Lin Wei-Che, Chou Kun-Hsien, Lee Pei-Lin, Tsai Nai-Wen, Chen Hsiu-Ling, Hsu Ai-Ling, Chen Meng-Hsiang, Huang Yung-Cheng, Lin Ching-Po, Lu Cheng-Hsien
Department of Diagnostic Radiology, Kaohsiung Chang Gung Memorial Hospital and Chang Gung University College of Medicine, Kaohsiung, Taiwan.
Brain Research Center, National Yang-Ming University, Taipei, Taiwan.
Neuroradiology. 2017 Apr;59(4):367-377. doi: 10.1007/s00234-017-1808-0. Epub 2017 Mar 16.
This paper aims to examine the effectiveness of structural imaging as an aid in the diagnosis of Parkinson's disease (PD).
High-resolution T -weighted magnetic resonance imaging was performed in 72 patients with idiopathic PD (mean age, 61.08 years) and 73 healthy subjects (mean age, 58.96 years). The whole brain was parcellated into 95 regions of interest using composite anatomical atlases, and region volumes were calculated. Three diagnostic classifiers were constructed using binary multiple logistic regression modeling: the (i) basal ganglion prior classifier, (ii) data-driven classifier, and (iii) basal ganglion prior/data-driven hybrid classifier. Leave-one-out cross validation was used to unbiasedly evaluate the predictive accuracy of imaging features. Pearson's correlation analysis was further performed to correlate outcome measurement using the best PD classifier with disease severity.
Smaller volume in susceptible regions is diagnostic for Parkinson's disease. Compared with the other two classifiers, the basal ganglion prior/data-driven hybrid classifier had the highest diagnostic reliability with a sensitivity of 74%, specificity of 75%, and accuracy of 74%. Furthermore, outcome measurement using this classifier was associated with disease severity.
Brain structural volumetric analysis with multiple logistic regression modeling can be a complementary tool for diagnosing PD.
本文旨在研究结构成像在帕金森病（PD）诊断中的辅助作用。
对72例特发性帕金森病患者（平均年龄61.08岁）和73名健康受试者（平均年龄58.96岁）进行高分辨率T加权磁共振成像。使用复合解剖图谱将全脑划分为95个感兴趣区域，并计算区域体积。使用二元多重逻辑回归模型构建了三个诊断分类器：（i）基底神经节先验分类器，（ii）数据驱动分类器，以及（iii）基底神经节先验/数据驱动混合分类器。采用留一法交叉验证来无偏地评估成像特征的预测准确性。进一步进行Pearson相关性分析，以使用最佳帕金森病分类器将结果测量与疾病严重程度相关联。
易感区域体积较小对帕金森病具有诊断意义。与其他两个分类器相比，基底神经节先验/数据驱动混合分类器具有最高的诊断可靠性，灵敏度为74%，特异性为75%，准确率为74%。此外，使用该分类器的结果测量与疾病严重程度相关。
采用多重逻辑回归模型的脑结构体积分析可作为诊断帕金森病的辅助工具。
