Study on multi-parametric texture analysis for quantifying brain magnetic susceptibility in patients with Parkinson′s disease

Objective·To quantify brain iron content in Parkinson′s disease (PD) patients by using quantitative susceptibility mapping (QSM) based on phase linearity fitting. Combined with texture analysis methods, the magnetic susceptibility distribution characteristics of gray matter nuclei in PD patients were quantitatively analyzed with multiple parameters and dimensions, and the sensitivity of texture features was evaluated with clinical scoring.Methods·Quantitative susceptibility images from 20 PD patients and 20 healthy controls (HC) were analyzed retrospectively. Regions of interest in basal ganglia were manually segmented, followed by three-dimensional texture analysis by using gray-level run-length matrix (GLRLM). One-way analysis of variance (ANOVA) was performed to compare differences between the two groups, and the bilateral Pearson linear correlation coefficient (r) was calculated to evaluate the correlation between texture parameters and UPDRS-III clinical scores.Results·The analysis of texture feature parameters showed that there were significant differences between the PD and HC groups in the gray matter nuclei. Among all the texture feature parameters of GLRLM, LngREnch showed significant differences between the PD group and the HC group in the five gray matter nuclei measured. The average magnetic susceptibility of gray matter nuclei and GLRLM texture parameters were sensitive in distinguishing PD from HC (AUC>0.5). The AUC values of RLNonUni, LngREnch, ShrtREmp, and Fraction were higher than that of the average magnetization susceptibiliyt. The correlation analysis showed that RLNonUni and GLevNonU in the caudate nucleus (CN), as well as GLevNonU in the red nucleus (RN), were significantly correlated with UPDRS-III scores, while no significant clinical correlations were found for the remaining parameters.Conclusion·Compared to the mean magnetic susceptibility values, GLRLM texture parameters provide better differentiation between the PD and HC groups. Multiparameter texture analysis offers a novel approach to QSM-based quantitative assessment of brain iron content, which can provide additional multidimensional quantitative information for the non-invasive diagnosis of PD.