• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Annual International Conference of the IEEE Engineering in Medicine and Biology Society >Multimodal MRI classification in vascular mild cognitive impairment
【24h】

Multimodal MRI classification in vascular mild cognitive impairment

机译：血管性轻度认知障碍的多模式MRI分类

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Vascular mild cognitive impairment (VMCI) is a disorder in which multimodal MRI can add significant value by combining diffusion tensor imaging (DTI) with brain morphometry. In this study we implemented and compared machine learning techniques for multimodal classification between 58 VMCI patients and 29 healthy subjects as well as for discrimination (within the VMCI group) between patients with different cognitive performances. For each subject, a cortical feature vector was constructed based on cortical parcellation and cortical and subcortical volumetric segmentation and a DTI feature vector was formed by combining descriptive statistical metrics related to the distribution of DTI invariants within white matter. We employed both a sequential minimal optimization and a functional tree classifier, using feature selection and 10-fold cross-validation, and compared their performances in monomodal and multimodal classification for both classification problems (healthy subjects vs VMCI and prediction of cognitive performance). While monomodal classification resulted in satisfactory performance in most cases, turning from monomodal to multimodal classification resulted in an improvement of the performance in the discrimination between VMCI patients with low cognitive performance and healthy subjects by up to 10% in sensitivity (leaving specificity unchanged). We therefore are able to confirm the usefulness of machine learning techniques in discriminating diseased states based on neuroimaging data.
机译：血管性轻度认知障碍（VMCI）是一种疾病，其中多模式MRI可通过将扩散张量成像（DTI）与脑形态计量学相结合来增加显着价值。在这项研究中，我们实施并比较了机器学习技术，用于58名VMCI患者和29名健康受试者之间的多模式分类，以及针对具有不同认知表现的患者（在VMCI组内）进行的区分。对于每个受试者，基于皮质细胞分裂以及皮质和皮质下体积分割构建皮质特征向量，并通过组合与白质内DTI不变量分布相关的描述性统计指标来形成DTI特征向量。我们使用了顺序最小优化和功能树分类器，使用了特征选择和10倍交叉验证，并针对两种分类问题（健康受试者vs VMCI和认知表现的预测）比较了它们在单峰和多峰分类中的表现。尽管单峰分类在大多数情况下均能获得令人满意的表现，但从单峰分类转变为多峰分类可将识别能力低的VMCI患者与健康受试者之间的辨别性能提高多达10％（敏感性保持不变）。因此，我们能够确认机器学习技术在基于神经影像数据区分疾病状态中的有用性。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号