit model may be created automatically for clinical use with the automatic TDFDFit model creator, or manually by a spectroscopist using the expert TDFDFit modelling frame. Once saved, the model becomes a reusable prior-knowledge template for datasets with matching acquisition conditions.
Applying a model is the practical quantification step. The model is selected in the main spectrIm-QMRS interface, applied to the selected spectra or voxels, and the fitted results are collected for inspection and map generation.
Select and refresh the model
After creating or selecting a prior-knowledge model, return to the main spectrIm-QMRS GUI. If a new model has just been created, refresh the toolbar dropdown boxes so the model list reflects the available models.
After refreshing, the selected default model is visible in the fitting
controls. In the legacy example, the model is named
SpectralModel.tfm. In version 3.0.1 Alpha this model may
instead be a model produced by the automatic model creator, by the
ILV basis-set workflow, or by the expert modelling frame.
To be checked manually: these screenshots originate from the older website. They document the workflow accurately, but a current version 3.0.1 Alpha screenshot should replace them when the final GUI state is stable.
Run fitting on selected voxels
The model is then applied to the selected spectra or MRSI voxels. In the historical example, selected voxels are marked blue and the TDFDFit action is launched from the main toolbar/menu.
In version 3.0.1 Alpha, this step connects conceptually to Parallel pthreads-TDFDFit CPU fitting: the model remains the prior-knowledge description, while the fitting engine performs the computational work over the selected spectra.
During fitting, progress information is displayed. The original page described an upper progress indicator for active threads and a lower progress indicator for completed quantification steps. The exact implementation has evolved, but the practical meaning remains the same: the user can monitor how far the fitting process has progressed.
When fitting terminates, spectrIm-QMRS collects the fitted results. The fitted sum spectrum can be overlaid on the measured sum spectrum, allowing the user to judge visually whether the prior-knowledge model and the data are in reasonable agreement.
Analyze fitted data
After fitting, a compact map-quantification dialog can be used to select the spectral component and fitted parameter to display. This is the bridge from spectral fitting to metabolite or parameter maps.
Metabolite or component
Fitted parameter
The upper dropdown lists model components such as metabolites or resonance groups. The second dropdown selects the fitted parameter to display, such as area, linewidth, phase, or other model parameters. The current version 3.0.1 Alpha dialog is described separately, including the real-time auto-scale clipping controls for handling outlier voxels: Map Quantification Results.
Create metabolite and parameter maps
Selecting a metabolite component and a parameter displays a map over the spectroscopy grid. These maps are not a substitute for reviewing individual fits, but they are essential for spatial interpretation of MRSI results.
A second example is the Cho-CH3 peak area map. Such maps can reveal spatial differences in the fitted metabolite-related signal.
Ratio maps can also be generated, for example a Cho-CH3/NAA ratio map. Ratio maps are often clinically interesting, but their meaning depends on model quality, preprocessing, voxel selection, and the biological question.
Parameter maps such as Lorentzian damping or zero-order phase can be used for quality assessment and for identifying spatial patterns that may reflect linewidth, phase, or fitting behavior rather than pure metabolite concentration.
How this fits into version 3.0.1 Alpha
This page describes the application step: taking a predefined model and using it to fit measured spectra. In version 3.0.1 Alpha, models may be created with the automatic model creator for routine clinical work or with the expert modelling frame for spectroscopy development.
The actual fitting of larger selections or MRSI volumes is the role of the parallel TDFDFit CPU implementation. The model defines what is fitted; the parallel fitting engine determines how the selected spectra are processed efficiently.