分享:
分享到微信朋友圈
X
临床研究
DCE-MRI药代动力学参数直方图预测前列腺癌内分泌治疗反应的研究
李静 邹彩霞 潘妮妮 陈俊 黄刚

Cite this article as: LI J, ZOU C X, PAN N N, et al. A study on the histogram of DCE-MRI pharmacokinetic parameters for predicting endocrine therapy response in prostate cancer[J]. Chin J Magn Reson Imaging, 2025, 16(4): 70-80.本文引用格式:李静, 邹彩霞, 潘妮妮, 等. DCE-MRI药代动力学参数直方图预测前列腺癌内分泌治疗反应的研究[J]. 磁共振成像, 2025, 16(4): 70-80. DOI:10.12015/issn.1674-8034.2025.04.011.


[摘要] 目的 探讨基于动态对比增强磁共振成像(dynamic contrast enhancement magnetic resonance imaging, DCE-MRI)药代动力学参数直方图特征预测前列腺癌(prostate cancer, PCa)内分泌治疗反应的价值。材料与方法 回顾性分析2018年1月至2023年10月河西学院附属张掖人民医院(中心1)和2020年2月至2023年2月甘肃省人民医院(中心2)PCa患者在内分泌治疗前2周的临床、影像资料,将中心1收集的105例病例按7∶3的比例分为训练集(73例)和内部验证集(32例),将中心2收集的47例病例作为外部验证集。选取DCE-MRI原始图像,通过Siemens Syngo.via工作站获得药代动力学参数容积转运常数(volume transfer contrast, Ktrans)、速率常数(rate contrast, Kep)、血管外细胞外容积分数(extravascular extracellular volume fraction, Ve)伪彩图。在3D Slicer软件中参照轴位T2WI在药代动力学参数伪彩图上逐层勾画全前列腺腺体感兴趣区(region of interest, ROI)后提取直方图特征,经最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)降维筛选出8个最优特征并计算直方图特征。采用单因素及后向多因素logistic回归分析内分泌治疗反应良好组和不良组的独立预测因素,并构建临床模型、直方图特征模型、联合模型。采用受试者工作特性曲线、校准曲线和决策曲线评价模型的效能,通过DeLong检验评估各模型曲线下面积(area under the curve, AUC),最后基于联合模型的独立预测因素绘制列线图。结果 训练集、内部验证集和外部验证集中治疗反应良好组和不良组之间Gleason评分、MRI-T分期、直方图特征差异均存在统计学意义(P<0.001)。后向多因素logistic回归分析显示Gleason评分(OR=0.925,95% CI:0.859~0.958,P=0.038)、MRI-T分期(OR=0.871,95% CI:0.800~0.949,P=0.002)及直方图特征(OR=0.096,95% CI:0.056~0.137,P<0.001)是PCa内分泌治疗反应的独立预测因素;临床模型在训练集、内部验证集及外部验证集的AUC分别为0.857(95% CI:0.774~0.939)、0.953(95% CI:0.888~0.996)、0.808(95% CI:0.676~0.941);直方图特征模型在训练集、内部验证集及外部验证集的AUC为0.874(95% CI:0.769~0.951)、0.816(95% CI:0.664~0.967)、0.674(95% CI:0.517~0.831);联合模型在训练集、内部验证集及外部验证集的AUC为0.951(95% CI:0.906~0.994)、0.973(95% CI:0.922~0.995)、0.830(95% CI:0.699~0.960);决策曲线和校准曲线分析表明,联合模型具有良好的临床应用价值和稳定性;DeLong检验及NRI值显示联合模型的预测效能优于临床模型和直方图特征模型。结论 DCE-MRI药代动力学参数直方图特征是预测PCa内分泌治疗反应的独立预测因素,联合模型在预测PCa内分泌治疗反应方面具有较好的价值,为临床治疗决策提供了新的思路。
[Abstract] Objective To explore the value of predicting the response of prostate cancer (PCa) to endocrine therapy based on the histogram features of pharmacokinetic parameters of dynamic contrast enhancement magnetic resonance imaging (DCE-MRI).Materials and Methods Retrospectively collect the clinical and imaging data of PCa patients from Zhangye People's Hospital Affiliated to Hexi University from January 2018 to October 2023 and Gansu Provincial People's Hospital from February 2020 to February 2023, two weeks before endocrine therapy. A total of 105 cases were collected from Zhangye People's Hospital Affiliated to Hexi University, which were divided into a training set (73 cases) and an internal validation set (32 cases) at a ratio of 7∶3. A total of 47 cases were collected from Gansu Provincial People's Hospital as an external validation set. Select the original DCE-MRI images, and obtain the pseudo-color maps of pharmacokinetic parameters including volume transfer constant (Ktrans), rate constant (Kep), and extravascular extracellular volume fraction (Ve) through the Siemens Syngovia workstation. In the 3D Slicer software, referring to the axial T2-weighted imaging (T2WI), delineate the region of interest (ROI) of the whole prostate gland layer by layer on the pseudo-color maps of pharmacokinetic parameters, and then extract the histogram features. Through dimensionality reduction by the least absolute shrinkage and selection operator (LASSO), 8 optimal features were screened out and the histogram features was calculated. Univariate and backward multivariate logistic regression were used to analyze the independent predictive factors of the good-response group and the poor-response group of endocrine therapy, and a clinical model, a histogram features model, and a combined model were constructed. The area under the curve (AUC) was calculated using the receiver operating characteristic (ROC) curve, and the calibration curve and decision curve were used to evaluate the performance of the model. The efficacy of each model was evaluated by the DeLong test. Finally, a nomogram was drawn based on the independent predictive factors of the combined model.Results In the training set, internal validation set, and external validation set, there were statistically significant differences in Gleason score, MRI-T stage, and histogram features between the good-response group and the poor-response group (P < 0.001). Backward multivariate logistic regression analysis showed that the Gleason score (OR = 0.925, 95% CI: 0.859 to 0.958, P = 0.038), MRI-T stage (OR = 0.871, 95% CI: 0.800 to 0.949, P = 0.002), and histogram features (OR = 0.096, 95% CI: 0.056 to 0.137, P < 0.001) were independent predictive factors for the response of PCa to endocrine therapy. The AUCs of the clinical model in the training set, internal validation set, and external validation set were 0.857 (95% CI: 0.774 to 0.939), 0.953 (95% CI: 0.888 to 0.996), and 0.808 (95% CI: 0.676 to 0.941), respectively. The AUCs of the histogram features model in the training set, internal validation set, and external validation set were 0.874 (95% CI: 0.769 to 0.951), 0.816 (95% CI: 0.664 to 0.967), and 0.674 (95% CI: 0.517 to 0.831), respectively. The AUCs of the combined model in the training set, internal validation set, and external validation set were 0.951 (95% CI: 0.906 to 0.994), 0.973 (95% CI: 0.922 to 0.995), and 0.830 (95% CI: 0.699 to 0.960), respectively. The analysis of the decision curve and calibration curve showed that the combined model had good clinical application value and stability. The DeLong test and NRI value showed that the predictive efficacy of the combined model was better than that of the clinical model and the histogram features model.Conclusions The histogram features of DCE-MRI pharmacokinetic parameters is an independent predictive factor for predicting the response of PCa to endocrine therapy. The combined model has good value in predicting the response of PCa to endocrine therapy, providing new ideas for clinical treatment decisions.
[关键词] 前列腺癌;磁共振成像;动态对比增强;直方图;内分泌治疗
[Keywords] prostate cancer;magnetic resonance imaging;dynamic contrast enhancement;histogram;endocrine therapy

