通過一個具有代表性的人群樣本,描述大腦動脈的復(fù)雜結(jié)構(gòu),對于診斷、分析和預(yù)測病理狀態(tài)具有重要意義。磁共振血管造影可顯示腦動脈血管。隨著自動追蹤和重建技術(shù)的出現(xiàn),神經(jīng)元三維重建數(shù)據(jù)的數(shù)量激增,神經(jīng)形態(tài)學(xué)研究也隨之興起。然而,缺乏機器驅(qū)動的注釋模式來自動檢測和預(yù)測受試者可能患腦血管疾病的障礙或風(fēng)險,使用神經(jīng)形態(tài)學(xué)測量值仍然是這門學(xué)科的一個障礙。隨機森林(RF)是一種常用的機器學(xué)習(xí)方法,在生物科學(xué)、經(jīng)濟學(xué)、化學(xué)工程、農(nóng)業(yè)科學(xué)、醫(yī)學(xué)研究等多個領(lǐng)域,在病理狀態(tài)的診斷和預(yù)測等關(guān)鍵應(yīng)用中都取得了競爭性的成功。它是一種由一個簡單的樹預(yù)測器組成的技術(shù),當(dāng)一組預(yù)測器值作為輸入時,每棵樹都會產(chǎn)生一個響應(yīng)。本文對機器學(xué)習(xí)領(lǐng)域進(jìn)行了全面的研究,采用基于決策樹的隨機算法randomforest對來自BraVa數(shù)據(jù)集的44名受試者的腦血管疾病的可能性進(jìn)行了預(yù)測。在腦血管疾病的分析和預(yù)測方面,本文利用SPSS統(tǒng)計工具,實現(xiàn)了被診斷為腦血管疾病的各指標(biāo)之間的獨立性和相關(guān)性。計算了描述整個血管結(jié)構(gòu)的各種定標(biāo)器參數(shù)在總體尺寸、分支特征、分叉角和對稱性方面的匯總統(tǒng)計。本研究所用數(shù)據(jù)的總體尺寸變異性與人體尺寸其他參數(shù)的報告值相似。...

【文章頁數(shù)】：79 頁

【學(xué)位級別】：碩士

【文章目錄】：
摘要
ABSTRACT
1 INTRODUCTION
    1.1 BACKGROUND OF STUDY
    1.2 SIGNIFICANCE OF RF FOR MORPHOMETRIC ANALYSIS AND PREDICTION
    1.3 PURPOSE OF RANDOM FOREST FOR AUTOMATIC DIAGNOSES
    1.4 OBJECTIVES OF THE RF ALGORITHM FOR NEUROMORPHOLOGICAL ANALYSIS
2 LITERATURE REVIEW OF RF IN NEUROSCIENCE
    2.1 NEUROSCIENCE
    2.2 NEURONS
        2.2.1 NEUROMORPHOLOGY
        2.2.2 NEURON SIZE AND SHAPE
        2.2.3 DEVELOPMENT IN CELL GROWTH
        2.2.4 MORPHOLOGICAL PLASTICITY
        2.2.5 EXTRINSIC VS. INTRINSIC INFLUENCES
    2.3 NEURON MINING PIPELINE
        2.3.1 NEURON DATA ACQUISITION
        2.3.2 NEURON FEATURE EXTRACTION
        2.3.3 PROCESSING DATA FROM NEUROMORPHOLOGICAL FEATURES
        2.3.4 NORMALIZATION
        2.3.5 MISSING VALUE TREATMENT
        2.3.6 DATA UNIFICATION AND CONSOLIDATION
        2.3.7 ADDRESS IMBALANCE DATASET
        2.3.8 EXCLUSION OF CONFOUNDING VARIABLES
        2.3.9 DIMENSIONALITY REDUCTION
        2.3.10 FEATURE SELECTION
        2.3.11 UNSUPERVISED LEARNING
        2.3.12 SUPERVISED LEARNING
        2.3.13 MULTILABEL AND MULTICLASS CLASSIFICATION
    2.4 RANDOM FOREST AS A TOOL FOR AUTOMATIC PREDICTION
        2.4.1 TYPES OF DECISION TREES
        2.4.2 RANDOM FOREST AS METRIC FOR PREDICTION
        2.4.3 RANDOM FOREST ANALYSIS ON DIABETES COMPLICATION DATA
        2.4.4 RF ENSEMBLES FOR DETECTION AND PREDICTION OF ALZHEIMER'S DISEASE WITH A GOOD BETWEEN-COHORT ROBUSTNESS
        2.4.5 RANDOM FOREST ATTRIBUTION SELECTION MEASURES
3 RANDOM FOREST AND STATISTICAL METHODOLOGY
    3.1 SUBJECT AND BRAVA DATA ACQUISITION
        3.1.1 DIGITAL RECONSTRUCTION
        3.1.2 MORPHOLOGICAL ANALYSIS
    3.2 EXPERIMENTAL SETTINGS AND RESULTS FROM THE BRAVA DATASET AND DIABETES DATASET
        3.2.1 THE DATASETS EMPLOYED
        3.2.2 EXPERIMENT 3A
        3.2.3 EXPERIMENT 3B AND 3C
4 RANDOM FOREST AND SPSS RESULTS AND ANALYSIS
    4.1 QUANTITATIVE ANATOMY OF CEREBRAL ARTERIES
    4.2 RANDOM FOREST FOR THE BRAVA DATASET
        4.2.1 TESTING RANDOM FOREST WITH BRAVA (AGE/CONTRACTION)
        4.2.2 TESTING RANDOM FOREST ALGORITHM WITH THE DIABETES DATASET
5 CONCLUSIONS
    5.1 RF ALGORITHM CONCLUSION
    5.2 RECOMMENDATION
ACKNOWLEDGEMENT
REFERENCES
APPENDIX OF CORRELATION BETWEEN THE VARIOUS ARTERIAL METRICS



本文編號：4025544