Nonlinear feature transformation and deep fusion for Alzheimer's Disease staging analysis
作者：
Highlights：
• A novel nonlinear metric learning method is proposed to improve patient classifications of AD/MCI vs. normal controls.
• Thin-plate splines are integrated with SVM classifiers to make data points more (linearly) separable.
• Cross-sectional and longitudinal neuroimaging features estimated from MR brain images are fused through stacked denoising sparse auto-encoder.
• The effectiveness of the proposed feature transformation and fusion strategies is evaluated with ADNI dataset.
摘要
Highlights•A novel nonlinear metric learning method is proposed to improve patient classifications of AD/MCI vs. normal controls.•Thin-plate splines are integrated with SVM classifiers to make data points more (linearly) separable.•Cross-sectional and longitudinal neuroimaging features estimated from MR brain images are fused through stacked denoising sparse auto-encoder.•The effectiveness of the proposed feature transformation and fusion strategies is evaluated with ADNI dataset.
论文关键词：Metric learning,Alzheimer's Disease (AD),Mild Cognitive Impairment (MCI),SVM classifier,Feature fusion,Deep neural networks
论文评审过程：Received 1 February 2016, Revised 17 September 2016, Accepted 21 September 2016, Available online 30 September 2016, Version of Record 27 November 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.09.032
