【作者】 高敏；

【导师】 刘军；

【作者基本信息】 中南大学 ， 临床医学（专业学位）， 2025， 博士

【摘要】 第一部分:基于多参数MRI的脑肿瘤类淋巴功能评估及其组织病理学相关性研究背景与目的:脑类淋巴系统通过脑脊液-间质液交换参与神经代谢废物清除,但其在脑肿瘤中的机制未明。本研究应用基于扩散张量成像(Diffusion Tensor Imaging,DTI)的脑室周围血管间隙指数(Along the Perivascular Space,ALPS)和多参数MRI(Multiparametric Magnetic Resonance Imaging,MTP),探讨类淋巴功能与水通道蛋白4(Aquaporin-4,AQP4)表达、瘤周水肿及微环境的关系。方法:纳入84例脑肿瘤患者及59例健康对照,行DTI及MTP扫描,测量ALPS指数、脑脊液、肿瘤及瘤周水肿(Peritumoral Brain Edema,PTBE)体积,并评估组织AQP4表达。结果:肿瘤组ALPS指数显著低于对照组(2.315 vs.2.879,p=0.001),脑脊液体积明显增加(p=0.001);肿瘤体积及脑脊液体积与ALPS指数负相关(r=-0.715,-0.747);PTBE体积与AQP4表达强正相关(r=0.989)。肿瘤区质子密度与ALPS指数正相关(ρ=0.503),PTBE区质子密度与AQP4负相关(ρ=-0.506,均p<0.001)。结论:DTI-ALPS及MTP有效评估脑肿瘤类淋巴功能及AQP4调控机制,揭示肿瘤水肿多系统相互作用。第二部分:基于连续时间随机游走模型的脑肿瘤微观异质性表征及免疫细胞浸润评估背景与目的:肿瘤微环境(Tumor Microenvironment,TME)异质性是脑肿瘤诊疗的关键难点。本研究基于连续时间随机游走(Continuous-time Random Walk,CTRW)模型评估脑肿瘤异质性特征,并构建TME免疫细胞浸润预测模型。方法:纳入98例脑肿瘤患者,术前行多b值扩散加权成像(Diffusion-Weighted Imaging,DWI),提取空间异质性(β)、时间异质性(α)及异常扩散系数(Dm)。通过病理及机器学习(Machine Learning,ML)定量评估组织与细胞异质性,构建ML预测模型评估免疫细胞浸润。结果:β值自正常脑组织至肿瘤区呈梯度递减(p<0.01),与病理人工(ρ=-0.948)及ML分析(ρ=-0.870)结果高度相关;α值在脑转移瘤显著低于原发肿瘤(p<0.05),与肿瘤细胞群复杂性负相关(ρ=-0.941);基于梯度提升算法的ML模型对中性粒细胞(R~2=0.994)和巨噬细胞(R~2=0.992)浸润预测精准。结论:CTRW模型有效表征脑肿瘤异质性,结合ML预测免疫浸润,指导个体化治疗决策。第三部分:基于多维扩散MRI与方差分解技术的脑肿瘤微结构特征解析背景与目的:不同类型脑肿瘤具特异微观结构差异。多维扩散MRI(Multidimensional Diffusion Magnetic Resonance Imaging,MDD MRI)及扩散方差分解(Diffusion Variance Decomposition,DIVIDE)技术可定量脑肿瘤微结构特征。本研究评估其与增殖标记物Ki-67、肿瘤抑制基因p53及细胞形态的关联。方法:纳入67例脑肿瘤患者,术前行MDD MRI扫描,测量分数各向异性(Fractional Anisotropy,FA)、平均扩散率(Mean Diffusivity,MD)、各向异性平均峰度(Mean Kurtosis Anisotropy,MKA)、各向同性平均峰度(Mean Kurtosis Isotropic,MKI)、总平均峰度(Mean Kurtosis Total,MKT)、取向有序参数(Orientation Order Parameter,OP)及微观分数各向异性(Microscopic Fractional Anisotropy,μFA)。术后评估Ki-67、p53表达,数字病理分析细胞直径变异系数(Coefficient of Variation,CV)及密度。结果:脑膜瘤与恶性肿瘤间各扩散参数差异显著,μFA与FA鉴别效能最佳(ROC曲线下面积分别为0.952与0.955);MD、MKA及μFA与Ki-67强相关(|ρ|>0.7),μFA最高(ρ=-0.782);MKA及μFA与细胞直径CV显著负相关(ρ≈-0.7,均p<0.001);OP有效区分原发肿瘤与转移瘤(p=0.003)。结论:MDD MRI与DIVIDE技术多尺度解析脑肿瘤微结构特征,与病理指标密切相关,为精准诊断及治疗提供重要依据。图23幅,表5个,参考文献223篇更多还原

