汉字书写对区分遗忘型轻度认知障碍的可行性研究

发布时间：2018-03-15 03:38

本文选题：汉字书写量表　切入点：遗忘型轻度认知障碍　出处：《杭州电子科技大学》2016年硕士论文　论文类型：学位论文

【摘要】：阿尔茨海默病(Alzheimer's disease,AD)是最主要的老年期痴呆类型,一旦发生便不可逆转。AD慢性损害大脑高级皮层功能,包括记忆力、视空间能力、认知能力、情绪情感和语言能力。如何通过客观有效的方法来识别早期AD患者是医学界面临的难题之一,神经心理学检查是目前临床上常用的评估方法。遗忘型轻度认知障碍(amnestic Mild Cognitive Impairment,a-MCI)作为AD的亚临床(前期)状态,近年来受到极大关注。因此,我们将研究重点转移到a-MCI之上,而对于a-MCI的筛选和检测手段,经过临床上长时间的研究证实,无论是在效率还是敏感性和特异性方面,神经心理学量表都占有一席之地,对识别a-MCI患者及早期AD患者具有极其重要的意义。目前用于AD诊断的汉语神经心理学量表十分缺乏,我们常用的量表都是从国外引进,例如,简明心理状况量表(mini-mental state examination,MMSE)和阿尔茨海默病评估量表(Alzheimer's Disease Assessment Scale,ADAS),其中涉及语言功能评定的条目,只适用于拼音语系患者,尤其是说英语国家的被试使用,而对说汉语写汉字的中国人来说不太适合。因此,研究专门的汉语书写量表,用于说汉语的AD患者的诊断是非常有价值的。本文利用目前北京大学第一医院失语检查的书写部分作为汉字书写量表(Chinese Character Writing Scale,CCWS),使用深度学习算法对其进行建模分类,并将结果与MMSE量表和ADAS量表结果作对比,研究该汉字书写量表对区分a-MCI的可行性。针对这样的需求,本实验先利用t检验对各组间各量表总分进行检验,初步验证在总分上是否可将该量表用于区分a-MCI。实验得到CCWS量表总分在三组之间均有显著性差异。MMSE量表总分仅在a-MCI组与AD组之间有显著性差异。ADAS量表总分三组之间差异均极其显著。实验结果提示,就总分而言,CCWS可用于区分a-MCI。之后再利用深度学习算法基于CCWS各分测验得分对NC组、a-MCI组以及AD组患者进行分类识别,并将结果与ADAS量表和MMSE量表结果相比较。所有实验均在以下两个组之间(NC组和a-MCI组、a-MCI组和AD组)展开。经过实验研究,与MMSE量表和ADAS量表分类精度相比,CCWS量表区分NC组和a-MCI组的分类准确率高于MMSE量表,与ADAS量表较为接近,且其最敏感的分测验是听写-偏旁、听写-短句、看图写字、写病情而其他分测验相对保存。实验表明使用CCWS量表各项分测验得分训练得到的分类器可以从正常老年人中筛选出a-MCI患者,临床上可以通过对这些患者进一步的检查,以期达到AD早期诊断的目的。因此,可在临床上使用CCWS量表作为a-MCI筛查和检测的辅助工具,而实验中得到的最敏感的测验子项目可以为下一步创制出一套重点评估AD患者书写能力的量表提供理论依据。
[Abstract]:Alzheimer's disease (AD) is the most common type of Alzheimer's disease. Once it occurs, AD is irreversible and chronic damage to the superior cortical function of the brain, including memory, visual and spatial abilities, and cognitive ability. Emotional emotion and language ability. How to identify early AD patients through objective and effective methods is one of the problems facing the medical profession. Neuropsychological examination is a commonly used clinical evaluation method at present. Amnestic Mild Cognitive Impairmenta-MCI, as the subclinical (preclinical) state of AD, has received great attention in recent years. For the screening and detection of a-MCI, it has been proved by clinical research for a long time that the neuropsychological scale has a place in the aspects of efficiency, sensitivity and specificity. It is of great significance to identify a-MCI patients and early AD patients. At present, the Chinese neuropsychological scale used for AD diagnosis is very scarce. Our commonly used scales are imported from abroad, for example, The mini-mental state examination scale (MMSE) and the Alzheimer's disease assessment scale (Alzheimer's disease), which involve language function assessment items, are only applicable to patients with phonetic families, especially in English-speaking countries. However, it is not suitable for Chinese who speak Chinese and write Chinese characters. Therefore, a special Chinese writing scale is studied. The diagnosis of AD patients who speak Chinese is very valuable. This paper uses the writing part of aphasia examination in the first Hospital of Peking University as the Chinese Character Writing scale CCWSUS, and uses the depth learning algorithm to model and classify it. Compared with the results of MMSE and ADAS, the feasibility of the Chinese character writing scale to distinguish a-MCI was studied. In order to meet this demand, the total scores of each group were tested by t test. The experiment shows that the total score of CCWS scale is significantly different among the three groups. The total score of MMSE scale is only significant difference between the a-MCI group and AD group. The total score of ADAS scale is between the three groups. The differences are extremely significant. The experimental results suggest that, As far as the total score is concerned, CCWS can be used to distinguish a-MCIs. Then, based on the scores of each subtest of CCWS, the patients in NC group and AD group are classified and identified by using depth learning algorithm. The results were compared with those of ADAS and MMSE. All the experiments were carried out between NC group and a-MCI group, a-MCI group and AD group. Compared with the classification accuracy of MMSE and ADAS, the accuracy of classification of NC and a-MCI was higher than that of MMSE, and the most sensitive subtest was dictation, dictation and short sentences. The experiment shows that the classifier trained by each subtest of the CCWS scale can screen a-MCI patients from the normal elderly, and can be further examined clinically by these patients. In order to achieve the purpose of early diagnosis of AD, CCWS scale can be used as a supplementary tool for screening and detection of a-MCI in clinic. The most sensitive test subitems obtained in the experiment can provide theoretical basis for the next step to develop a set of scales that focus on the evaluation of AD patients' writing ability.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R749.1

【相似文献】