轻度认知功能障碍(MCI)是正常老化向痴呆转变的过渡阶段,目前被认为是老年痴呆症(AD)的一种先期征兆,其相关研究对于AD的早期诊断与干预具有重要意义。MCI的诊断一般通过认知和记忆的测查进行,各项指标均为正常或MCI状态时可直接确诊,如果不一致则需医生依据经验进一步判断。本研究从已确诊的被试中训练得出支持向量机分类模型,然后对需要医生诊断的被试做预测,实验表明,以医生的诊断为准,预测的符合率最高可达85.7%,有助于MCI的计算机辅助诊断。
Mild Cognitive Impairment (MCI), the transitional stage from normal aging to the Alzheimer Disease (AD), is now regarded as the early stage of AD. and the research on MCI is significant for the early diagnosis and therapy of AD. Generally cognition and memory function examination are performed in diagnosis of MCI. It is easy to diagnose subjects as normal or MCI when all test indexes are identical, but the final diagnosis needs to be made by doctors according to their experiences if there is any difference among indexes. In this paper, a classifier was trained based on support vector machine (SVM) using the data of the subjects with confirmed diagnosis, and then to predict the state of those undiagnosed. The experiment showed the highest prediction accuracy achieved 87.6% according to doctors' diagnoses, and the method can be used in computer-aided diagnosis of MCI.