李静 1, 2   邹彩霞 1   潘妮妮 3   陈俊 4   黄刚 3*  

1 甘肃省中医药大学第一临床医学院,兰州 730000

2 河西学院附属张掖人民医院影像研究所,张掖 734000

3 甘肃省人民医院放射科,兰州 730000

4 拜耳医药保健有限公司影像诊断业务部,武汉 430000

通信作者:黄刚,E-mail:keen0999@163.com

作者贡献声明:黄刚设计本研究的方案,对稿件重要的学术内容进行了修改;李静起草和撰写稿件,获取、分析和解释本研究的数据;邹彩霞、潘妮妮、陈俊分析本研究的数据,对稿件的重要内容进行了修改;黄刚、李静获得了甘肃省卫生健康委卫生健康行业科研项目、甘肃省教育厅高校教师创新基金项目、2024年度张掖市市级科技计划项目、2024年河西学院校长基金创新团队项目资助;全体作者都同意发表最后的修改稿,同意对本研究的所有方面负责,确保本研究的准确性和诚信。


基金项目: 甘肃省卫生健康委卫生健康行业科研项目 GS-62000000001-2024-010 甘肃省教育厅高校教师创新基金项目 2024A-156 2024年度张掖市市级科技计划项目 ZY2024BJ08 河西学院校长基金创新团队项目 CXTD2024014
收稿日期:2024-10-22
接受日期:2025-04-10
中图分类号:R445.2  R737.25 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2025.04.011
本文引用格式:李静, 邹彩霞, 潘妮妮, 等. DCE-MRI药代动力学参数直方图预测前列腺癌内分泌治疗反应的研究[J]. 磁共振成像, 2025, 16(4): 70-80. DOI:10.12015/issn.1674-8034.2025.04.011.

0 引言

       前列腺癌(prostate cancer, PCa)是男性生殖系统常见的恶性肿瘤之一,2022年全球癌症统计显示PCa的发病率和死亡率分别为第2位和第5位[1]。我国PCa的发病率和死亡率分别为第6位和第7位,但大部分患者初次就诊时已处于晚期,错过了手术治疗的最佳时间[2],内分泌治疗是晚期PCa患者标准的治疗方法[3]。但内分泌治疗在不同个体PCa患者中的治疗反应存在差异[4],因此在治疗前如何筛选内分泌治疗有效的患者是临床存在的难题。