【Abstract】 Part Ⅰ:A Histopathologic Correlation Study Evaluating Glymphatic Function in Brain Tumors by Multi-Parametric MRIBackground and Objective:The glymphatic system facilitates waste clearance in the brain through cerebrospinal fluid-interstitial fluid exchange,but its role in brain tumors remains unclear.This study used diffusion tensor imaging(DTI)-based along the perivascular space(ALPS)index and multiparametric magnetic resonance imaging(MRI)to investigate the relationship among glymphatic function,Aquaporin-4(AQP4)expression,peritumoral brain edema(PTBE),and tumor microenvironment.Methods:Eighty-four patients with brain tumors and 59 healthy controls underwent DTI and multiparametric MRI scans.ALPS index,cerebrospinal fluid,tumor,and PTBE volumes were measured,and AQP4expression was evaluated immunohistochemically.Results:ALPS index was significantly lower(2.315 vs.2.879,p=0.001)and cerebrospinal fluid volume significantly larger(p=0.001)in tumor patients compared to controls.Tumor and cerebrospinal fluid volumes negatively correlated with ALPS index(r=-0.715,r=-0.747,respectively);PTBE volume positively correlated with AQP4 expression(r=0.989).Proton density positively correlated with ALPS in tumor regions(ρ=0.503)but negatively with AQP4 in PTBE regions(ρ=-0.506,both p<0.001).Conclusions:DTI-ALPS combined with multiparametric MRI effectively assessed glymphatic dysfunction and AQP4 modulation in brain tumors,elucidating interactions among multiple systems in tumor-associated edema.Part Ⅱ:Characterization of Brain Tumor Microstructural Heterogeneity and Evaluation of Immune Cell Infiltration Using the Continuous-Time Random Walk ModelBackground and Objective:Tumor microenvironment(TME)heterogeneity poses diagnostic and therapeutic challenges in brain tumors.This study applied the continuous-time random walk(CTRW)model to diffusion-weighted imaging(DWI)to characterize TME heterogeneity and developed a predictive model for immune cell infiltration.Methods:Ninety-eight brain tumor patients underwent preoperative multi-b-value DWI.Spatial heterogeneity(β),temporal heterogeneity(α),and anomalous diffusion coefficient(Dm)were extracted.Tissue and cellular heterogeneity were quantified via pathology and machine learning(ML),and predictive models for immune infiltration were established.Results:βvalues showed a decreasing gradient from normal tissue to tumor regions(p<0.01),highly correlating with manual(ρ=-0.948)and ML-based pathological assessments(ρ=-0.870).αvalues were significantly lower in brain metastases than in primary tumors(p<0.05),negatively correlating with tumor cell complexity(ρ=-0.941).ML models using gradient boosting accurately predicted neutrophil(R~2=0.994)and macrophage infiltration(R~2=0.992).Conclusions:The CTRW model effectively characterized brain tumor heterogeneity.Coupled with ML,it accurately predicted immune infiltration,facilitating personalized therapeutic strategies.Part Ⅲ:Analysis of Brain Tumor Microstructural Features Using Multidimensional Diffusion MRI and Diffusional Variance DecompositionBackground and Objective:Brain tumors display distinct microstructural differences.Multidimensional diffusion MRI(MDD MRI)combined with diffusion variance decomposition(DIVIDE)quantitatively assesses these differences.This study explored correlations among diffusion parameters,Ki-67 proliferation marker,p53 expression,and cell morphology.Methods:Sixty-seven brain tumor patients underwent MDD MRI scans.Fractional anisotropy(FA),mean diffusivity(MD),mean kurtosis anisotropy(MKA),mean kurtosis isotropic(MKI),mean kurtosis total(MKT),orientation order parameter(OP),and microscopic fractional anisotropy(μFA)were measured.Ki-67 and p53 expression,cell diameter coefficient of variation(CV),and density were evaluated postoperatively.Results:Significant differences in diffusion parameters between meningiomas and malignant tumors were observed;μFA and FA showed the highest diagnostic accuracy(AUC=0.952,0.955 respectively).MD,MKA,andμFA strongly correlated with Ki-67(|ρ|>0.7,highestμFAρ=-0.782).MKA andμFA negatively correlated with cell diameter CV(ρ≈-0.7,both p<0.001).OP effectively distinguished primary tumors from metastases(p=0.003).Conclusions:MDD MRI with DIVIDE quantitatively characterized brain tumor microstructural heterogeneity,closely correlating with pathological markers,thus enhancing precision diagnosis and treatment planning.更多还原