HIV-associated Neuropathic Pain Classification of MRI Brain Images

Open Access
Wang, Dongzhe
Graduate Program:
Electrical Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
April 02, 2014
Committee Members:
  • David Jonathan Miller, Thesis Advisor
  • HIV
  • neuropathic pain
  • MRI
  • medical imaging
  • support vector machines
  • feature extraction
  • recursive feature elimination
HIV-associated sensory neuropathy influences over 50% of HIV patients. The clinical expression of HIV neuropathy is dramatically variable. Although many HIV patients report few symptoms, approximately half report distal neuropathic pain (DNP). To better understand how the central nervous system is associated with HIV DNP, in this thesis, an analysis of HIV-infected participants’ brain structural magnetic resonance imaging (MRI) volumes was performed. Using multivariable regression analysis (involving demographic and clinical variables), the relationship between HIV DNP and the MRI results was investigated. Our study concluded that worse severity of DNP symptoms was correlated with smaller cerebral cortical gray matter [1]. According to this conclusion, we performed a statistical classification analysis on the presence of DNP symptoms in the structural MRI images. We generated three relevant feature extraction schemes, leading to three separate experiments. These three experiments will be helpful and informative for our study on clinical HIV DNP diagnosis. The novelty in this work relative to existing HIV DNP studies is the optimization of DNP classification performance based on the MRI data sets, using low dimensional features and computationally efficient models.