基于NM-MRI的影像组学在帕金森病诊断中的价值

Value of imaging omics based on NM-MRI in the diagnosis of Parkinson’s disease

目的 应用神经黑色素敏感磁共振成像(NM-MRI)影像组学纹理特征探究其诊断帕金森病(PD)的价值。方法 收集牡丹江医学院附属红旗医院2020年1月至2022年2月PD患者及健康对照者各60例。以中脑黑质致密部(SNc)为感兴趣区域(ROI)进行勾画提取纹理特征参数,使用费希尔参数法(Fisher)、最小分类误差与最小平均相关系数法(POE+ACC)、相关信息测度法(MI)对取得的影像组学纹理特征参数降维处理,导入Ma Zda的B11处理器进行分析。比较PD组、健康对照组(HC组)两者影像组学纹理特征参数差异,绘制受试者工作特征曲线(ROC)。结果 三种降维方法中Fisher提取的最佳参数与NDA相结合的错判率最低(5.83%),PD与HC组SNc影像组学纹理特征参数Teta4差异无统计学意义(P> 0.05),其余均具有统计学意义(P <0.05),其中S(0,1)Dif Varnc曲线下面积(AUC)最大(0.889)。结论 NM-MRI影像组学纹理分析能够为PD诊断提供诊断依据,在区分PD与正常对照组中具有较好的参考价值。

Objective To explore the value of neuromelanin-sensitive magnetic resonance imaging(NM-MRI) in the diagnosis of Parkinson’s disease(PD) by using its imaging omics texture features. Methods A total of 60 PD patients and 60 healthy control patients admitted to Hongqi Hospital Affiliated to Mudanjiang Medical University from January 2020 to February 2022 were selected. The midbrain substantia nigra compacta(SNc) was drawn as a region of interest(ROI) to extract texture feature parameters. The Fisher parameter method, the minimum classification error and the minimum average correlation coefficient method(POE+ACC), and the mutual information measure(MI) were adopted to reduce the dimension of the texture feature parameters obtained from imaging omics, which were imported into the B11 processor of Ma Zda for analysis. The differences of the imaging omics texture feature parameters between the PD group and the healthy controlgroup(HC group) were compared, and the receiver operating characteristic curve(ROC) was drawn. Results Among the three dimensionality reduction methods, the best parameter extracted by Fisher in combination with NDA had the lowest error rate(5.83%). There was no significant difference in Teta4, the SNc imaging omics texture feature parameter, between the PD group and the HC group(P > 0.05), while there were significant differences in other parameters between the two groups(P < 0.05), among which the difference of the area under the curve(AUC) of S(0,1)Dif Varnc between the two groups was the most significant(0.889). Conclusion NM-MRI imaging omics texture analysis can provide diagnostic basis for PD diagnosis, and has good reference value in distinguishing PD from normal control group.