       磁共振成像(magnetic resonance imaging, MRI)是无创性诊断PCa首选的影像学方法。基于动态对比增强MRI(dynamic contrast enhancement MRI, DCE-MRI)的药代动力学参数已经被证明可以通过反映肿瘤血管灌注情况及血管通透性评估PCa的内分泌治疗反应[5]。PCa患者经过内分泌治疗后,肿瘤组织和正常前列腺组织的DCE-MRI药代动力学参数容积转运常数(volume transfer contrast, Ktrans)均减低,但是相较于治疗前,治疗后肿瘤组织与正常前列腺组织的Ktrans值差异变小[6],因此与正常前列腺组织相比,肿瘤组织的Ktrans更易受到内分泌治疗影响而产生变化。尽管有研究显示DCE-MRI药代动力学参数图能预测PCa放疗后的疗效[7],但是目前尚缺乏研究探讨其对内分泌治疗反应的预测价值。基于正电子发射计算机断层显影(positron emission tomography computed tomography, PET/CT)的影像组学模型也能预测PCa患者内分泌治疗的疗效(AUC=0.854)[8],但因价格昂贵推广应用较难。有研究构建了基于双参数MRI的深度学习模型预测去势抵抗性PCa发生风险,但其效能并不高(AUC=0.768)[9],并且深度学习模型可解释性差,临床应用受限较多。直方图参数是一种一阶图像特征,反映了区域内像素值大小的空间分布情况。治疗前DCE-MRI药代动力学参数直方图分析在预测食管鳞癌、直肠癌放化疗疗效中均具有较高价值[10, 11]。由于80%的PCa为多灶性生长[12],且研究已经证实肿瘤周围的区域图像特征也能反映肿瘤的生物学行为和治疗应答状态[13]。因此,本研究以全前列腺体作为感兴趣区(region of interest, ROI)提取PCa患者治疗前DCE-MRI药代动力学参数直方图特征,并联合临床资料构建机器学习模型预测内分泌治疗的疗效,为临床分层PCa患者提供新的思路。

1 材料与方法

1.1 研究对象

       本研究遵守《赫尔辛基宣言》,经河西学院附属张掖人民医院(批准号:HFYER-2023.07)和甘肃省人民医院(批准号:2022-458)伦理委员会审查及批准,并免除了受试者书面知情同意。

       回顾性收集2018年1月至2023年10月河西学院附属张掖人民医院(中心1)和2020年2月至2023年2月甘肃省人民医院(中心2)经穿刺活检病理证实的PCa患者资料,包括:初始前列腺特异性抗原(prostate-specific antigen, PSA)、治疗后PSA、年龄、Gleason评分、MRI-T分期、淋巴结转移、骨转移、DCE-MRI影像资料。Gleason评分取细针穿刺活检中的最高评分;MRI-T分期由一位具有10年MRI诊断经验的医生根据MRI图像进行评估;有无淋巴结转移由10年诊断经验的放射科医师参照计算机断层扫描(computed tomography, CT)或MRI或PET-CT评估;骨转移根据骨扫描或PET-CT或MRI评估。将中心1患者按7∶3比例分为训练集和内部验证集,中心2患者作为外部验证集。

       纳入标准:(1)患者经穿刺活检病理证实为PCa;(2)患者在治疗前2周内进行MRI检查;(3)内分泌治疗后睾酮<20 ng/dL。排除标准:(1)临床资料不完整;(2)接受过电切手术、根治手术、放疗、化疗;(3)DCE-MRI图像有伪影;(4)合并其他肿瘤。具体筛选流程详见图1

图1  两个中心病例筛选过程。
Fig. 1  The two-center case screening process.

1.2 MRI扫描方案

       MRI扫描采用3.0 T磁共振扫描仪(中心1:Siemens lumina 214707;中心2:MAGNETOM Skyra,Siemens Healthcare)及采用18通道体部相控阵线圈。患者取仰卧位,扫描序列包括:T1WI轴位、T2WI轴位和矢状位、弥散加权成像(diffusion-weighted imaging, DWI)轴位和动态对比增强(dynamic contrast enhancement, DCE)序列。DCE-MRI序列使用3D容积内插屏气扫描(three-dimensional volume interpolated breath-hold examination, FLASH 3D-VIBE)T1WI轴位,按0.2 mmol/kg的标准准备钆喷酸葡胺注射液(Gd-DTPA,广州康臣药业有限公司)及20 mL生理盐水,以2.5 mL/s的速率用高压注射器在第4期开始同时注射Gd-DTPA和生理盐水,扫描36期。扫描参数详见表1

表1  磁共振序列扫描参数
Tab. 1  Magnetic resonance sequence scanning parameters

1.3 内分泌治疗方案及疗效评价标准

1.3.1 治疗方案

       所有患者在穿刺活检得到病理证实后8周内采用去势药物联合抗雄性激素药物进行内分泌治疗,治疗周期为6个月(约168天)。中心1的105例患者中,57例患者用戈舍瑞林+比卡鲁胺(醋酸戈舍瑞林3.6 mg,每28日1次,或10.8 mg,每84日1次,皮下注射;比卡鲁胺50 mg,每日1次,口服);48例患者用亮丙瑞林+恩杂鲁胺(醋酸亮丙瑞林3.75 mg,每28日1次,皮下注射;恩杂鲁胺50 mg,每日1次,口服)。中心2的47例患者中,25例患者进行戈舍瑞林+比卡鲁胺(醋酸戈舍瑞林3.6 mg,每28日1次,或10.8 mg,每84日1次,皮下注射;比卡鲁胺50 mg,每日1次,口服),22例进行亮丙瑞林+比卡鲁胺(醋酸亮丙瑞林3.75 mg,28日1次,皮下注射;比卡鲁胺50 mg,每日1次,口服)。

1.3.2 疗效评价标准

       所有患者在内分泌治疗后6个月通过PSA检测评价其治疗效果,根据中国临床肿瘤学会(Chinese Society of Clinical Oncology, CSCO)《CSCO前列腺癌诊疗指南》[14],0.2 ng/mL<PSA<4 ng/mL生存时间中位数为44个月,PSA≤0.2 ng/mL生存时间中位数为75个月,所以本研究将PSA>0.2 ng/mL的患者分为治疗反应不良组,PSA≤0.2 ng/mL的患者分为治疗反应良好组。

