A0725
Title: Classifying Alzheimers Disease patients and identifying related BOLD signals using penalized logistic regression
Authors: Hyeonjeong Lim - Chungnam National University (Korea, South) [presenting]
Eunjee Lee - Chungnam National University (Korea, South)
Jeong Yeon Park - Chungnam National University (Korea, South)
Abstract: Alzheimers Disease is one of the most prevalent types of dementia. As there is currently no complete cure for AD, it is crucial to detect and treat it early with proper care. Blood-Oxygen-Level Dependent (BOLD) signals can be used to identify abnormal patterns of brain activity in patients, which can facilitate early diagnosis of AD. Therefore, this study aims to explore the BOLD signals associated with the onset of AD and construct a model for classifying AD and NC patients based on these signals. We analyze the fMRI data provided by ADNI (Alzheimers Disease Neuroimaging Initiative). The study was conducted on 307 patients, each with 116 BOLD signals and corresponding demographic information. We extract the functional characteristics of the 116 BOLD signals by using functional principal component analysis (PCA) to calculate PC scores. We use the PC scores as explanatory variables in a logistic regression model. We consider LASSO, elastic-net, and SCAD penalties for variable selection. The prediction performance of the proposed method is compared with that of competing methods, including decision tree, random forest, and boosting models. We conduct a receiver operating characteristic (ROC) analysis to evaluate the model selection performance. As a result, we can identify BOLD signals related to Alzheimers Disease and proactively classify AD and NC by using the proposed model.