Department of Computer Science, University of Ioannina, Greece.
Artif Intell Med. 2011 Sep;53(1):35-45. doi: 10.1016/j.artmed.2011.05.005. Epub 2011 Jun 23.
The aim of this work is to provide a supervised method to assist the diagnosis and monitor the progression of the Alzheimer's disease (AD) using information which can be extracted from a functional magnetic resonance imaging (fMRI) experiment.
The proposed method consists of five stages: (a) preprocessing of fMRI data, (b) modeling of the fMRI voxel time series using a generalized linear model, (c) feature extraction from the fMRI experiment, (d) feature selection, and (e) classification using the random forests algorithm. In the last stage we employ features that were extracted from the fMRI and other features such as demographics, behavioral and volumetric measures. The aim of the classification is twofold: first to diagnose AD and second to classify AD as very mild and mild.
The method is evaluated using data from 41 subjects. The stage of AD is established using the Washington University Alzheimer's Disease Research Center recruitment and assessment procedures. The method classifies a patient as healthy or demented with 84% sensitivity and 92.3% specificity, and the stages of AD with 81% and 87% accuracy for the three class and the four class problem, respectively.
The method is advantageous since it is fully automated and for the first time the diagnosis and staging of the disease are addressed using fMRI.
本研究旨在提供一种基于监督学习的方法，通过对功能磁共振成像（fMRI）实验中提取的信息，辅助阿尔茨海默病（AD）的诊断和病情进展监测。
该方法共分为五个阶段：（a）fMRI 数据预处理，（b）使用广义线性模型对 fMRI 体素时间序列进行建模，（c）从 fMRI 实验中提取特征，（d）特征选择，（e）使用随机森林算法进行分类。在最后一个阶段，我们同时使用了从 fMRI 中提取的特征以及其他特征，如人口统计学、行为和容积测量。分类的目的有两个：一是诊断 AD，二是将 AD 分为非常轻度和轻度。
该方法使用了 41 名受试者的数据进行评估。AD 分期采用华盛顿大学阿尔茨海默病研究中心的招募和评估程序确定。该方法对健康和痴呆患者的分类具有 84%的敏感性和 92.3%的特异性，对 AD 的三个分期和四个分期的分类准确率分别为 81%和 87%。
该方法具有优势，因为它是全自动的，并且首次使用 fMRI 进行疾病的诊断和分期。