1.4 DCE-MRI药代动力学参数图像处理及分割

       将DCE-MRI原始图像传至Siemens Syngo.via工作站,对原始图像进行运动校正及配准;选择Tofts模型,通过计算生成前列腺Ktrans、速率常数(rate contrast, Kep)、血管外细胞外容积分数(extravascular extracellular volume fraction, Ve)伪彩图;导入3D Slicer软件(版本号5.4.0),由一名医学影像科从事前列腺诊断工作的副主任医生参考T2WI图像在DCE-MRI参数图中逐层勾画全前列腺腺体为ROI(图2);1周后随机选取50例患者再次勾画前列腺ROI评估直方图特征的一致性和稳定性。

图2  全前列腺分割示意图。2A:T2WI图;2B:容积转运常数(Ktrans)伪彩图;2C:三维感兴趣区(ROI)图。
Fig. 2  Schematic diagram of Total prostate segmentation. 2A: T2-weighted image; 2B: Pseudo color plot of volumetric transport constant (Ktrans); 2C: The diagram of three-dimensional region of interest (ROI).

1.5 DCE-MRI药代动力学参数直方图特征提取、筛选

       使用FeAture Explorer(V.0.55;http://github.com/salan668/FAE)软件导入分割后ROI,为标准化体素间距将图像重新采样到体素大小为1 mm×1 mm×1 mm,通过使用25 SI的固定箱宽度(binwidth)对图像体素强度值进行离散化,将体素强度归一化至1~100之间,以减少不同机器采集图像信号强度的差异。从每个药代动力学参数图中提取18个直方图特征,包括10%灰度强度(10Percentile)、90%灰度强度(90Percentile)、能量(Energy)、熵(Entropy)、四分位范围(Interquartile Range)、峰度(Kurtosis)、最大灰度强度(Maximum)、平均灰度强度(Mean Absolute Deviation)、均值(Mean)、中位灰度强度(Median)、最小灰度强度(Minimum)、灰度值范围(Range)、平均绝对偏差(Robust Mean Absolute Deviation)、均方根(Root Mean Squared)、偏度(Skewness)、总能量(Total Energy)、均匀度(Uniformity)、方差(Variance)。直方图特征遵循成像生物标志物标准化倡议指南[15],对直方图特征通过Z-score进行标准化处理,用R语言最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)算法筛选出价值最大的特征,通过选择的特征线性组合计算,转化为直方图特征评分。

1.6 模型构建及评估效能

       对单因素及后向多因素logistic回归分析有统计学意义的临床资料构建临床模型;利用DCE-MRI药代动力学直方图特征构建机器学习模型,并计算其直方图特征评分;联合直方图特征及临床资料构建联合模型。临床模型、直方图特征模型、联合模型的判别效能通过受试者工作特性曲线下面积(area under the curve, AUC)、敏感度、特异度等方法评估,并进行DeLong检验比较所构建模型的AUC;校准曲线用来评估各模型的预测值与真实值之间的拟合度;用决策曲线估计不同阈值概率的净效益,从而评估各模型的临床适用性;最后绘制列线图直观展示联合模型的临床应用价值。

1.7 统计学分析

       使用SPSS 22.0、R语言以及Python(version 3.7.12)软件进行统计学分析以及机器学习模型构建,检验水准α=0.05。使用Kolmogorv-Snimov检验验证连续性变量是否服从正态分布,正态分布的变量用均值±标准差表示,采用独立样本t检验比较组间差异,非正态分布的变量用中位数(第25,75百分位数)表示,采用Mann-Whitney U检验比较组间差异。分类数据以频数和百分比表示,并使用卡方检验或Fisher精确检验进行差异性检验。对临床资料、直方图特征评分采用logistic回归进行单因素、后向多因素分析,多因素分析有统计学意义的结果用于构建模型,通过AUC、校准曲线和决策曲线、DeLong检验及NRI值分析各模型的效能和实用性。采用组间相关系数(intra-class correlation coefficient, ICC)来评估两次勾画的一致性,ICC值大于0.75表示可信度良好。

2 结果

2.1 一般资料

       根据纳排标准,最终纳入中心1病例105例和中心2病例47例。临床资料分析结果显示MRI-T分期、Gleason评分在训练集、内部验证集、外部验证集中差异均有统计学意义(P<0.05)(表2)。

表2  训练集、内部验证集、外部验证集前列腺癌患者临床资料比较
Tab. 2  Comparison of clinical data of PCa patients in the training set, internal validation set, and external validation set

2.2 DCE-MRI药代动力学参数直方图特征筛选

       对观察者两次勾画的ROI提取54个直方图特征进行ICC分析,选择ICC>0.75的48个特征,通过Z-score标准化及LASSO特征降维最终从Ktrans、Kep、Ve中筛选出8个最优特征(表3),对8个最优特征参数在不同治疗反应组做箱线图分析(图3),结果显示两组之间差异具有统计学意义(P<0.01)。

图3  DCE-MRI 药代动力学参数直方图特征的箱线图。*表示P<0.01;**表示P<0.001。DCE:动态对比增强。
Fig. 3  Box plot of histogram features of DCE-MRI pharmacokinetic parameters. * represents P<0.01; ** represents P<0.001. DCE: dynamic contrast enhancement.
表3  DCE-MRI药代动力学参数直方图特征和系数
Tab. 3  Histogram characteristics and coefficients of DCE-MRI quantitative parameters

