Comparison of MRI Spectroscopy software packages performance and application on HCV-infected patients’ real data

Abstract

1H MRS is conceived as a pioneer methodology for brain metabolism inspection and health status appraisal. Post-processing interventions are required to obtain explicit metabolite quantification values from which to derive diagnosis. On the grounds of addressing and covering such operation, multiple software packages have been recently developed and launched leading to an amorphous assortment of spectroscopic image processing tools, with lack of standardization and regulation. The current study thereby intends to judge the coherence and consistency of compound estimation outputs in terms of result variability by intercorrelation and intracorrelation analyses between appointed programs, being LCModel, Osprey, TARQUIN, and spant toolbox. The examination is performed on a 83-subject SVS short-TE 3T SIEMENS PRESS spectroscopic acquisitions’ collection, including healthy controls and HCV-infected patients assisted with DAA treatment. The analytical core of the project assesses software performance through the creation of a Python script in order to automatically compute and display the results sought. The statistical tests providing enough information to draw substantial conclusions stem from extraction of coefficient of determination (R2 ), Pearson’s coefficient (r), and intraclass correlation coefficient (ICC) together with representation of boxplots, rainclouds, and scatter plots easing data visualization. A clinical implementation is also entailed on the same basis, whose purpose is to reveal actual DAA treatment effect on HCV-infected patients by means of metabolite concentration alteration and hypothetical restoration. Conclusions declare evident and alarming variability among MRS platforms compromising the rigor, sharpness and systematization demanded in this discipline since quantification results hold incoherences, although they do not seem to affect or oppose medical determinations jeopardizing patient’s health. However, it would be interesting to extend the analysis to a greater cohort of subjects to reinforce and get to more solid resolutions.