Histogram Analysis of Quantitative Susceptibility Mapping for the Diagnosis of Parkinson's Disease

Abstract

Rationale and objectives: To investigate the diagnostic performance of histogram analysis combined with quantitative susceptibility mapping (QSM) for differentiating Parkinson's disease (PD) patients from healthy controls.

Methods: We included 35 patients with PD diagnosed by two neurologists from August 2019 to January 2020 in our hospital in this prospective study. The clinical diagnosis was based on the Movement Disorder Society Clinical Diagnostic Criteria for PD. At the same time, 23 healthy volunteers matched for age and sex were recruited as controls. The Mini Mental State Examination, the third part of the Parkinson's Disease Rating Scale, the Hoehn & Yahr stages, and disease duration (year) were used to assess the PD patients. QSM was performed using a 3T MR scanner. The regions of interest were depicted according to the head of the caudate nucleus(CN), globus pallidus(GP), putamina (PUT), thalmus(TH), substantia nigra (SN), red nucleus(RN), and dentate nucleus. Then the corresponding histogram features were extracted. The Mann-Whitney U test was used to identify significant histogram features for differentiating PD patients from healthy controls. Area under the receiver operating characteristics curve (AUC) analysis was conducted to evaluate the diagnostic performance of all significant histogram features. Multivariate logistic regression analysis was performed to identify the best combined model for all seven nuclei. Differences among the AUCs were compared pairwise.

Results: Histogram features in all nuclei except TH showed significant differences between the groups. Among the single features, the 10th percentile of SN (SN_P10) yielded the highest AUC of 0.894, with the highest specificity of 86.86% for differentiating PD patients from healthy controls. The 75th percentile of PUT (PUT_P75) yielded the highest sensitivity of 97.14%. In the multivariate logistic regression analysis, SN_P10 combined with PUT_P75 yielded the highest diagnostic performance with the highest AUC of 0.911, the highest specificity of 91.30% and an excellent sensitivity of 92.40%.

Conclusion: QSM combined with histogram analysis successfully distinguished PD patients from healthy controls, and the result was notably superior to the mean value.

Keywords: Histogram analysis; Iron deposition; Parkinson's disease; Quantitative susceptibility mapping.