2.3 模型构建及效能比较

       后向多因素logistic回归分析显示Gleason评分、MRI-T分期、直方图特征评分差异有统计学意义(P<0.001)(表4),分别构建基于Gleason评分和MRI-T分期的临床模型、直方图特征模型及联合模型并比较模型效能(表5图4)。临床模型的AUC在训练集为0.857,内部验证集为0.953,外部验证集0.808,直方图特征模型的AUC在训练集为0.874,内部验证集0.816,外部验证集0.674,联合模型的AUC在训练集为0.951,内部验证集0.973,外部验证集0.830,联合模型的预测效能最优,Hosmer-Lemesow检验P值>0.05,决策曲线评估临床适用性较好。DeLong检验结果显示联合模型与直方图特征模型在训练集、内部验证集、外部验证集的P值均<0.05,但联合模型与临床模型只在训练集P值<0.05,内部验证集和外部验证集P值均>0.05(表6)。NRI值显示联合模型与临床模型在训练集、内部验证集、外部验证集中分别为0.187、0.184、0.098,说明联合模型的预测效能比临床模型是有改进的。基于联合模型的独立预测因素绘制列线图并附典型病例(图5, 6, 7)。

图4  训练集、内部验证集、外部验证集的ROC 曲线(4A~4C)、校准曲线(4D~4F)以及决策曲线(4G~4I)。ROC:受试者工作特征;AUC:曲线下面积。Histogram model 为直方图模型;Clinical model 为临床模型;Nomogram model 为列线图模型即联合模型。
Fig. 4  ROC curves (4A-4C), calibration curves (4D-4F), and decision curves (4G-4I) for the training, internal, and external validation sets. ROC: receiver operating characteristic; AUC: area under the curve. Nomogram model represents combined model.
图5  列线图。HistogramScore为直方图特征,Gleason为Gleason评分,T为MRI-T分期。
Fig. 5  Nomogram. HistogramScore is histogram feature; Gleason is Gleason score, T is MRI-T stage.
图6  男,81岁,前列腺癌内分泌治疗反应良好患者。6A~6D分别为T2WI图、Ktrans伪彩图、Kep伪彩图、Ve伪彩图,Ktrans均值3.978 7,中位数39.897 8,最大值453,90%百分位数26.189 3,偏度26.189 3,Kep偏度26.189 3,Ve中位数103,Ve最大值380,计算直方图特征评分0.673 9(93分),Gleason评分为4+4=8(22分),MRI-T分期为3a(14分),最终得分93分+22分+14分=129分,预测治疗反应好的概率为98.7%。
图7  男,72岁,前列腺癌内分泌治疗反应不良患者。7A~7D分别为T2WI图、Ktrans伪彩图、Kep伪彩图、Ve伪彩图,Ktrans均值3.973 8,中位数26.572 5,最大值524,90%百分位数39.117 5,偏度39.117 5,Kep偏度39.117 5,Ve中位数177,Ve最大值999,计算直方图特征评分-2.345 3(52分),Gleason评分为5+4=9(8分),MRI-T分期为3b(8分),最终得分52分+8分+8分=68分,预测治疗反应好的概率为18.2%。
Fig. 6  Male, 81 years old, PCa responded well to endocrine therapy patient. 6A-6D is T2WI image, pseudo color plot of Ktrans, pseudo color plot of Kep, and pseudo color plot of Ve, respectively. Ktrans mean of 3.978 7, median of 39.897 8, maximum of 453, 90% percentile of 26.189 3, skewness of 26.189 3, Kep skewness of 26.189 3, Ve median of 103, Ve maximum of 380, calculated histogram feature score of 0.673 9 (93 points), Gleason score of 4+4=8 (22 points), MRI-T stage of 3a (14 points), final score of 93 points + 22 points + 14 points = 129 points, predicting a good response to treatment with a probability of 98.7%.
Fig. 7  Male, 72 years old, poor response to PCa endocrine therapy patient. 7A-7D is T2WI image, pseudo color plot of Ktrans, pseudo color plot of Kep, and pseudo color plot of Ve, respectively. Ktrans mean of 3.973 8, median of 26.572 5, maximum of 524, 90% percentile of 39.117 5, skewness of 39.117 5, Kep skewness of 39.117 5, median of Ve of 177, maximum of Ve of 999, calculated histogram feature score of -2.345 3 (52 points), Gleason score of 5+4 = 9 (8 points), an MRI-T score of 3b (8 points), final score of 52 points + 8 points + 8 points = 68 points, predicting a good response to treatment is 18.2%.
表4  临床变量与直方图特征的单因素、多因素logistic回归分析结果
Tab. 4  Univariate and multivariate logistic regression analysis of clinical variables and histogram characteristics
表5  临床模型、直方图模型、联合模型对前列腺癌内分泌治疗的预测效能比较
Tab. 5  Comparison of the predictive performance of clinical models, histogram models, and combined models for PCa endocrine therapy
表6  各模型在训练集、内部验证集、外部验证集中的DeLong检验
Tab. 6  DeLong test of models in the training set, internal validation set, and external validation set

2.4 PCa不同内分泌治疗药物的亚组分析

       将所有患者按内分泌治疗药物的不同分为戈舍瑞林+比卡鲁胺、亮丙瑞林+恩杂鲁胺、亮丙瑞林+比卡鲁胺3组之后进行亚组分析,结果显示不同药物治疗组内差异无统计学意义(P>0.05),交互P值为0.957说明不同治疗药物组之间没有交互作用,治疗效果无显著性差异,不会影响模型的预测效能(图8)。

