Classification methods for HIV-1 medicated neuronal damage

HIV-1-associated dementia (HAD) is the most devastating disease happened in the central nervous system of AIDS patients. Neuronal damage, the early indicator of HAD, under different treatments can be applied to design and study specific therapies for the prevention or reversal of the neuronal death associated with HAD. A computer-based image program was used to quantitatively estimate the change of neurites, arbors, branch nodes, and cell bodies in cultured cortical neurons. Nine attributes (variables) and two classes G2 (non-treatment control group) and G4 (gp120-treatment group) were considered to describe the statuses of neuronal damage. Various classification methods have been carried out in our research group. In this paper, we focus on using logistic regression method for classification, and compare the resulting predictive accuracy with that of using two-class multiple criteria linear programming (MCLP) and neural networks (NN) models conducted earlier. The results show that logistic regression obtained the best classification accuracy. As a pilot study, it demonstrates the use and effectiveness of statistical method in the classification mining of neuronal damage associated with HAD.