Diagnostic Value of Texture Analysis Based on Quantitative Susceptibility Mapping in Parkinson's Disease / 中国医学影像学杂志

ABSTRACT

Purpose To explore the value of texture analysis in the diagnosis and course evaluation of Parkinson's disease(PD)by analyzing the texture features of gray matter nuclei and white matter on quantitative susceptibility mapping(QSM)sequences.Materials and Methods A total of 30 PD patients and 22 normal controls from July 2019 to November 2020 in Jiangyin People's Hospital were prospectively enrolled to perform enhanced gradient echo T2* weighted angiography(ESWAN)sequence scanning.All QSM images were obtained through post-processing.Region of interest was manually obtained,including bilateral caudate heads,globus pallidus,putamen,substantia nigra,red nucleus,cerebellar dentate nucleus and white matter at the center of the semicircle.The texture features of the region of interest were extracted.After dimension reduction and screening,a set of optimal texture parameters were obtained,and a random forest prediction model was constructed.The diagnostic efficiency of the model was analyzed and evaluated and the reliability of the model was evaluated.The correlation between the selected texture features and the clinical scale of PD patients was analyzed.Results A group(n=5)of the best texture feature parameters were obtained from QSM map.The area under curve range of independent prediction PD was 0.697-0.823,the area under curve of random forest model was 0.910,and the accuracy of cross validation was 0.888.Texture feature(r4_wavelet_LLL_firstorder_Energy)of PD group was negatively correlated with the scores of the mini mental state examination(r=-0.470,P=0.011).Conclusion The texture analysis based on QSM has a high diagnostic value for PD,and the texture features of the left putamen have a certain correlation with the cognitive function of PD patients.