图8  前列腺癌不同内分泌治疗药物组的亚组分析及森林图。0代表治疗反应不良;1代表治疗反应良好;Treat1代表戈舍瑞林+比卡鲁胺;Treat2代表亮丙瑞林+恩杂鲁胺,Treat3代表亮丙瑞林+比卡鲁胺;P for interraction代表交互P值。
Fig. 8  Subgroup analysis of different endocrine therapy drug groups and forest diagram. 0 represents poor treatment response; 1 represents a good response to treatment; Treat1 stands for goserelin bicalutamide; Treat2 stands for leuprolide and enzalutamide, and Treat3 stands for leuprolide and bicalutamide.

3 讨论

       本研究基于全前列腺DCE-MRI药代动力学参数直方图特征构建了机器学习模型预测PCa内分泌治疗反应,探讨了直方图特征模型、临床模型,以及联合模型对PCa内分泌治疗反应的预测价值,结果显示联合模型在训练集、内部验证集和外部验证集中均具有更高的预测能力(AUC分别为0.951,0.973,0.830)。本研究构建了具有较高效能的联合模型预测PCa内分泌治疗反应,为临床PCa患者的精准治疗提供了新的思路和方法。

3.1 DCE-MRI药代动力学参数在PCa内分泌治疗反应的预测价值

       内分泌治疗是局部晚期和转移性PCa的首选治疗方式[3],通过降低雄激素或阻断雄激素与雄激素受体(androgen receptor, AR)结合抑制前列腺腺体上皮产生血管内皮生长因子,诱导内皮细胞凋亡,从而抑制肿瘤细胞的生长和扩散[16]。对于晚期PCa其标准的治疗方式是最大限度的雄激素阻断[17],其治疗方案包括联合去势和抗雄激素治疗,去势药物主要作用是抑制雄激素的产生,抗雄药物通过抑制AR受体的信号传导来抑制肿瘤生长。内分泌治疗后去势抵抗性PCa如果发生在治疗后6个月内,其死亡率会明显上升[18],如果能在治疗前识别这些患者从而通过调整治疗方案进行干预能够提升PCa患者的治疗效果。

       DCE-MRI药代动力学参数是通过数学模型计算出定量反映肿瘤组织微循环灌注状态的一组参数,能够表征肿瘤组织新生血管生成、毛细血管通透性和肿瘤乏氧等状态[19, 20]。以往研究中已经证实了DCE-MRI定量参数在多个与PCa相关的重要领域具备较高的应用价值,在显著性PCa的诊断中通过对定量参数的分析能够更为准确地识别出具有临床意义的PCa病灶[21],在PCa侵袭性评估中可以为判断肿瘤的侵袭能力提供关键信息,辅助医生制订更合理的治疗方案[22],在PCa放疗后正常组织和肿瘤组织的变化等方面有较高价值[23],也可用于监测PCa放疗后正常组织和肿瘤组织的变化情况,帮助医生及时了解治疗效果及组织的反应[24]。但以往的研究尚不清楚DCE-MRI定量参数在PCa内分泌治疗前能否预测治疗反应。PCa内分泌治疗反应受到多种因素的影响,如肿瘤组织的特性及患者自身的身体状况[25],其中肿瘤组织的异质性是一个极为关键的影响因素[26]。基于MRI的影像组学模型在预测PCa放疗后以及根治性切除术后生化复发的发生方面均展现出了较高的价值[27, 28],但影像组学特征也存在一些局限性,其可解释性相对较差[15],而诸如纹理特征、小波特征等极易受到扫描设备的类型、扫描技术的差异以及图像伪影的影响[29],这些因素限制了组学特征模型在临床实践中的广泛应用。本研究中利用能够反映组织微循环状态的DCE-MRI定量参数图提取了能够反映肿瘤异质性且相对较为稳定的全前列腺腺体ROI直方图特征,以此构建模型来预测PCa患者的内分泌治疗反应。研究结果证实DCE-MRI定量参数不仅能够评估内分泌治疗疗效[6],还具备预测治疗反应的能力。由于DCE-MRI定量参数具有明确的病理生理学含义,这使得本研究所构建的基于DCE-MRI定量参数的直方图特征模型拥有良好的可解释性。BJÖRELAND等[6]研究显示Ktrans平均值、中位数、最大值、95%百分位数在PCa内分泌治疗前肿瘤组织和正常组织之间存在显著差异,而在治疗后肿瘤组织和正常组织中无显著差异,可见这些直方图特征的变化能反映内分泌治疗的敏感性。本研究直方图特征模型显示较高的Ktrans均值、中位数、最大值、90%百分位数、偏度会导致内分泌治疗反应不良,Ktrans是单位时间内从血浆进入到组织间隙的对比剂的转运容积[30, 31],它能够直观地反映肿瘤局部微血管密度及毛细血管通透性。随着PCa恶性程度的增加,其微血管的生长不均匀性和复杂性也随之上升,所以恶性程度越高的肿瘤往往具有更高的Ktrans值,导致其治疗效果变差。此外,高Ktrans值还与组织的乏氧诱导因子密切相关[32],Ktrans越高的肿瘤组织更容易处于乏氧状态,乏氧条件下AR的稳定性和转录活性会增强[33],故而内分泌治疗的效果不佳。在预测PCa根治术后手术切缘阳性的研究中显示PCa高危组的Ktrans最大值高于低危组[34],所以Ktrans最大值是预测根治术后手术切缘阳性的独立预测因素。但在宫颈癌新辅助化疗[35]、食管鳞癌放化疗[10]研究中表现出了不同的结果,即治疗反应好的肿瘤具有更高的Ktrans值,因为Ktrans高的肿瘤血管通透性高,能够将更大量的化疗药物运输到肿瘤,对肿瘤细胞起到直接杀灭的作用,表现出更好的治疗反应[36]。造成这种不同研究结果的原因可能主要是肿瘤间存在异质性,不同肿瘤的生物学特性差异较大;其他原因包括研究人群的年龄、基础疾病、遗传背景等因素的不同;还可能是不同肿瘤对不同药物的敏感性和反应机制的差异所致。在本研究中,Kep偏度以及Ve中位数和最大值也被纳入模型中,治疗反应不良组的Kep、Ve值较治疗反应良好组高。Kep是单位时间内从血管外细胞外间隙重新流回血管内的对比剂的转运速率,它与新生血管的数量、表面积及通透性密切相关[37, 38]。头颈部鳞状细胞癌患者中Kep偏度较大的肿瘤预后较差[39],可能是因为Kep偏度越大,肿瘤的复杂性和异质性会相应增加[40, 41],而且肿瘤内部缺氧和坏死也会增加[42, 43],肿瘤内的低氧细胞对治疗效果具有抵抗作用[44],所以Kep偏度越大的肿瘤治疗效果不好。Ve代表单位体积内血管外细胞外间隙的容积分数,可作为反映细胞密度的标志物[45],Ve高的肿瘤其内细胞凋亡速度快且肿瘤的坏死较多[46],肿瘤具有较大的血管外细胞外间隙,不利于药物从血管进入肿瘤组织,故而肿瘤的治疗效果不及Ve低的肿瘤好[35],本研究与以往的研究结果一致。基于DCE-MRI定量参数的直方图特征模型在训练集和内部验证集的AUC分别为0.816、0.874,但在外部验证集的AUC仅有0.674,分析原因可能是外部验证集的病例数相对较少,后续需要扩大样本量来全面提高模型的泛化能力;还可能与两家医院的DCE-MRI的时间分辨率、空间分辨率、翻转角等存在差异,影响了对比剂的动态变化过程及ROI的定位和信号的准确性,进而影响了定量参数值,后期可进行前瞻性研究统一扫描参数进而提高定量参数的准确性。

