Spatial Large-Margin Angle-Based Classifier for Multi-Category Neuroimaging Data

• Activity Number: 420
• Type: Topic Contributed
• Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Leo Yu-Feng Liu* and Yufeng Liu and Hongtu Zhu
• Companies: The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
• Keywords: ADMM ; Alzheimer's Disease ; Angle-Based Classifier ; Fused Lasso ; Large Margin Classifier ; Neuroimaging classification

Abstract

With the development of imaging techniques, scientists are interested in identifying imaging biomarkers that are related to different subtypes or transitional stages of various neuropsychiatric and neurodegenerative diseases. In this paper, we propose a novel Fused Lasso (Tibshirani et al., 2005) penalized Multi-category Angle-based Classifier (FLMAC) (Zhang and Liu, 2014) for the identification of such imaging biomarkers. The proposed FLMAC not only utilizes the spatial structure of imaging data, but also handles both binary and multi-category classification problems. Moreover, FLMAC is designed to effectively deal with the high dimensionality of most image data. We introduce an efficient algorithm based on an Alternative Direction Method of Multipliers (ADMM) algorithm to solve the large scale optimization problem for FLMAC. Both our simulation and real data experiments demonstrate the usefulness of FLMAC.