【摘要】：大腦是人身體中最重要、最復(fù)雜的器官,相當(dāng)于計(jì)算機(jī)的CPU,支配著人們的思維和各種行為,研究大腦的組織結(jié)構(gòu)和工作機(jī)理對(duì)于人類身體健康具有重要意義。阿爾茲海默癥(AD),是神經(jīng)系統(tǒng)退行性的一種疾病,而且該疾病的起病和發(fā)展不容易被人們所發(fā)現(xiàn),臨床上記憶力、溝通和語言能力、行為能力、認(rèn)知功能等持續(xù)下降甚至全面喪失,導(dǎo)致全面性癡呆,民間稱老年癡呆癥。輕度認(rèn)知障礙(MCI)是介于正常衰老和癡呆之間的一種中間狀態(tài),與正常(NC)老人相比,患者有一定的輕度認(rèn)知功能減退,但是還不足以表現(xiàn)為癡呆,是老年癡呆的高發(fā)人群。由于現(xiàn)今還無法治愈阿爾茲海默癥,其病情具有不可逆性,在還沒有轉(zhuǎn)化為阿爾茲海默癥的輕度認(rèn)知障礙這一階段進(jìn)行預(yù)防和干預(yù),對(duì)于控制患者的病情和延緩發(fā)病癡呆有明顯效果。在越來越多的相關(guān)研究中表明,AD患者與正常老年人、MCI患者與正常老年人之間腦部組織結(jié)構(gòu)發(fā)生了變化,腦皮層存在差異,借助核磁共振成像(MRI)技術(shù),對(duì)采集到的大腦圖像數(shù)據(jù)進(jìn)行處理分析,研究AD和MCI腦結(jié)構(gòu)組織的變化,對(duì)降低AD的發(fā)病和死亡率有重要的意義,為實(shí)現(xiàn)AD的早期預(yù)測(cè)和自動(dòng)診斷提供理論依據(jù)。本文的主要貢獻(xiàn)如下:1.對(duì)AD、MCI、NC這些被試人群的MRI腦影像數(shù)據(jù),以基于曲面形態(tài)學(xué)的分析方法,計(jì)算出被試人群腦形態(tài)學(xué)的三種指標(biāo):皮層厚度、灰質(zhì)體積、皮層復(fù)雜度。對(duì)三種被試人群,基于每種指標(biāo)數(shù)據(jù),利用統(tǒng)計(jì)分析的方法,比較每?jī)煞N人群的大腦結(jié)構(gòu)組織差異,證明每種指標(biāo)下AD和NC及MCI和NC兩種模型腦組織存在顯著差異,為多任務(wù)學(xué)習(xí)選取特征提供依據(jù),同時(shí)找出有哪些腦區(qū)存在異常。2.對(duì)AD和NC,MCI和NC兩種模型分別進(jìn)行分類實(shí)驗(yàn)。在該領(lǐng)域嘗試運(yùn)用多任務(wù)學(xué)習(xí)(MTL)方法來選擇特征,對(duì)腦皮層形態(tài)學(xué)的三種指標(biāo):腦皮層厚度、灰質(zhì)體積、皮層復(fù)雜度看成三個(gè)任務(wù),選擇特征時(shí)任務(wù)之間存在一定的聯(lián)系。同時(shí)與F-score、mRMR兩種特征選擇方法進(jìn)行比較。在特征選擇的基礎(chǔ)上,選擇一個(gè)合適的分類器更為重要,本文嘗試采用極限學(xué)習(xí)機(jī)(ELM)的分類方法,同時(shí)結(jié)合SVM-Linear、SVM-RBF兩種分類器,探索目前MRI中常用的特征選擇方法與分類器在AD和NC以及MCI和NC分類上的最優(yōu)組合。
[Abstract]:The brain is the most important and complex organ in the human body, which is equivalent to the computer CPU, which dominates people's thinking and various behaviors. It is of great significance to study the structure and working mechanism of the brain for human health. Alzheimer's disease (AD),) is a degenerative disease of the nervous system, and the onset and development of the disease are not easily discovered. Cognitive function and other continuous decline or even total loss, leading to comprehensive dementia, known as Alzheimer's disease. Mild cognitive impairment (MCI) is a kind of intermediate state between normal aging and dementia. Compared with normal (NC) patients, the patients have mild cognitive impairment, but not enough to show dementia, which is a high risk group of senile dementia. Because there is no cure for Alzheimer's disease, and its condition is irreversible, prevention and intervention are carried out at a stage that has not yet transformed into mild cognitive impairment of Alzheimer's disease. To control the patient's condition and delay the onset of dementia has obvious effect. More and more related studies have shown that the brain tissue structure has changed between AD patients and normal elderly people. There are differences in cerebral cortex between AD patients and normal elderly patients. Magnetic resonance imaging (MRI) technique is used. Processing and analyzing the collected brain image data and studying the changes of brain structure in AD and MCI are of great significance to reduce the morbidity and mortality of AD and provide theoretical basis for early prediction and automatic diagnosis of AD. The main contributions of this paper are as follows: 1. In this paper, the MRI brain image data of the subjects were analyzed based on curved surface morphology, and three indexes of brain morphology were calculated: thickness of cortex, volume of gray matter and complexity of cortex. Based on the data of each index, the differences of brain structure in each group were compared by statistical analysis. It was proved that there were significant differences between AD and NC and MCI and NC in each index. To provide the basis for selecting the characteristics of multitask learning, and to find out which brain regions have abnormal. 2. 2. Classification experiments were carried out on AD and NCU MCI and NC models respectively. In this field, the multi-task learning (MTL) method is used to select the features. The cortical thickness, gray matter volume, cortical complexity are regarded as three tasks, and there is a certain relationship between the tasks when selecting the features. At the same time, it is compared with two feature selection methods of F-scorex mRMR. On the basis of feature selection, it is more important to select a suitable classifier. In this paper, we try to use the classification method of extreme learning machine (ELM) and combine SVM-Linear-SVM-RBF classifier. This paper explores the optimal combination of feature selection methods and classifiers in AD and NC and MCI and NC classification in MRI.
【學(xué)位授予單位】：山東師范大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：R749.16;TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 姚旭;王曉丹;張玉璽;權(quán)文;;特征選擇方法綜述[J];控制與決策;2012年02期

2 丁世飛;齊丙娟;譚紅艷;;支持向量機(jī)理論與算法研究綜述[J];電子科技大學(xué)學(xué)報(bào);2011年01期

相關(guān)博士學(xué)位論文 前2條

1 姚志軍;輕度認(rèn)知障礙和阿爾茲海默病腦形態(tài)異常的磁共振影像研究[D];蘭州大學(xué);2011年

2 林磊;基于MRI的大腦皮層形態(tài)學(xué)研究[D];浙江大學(xué);2010年

相關(guān)碩士學(xué)位論文 前2條

1 梁傳江;基于曲面形態(tài)學(xué)的輕度認(rèn)知障礙腦結(jié)構(gòu)特性研究[D];蘭州大學(xué);2014年

2 李楊;基于多任務(wù)學(xué)習(xí)的人臉圖像識(shí)別[D];上海大學(xué);2014年

，

本文編號(hào)：2163374