       80%的PCa为多灶性生长[12],且肿瘤周围区域的肿瘤微环境(肿瘤周围的细胞、血管、生长因子及其他分子组成的环境)也是肿瘤进展的重要影响因素[26, 47, 48],所以本研究采用了全前列腺腺体ROI来提取DCE-MRI药代动力学参数直方图特征,全前列腺的兴趣区勾画相对可重复性更好,使得模型的效能会更少受到手动分割的影响。

3.2 临床因素对PCa内分泌治疗反应的预测价值

       以往研究关于治疗前PSA的差异是否会对内分泌治疗产生影响的研究结果不一致,有学者[49]认为治疗前PSA会显著影响内分泌治疗结局,也是发生去势抵抗性PCa的危险因素[50]。另有研究[51]显示治疗前PSA仅在单因素分析中有差异,而多因素分析无意义,与本研究结论一致,可能是因为Gleason评分比较高的患者前列腺正常腺泡结构破坏更明显,从而使PSA的产生减少,所以治疗前PSA低的患者不一定治疗反应会好,本研究PSA单因素分析有意义而多因素分析无意义还可能因为2个中心血清学检查设备和参数设置不同。本研究发现PCa组织的Gleason评分和治疗前的MRI-T分期也能够较好预测内分泌治疗反应,这为临床分层不同内分泌治疗反应的患者提供了简便可行的方法。Gleason评分是PCa组织中主要结构和次要结构病理分级的总和,反映了PCa的恶性程度[52],恶性程度高的PCa分化差,血管生成因子和AR显著增加[53],因此从内分泌治疗中获益较少[54, 55]。MRI-T分期是衡量肿瘤向外侵犯的指标,往往与肿瘤组织的体积和恶性程度密切相关[56, 57],越高的MRI-T分期实际上也是代表着肿瘤组织更大体积和更高的Gleason评分。高的Gleason评分与PCa根治术后[58]和放疗[59]患者较差的预后也相关。基于Gleason评分、MRI-T分期构建的临床模型在本研究的AUC达到了0.808以上,但基于穿刺获取的组织Gleason评分结果存在抽样偏倚,并不能完全代表肿瘤整体的Gleason评分,此外,穿刺存在的不适感和可能存在的并发症使得部分患者不能接受[60, 61]

