School of Mechanical, Electrical & Information Engineering, Shandong University, Weihai, Shandong Province, 264209, China.
CAS Key Laboratory of Molecular Imaging, Institute of Automation, Beijing, 100190, China.
Eur Radiol. 2017 Oct;27(10):4153-4162. doi: 10.1007/s00330-017-4820-1. Epub 2017 Apr 10.
To identify the white matter (WM) impairments of the antiretroviral therapy (ART)-naïve HIV patients by conducting a multivariate pattern analysis (MVPA) of Diffusion Tensor Imaging (DTI) data METHODS: We enrolled 33 ART-naïve HIV patients and 32 Normal controls in the current study. Firstly, the DTI metrics in whole brain WM tracts were extracted for each subject and feed into the Least Absolute Shrinkage and Selection Operators procedure (LASSO)-Logistic regression model to identify the impaired WM tracts. Then, Support Vector Machines (SVM) model was constructed based on the DTI metrics in the impaired WM tracts to make HIV-control group classification. Pearson correlations between the WM impairments and HIV clinical statics were also investigated.
Extensive HIV-related impairments were observed in the WM tracts associated with motor function, the corpus callosum (CC) and the frontal WM. With leave-one-out cross validation, accuracy of 83.08% (P=0.002) and the area under the Receiver Operating Characteristic curve of 0.9110 were obtained in the SVM classification model. The impairments of the CC were significantly correlated with the HIV clinic statics.
The MVPA was sensitive to detect the HIV-related WM changes. Our findings indicated that the MVPA had considerable potential in exploring the HIV-related WM impairments.
• WM impairments along motor pathway were detected among the ART-naïve HIV patients • Prominent HIV-related WM impairments were observed in CC and frontal WM • The impairments of CC were significantly related to the HIV clinic statics • The CC might be susceptible to immune dysfunction and HIV replication • Multivariate pattern analysis had potential for studying the HIV-related white matter impairments.
通过对弥散张量成像（DTI）数据进行多元模式分析（MVPA），来识别未接受抗逆转录病毒治疗（ART）的 HIV 患者的白质（WM）损伤。
本研究纳入了 33 名未接受 ART 的 HIV 患者和 32 名正常对照。首先，对每位受试者的全脑 WM 束中的 DTI 指标进行提取，并将其输入到最小绝对值收缩和选择算子程序（LASSO）-逻辑回归模型中，以识别受损的 WM 束。然后，基于受损 WM 束中的 DTI 指标构建支持向量机（SVM）模型，以实现 HIV-对照组分类。还研究了 WM 损伤与 HIV 临床统计学之间的 Pearson 相关性。
在与运动功能相关的 WM 束、胼胝体（CC）和额 WM 中观察到广泛的与 HIV 相关的损伤。通过留一法交叉验证，SVM 分类模型的准确率为 83.08%（P=0.002），受试者工作特征曲线下面积为 0.9110。CC 的损伤与 HIV 临床统计学显著相关。
MVPA 能够敏感地检测到与 HIV 相关的 WM 变化。我们的研究结果表明，MVPA 在探索与 HIV 相关的 WM 损伤方面具有相当大的潜力。
· 未接受 ART 的 HIV 患者中检测到沿运动通路的 WM 损伤。
· 在 CC 和额 WM 中观察到明显的与 HIV 相关的 WM 损伤。
· CC 的损伤与 HIV 临床统计学显著相关。
· CC 可能容易受到免疫功能障碍和 HIV 复制的影响。
· 多元模式分析具有研究与 HIV 相关的白质损伤的潜力。
