| 林椿森,路伟钊,李文勤,李晶磊,闵刚,石丽婷.MRI纹理分析评价注意缺陷多动障碍[J].中国医学影像技术,2022,38(2):167~171 |
| MRI纹理分析评价注意缺陷多动障碍 |
| MRI texture analysis for observation of attention-deficit hyperactivity disorder |
| 投稿时间：2021-04-07 修订日期：2021-08-25 |
| DOI：10.13929/j.issn.1003-3289.2022.02.003 |
| 中文关键词: 注意缺陷障碍伴多动 纹理分析 磁共振成像 |
| 英文关键词:attention deficit disorder with hyperactivity texture analysis magnetic resonance imaging |
| 基金项目:山东省重点研发计划（2017GGX201010）。 |
|
| 摘要点击次数: 1812 |
| 全文下载次数: 554 |
| 中文摘要: |
| 目的 观察MRI纹理分析诊断注意缺陷多动障碍（ADHD）及分型的效果。方法 基于纽约大学医学中心公开MRI数据选取88例ADHD患者（ADHD组）及67名健康受试者（对照组），将ADHD组分为注意力缺陷为主型（ADHD-I）亚组（n=32）和混合型（ADHD-C）亚组（n=56），提取并比较受试者脑白质和脑灰质的纹理特征差异。比较ADHD组与对照组、ADHD-I亚组与ADHD-C亚组间纹理特征差异，以Spearman相关分析剔除相关性较高的冗余特征。基于具有显著差异的纹理特征构建支持向量机（SVM）模型，绘制受试者工作特征（ROC）曲线，计算曲线下面积（AUC），评价SVM模型诊断ADHD及分型的效能。结果 ADHD-I亚组、ADHD-C亚组及对照组间共12个脑灰质纹理特征及14个脑白质纹理特征差异具有统计学意义（P均<0.05）。基于24个脑灰质特征的SVM模型鉴别ADHD组与对照组的AUC为0.85，准确率为72.00%，敏感度为80.00%，特异度为60.00%；联合1个脑灰质和18个脑白质特征的SVM模型区分ADHD分型的AUC为0.81，准确率为84.00%，敏感度为93.33%，特异度为70.00%。结论 MRI纹理分析可用于诊断ADHD并分型。 |
| 英文摘要: |
| Objective To observe the value of MRI texture analysis in diagnosis and classification of attention deficit hyperactivity disorder (ADHD). Methods Based on the open MRI data of New York University Medical Center, 88 ADHD patients (ADHD group) and 67 healthy subjects (control group) were selected. Patients in ADHD group were divided into predominantly inattentive ADHD (ADHD-I) subgroup (n=32) and combined ADHD (ADHD-C) subgroup (n=56). The texture features of white matter and gray matter were extracted, and the statistical differences among groups/subgroups were calculated. The texture features were compared between ADHD group and control group, as well as ADHD-I subgroup and ADHD-C subgroup. Spearman correlation analysis was used to reduce redundant features with high correlations. The remained features being significantly different between groups/subgroups were used to establish support vector machine (SVM) classification models. The receiver operating characteristic (ROC) curve was drawn, and the area under the curve (AUC) was calculated to evaluate the performance of SVM models for diagnosis and classifying of ADHD. Results There were 12 gray matter features and 14 white matter features being significant different among ADHD-I subgroup, ADHD-C subgroup and control group (all P<0.05). The AUC of SVM model based on 24 gray matter features for differentiating ADHD patients healthy subjects was 0.85, with accuracy of 72.00%, sensitivity of 80.00% and specificity of 60.00%. SVM model combining 1 gray matter and 18 white matter features had an AUC of 0.81, accuracy of 84.00%, sensitivity of 93.33% and specificity of 70.00% for distinguishing different types ADHD. Conclusion MRI texture analysis could be used to diagnose and classify ADHD. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
