Introducing Spatial Prior Knowledge Into ProFit-1D – A Spectral Fitting Strategy for MRSI

Motivation: Proton MR Spectroscopy Imaging requires specialized quantification software. Several packages are available; however, they can give significantly different outputs, which can cause errors in decision making. Goal(s): We theorize that quantification algorithms can be improved by introducing spatial-based constraints to the model alongside the traditional spectral constraints. Approach: We developed an alternative modality to our quantification software which incorporates the spatial relationship for confounding fitting parameters as a constraint for four human datasets. Improved fit performance is quantified using CRLB error measurements. Results: Reduced error estimates were observed for all metabolites. Metabolite distribution stayed consistent between fitting strategies and expected tissue distribution. Impact: The results of this study are promising for improving the interpretability of MRSI data. Our approach managed to improve the reproducibility of our quantification results and is translatable to a wide range of spectral fitting strategies.