3.3 联合模型在PCa内分泌治疗反应的预测价值

       本研究联合DCE-MRI定量参数直方图特征以及临床资料构建的联合模型在预测PCa内分泌治疗反应方面展现出最好的效能,在训练集的AUC为0.951,内部验证集的AUC为0.973,外部验证集上AUC也达到了0.830。联合模型整合了DCE-MRI定量参数丰富的影像信息以及病理信息,相较于单独使用直方图模型或临床模型其预测效能得到了显著提升。说明PCa内分泌治疗反应受到多种因素的综合影响,在构建预测模型时尽可能全面地纳入这些影响因素能够有效地提升模型的预测准确性。一项预测去势抵抗性PCa发生风险的研究中显示,当仅基于双参数MRI和病理指标分别构建深度学习模型时其AUC分别为0.768和0.752,但当将MRI和病理指标联合起来构建深度学习模型的AUC大幅提高至0.860[9]。这一对比鲜明地体现了多因素联合在提升模型预测能力方面的重要作用。另一项关于预测PCa根治术后生化复发的研究[62]构建的放射组学联合病理组学模型在训练集和测试集的C指数分别为0.867和0.804,显著优于单独的放射组学模型(C指数在训练集为0.837,验证集为0.742),这些系列研究都有力地证明了综合多种因素构建模型的优势。列线图是一种实用性很强的临床简易性工具,它将复杂的回归公式转化为可视化图形,近年来得到了广泛的临床应用[63, 64]。本研究展示了2例不同治疗反应的病例,图6病例患者Ktrans均值、中位数、最大值、90%百分位数、偏度、Kep偏度、Ve中位数、Ve最大值均低于图7所示病例,图7患者与图6患者相比较高的Ktrans导致组织更容易处于乏氧状态,乏氧条件下AR的稳定性和转录活性增强,而Kep偏度较高的肿瘤复杂性和异质性增加,肿瘤内部出现缺氧和坏死,低氧细胞对治疗效果产生抵抗作用,另外Ve高的肿瘤不利于药物从血管进入肿瘤组织,所以DCE-MRI定量参数直方图可以解释内分泌治疗反应不良的原因;从临床因素考虑图7患者Gleason评分和MRI-T分期较图6患者高,说明图7患者肿瘤组织体积更大且恶性程度更高,所以治疗反应不良的概率更高。由此可见,综合DCE-MRI定量参数直方图和临床因素不仅可以很好地解释治疗反应不良的原理,还能早期准确预测治疗反应不良的病例,为临床医生及时调整治疗方案提供依据。

3.4 本研究的局限性

       (1)本研究是回顾性研究,样本量较小,结果需要在更大的样本量中进一步验证;(2)本研究选取了DCE-MRI药代动力学参数图进行直方图分析,虽然模型的可解释性较好,但是没有充分利用多参数MRI信息,后期还需纳入多参数MRI进一步提高模型的稳定性和效能;(3)使用了两家医疗机构的影像数据和临床资料,尽管对影像图像进行了标准化处理,但临床指标存在一定的异质性,可能会影响模型的预测效能。

4 结论

       基于DCE-MRI药代动力学参数直方图特征构建的机器学习模型对预测PCa患者内分泌治疗反应有较好的价值,结合临床构建联合模型能够提升预测效能,为PCa患者制订精准治疗决策提供帮助。

[1]
FILHO A M, LAVERSANNE M, FERLAY J, et al. The GLOBOCAN 2022 cancer estimates: Data sources, methods, and a snapshot of the cancer burden worldwide[J]. Int J Cancer, 2025, 156(7): 1336-1346. DOI: 10.1002/ijc.35278.
[2]
HAN B F, ZHENG R S, ZENG H M, et al. Cancer incidence and mortality in China, 2022[J]. J Natl Cancer Cent, 2024, 4(1): 47-53. DOI: 10.1016/j.jncc.2024.01.006.
[3]
LIU W, YAO Y, LIU X, et al. Neoadjuvant hormone therapy for patients with high-risk prostate cancer: a systematic review and meta-analysis[J]. Asian J Androl, 2021, 23(4): 429-436. DOI: 10.4103/aja.aja_96_20.
[4]
WILKINSON S, HARMON S A, TERRIGINO N T, et al. A case report of multiple primary prostate tumors with differential drug sensitivity[J]. Nat Commun, 2020, 11(1): 837. DOI: 10.1038/s41467-020-14657-7.
[5]
ALONZI R, PADHANI A R, JANE TAYLOR N, et al. Antivascular effects of neoadjuvant androgen deprivation for prostate cancer: an in vivo human study using susceptibility and relaxivity dynamic MRI[J]. Int J Radiat Oncol Biol Phys, 2011, 80(3): 721-727. DOI: 10.1016/j.ijrobp.2010.02.060.
[6]
BJÖRELAND U, NYHOLM T, JONSSON J, et al. Impact of neoadjuvant androgen deprivation therapy on magnetic resonance imaging features in prostate cancer before radiotherapy[J]. Phys Imaging Radiat Oncol, 2021, 17: 117-123. DOI: 10.1016/j.phro.2021.01.004.
[7]
RAKAUSKAS A, SHAH T T, PETERS M, et al. Can quantitative analysis of multi-parametric MRI independently predict failure of focal salvage HIFU therapy in men with radio-recurrent prostate cancer?[J/OL]. Urol Oncol, 2021, 39(12): 830.e1-830.e8 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC8639607/. DOI: 10.1016/j.urolonc.2021.04.017.
[8]
QIAO J Y, LIU B T, XIN J, et al. Prediction of prognosis and response to androgen deprivation therapy in intermediate to high-risk prostate cancer using 18F-FDG PET/CT radiomics[J]. Acad Radiol, 2024, 31(12): 5008-5021. DOI: 10.1016/j.acra.2024.06.034.
[9]
ZHOU C, ZHANG Y F, GUO S, et al. Multimodal data integration for predicting progression risk in castration-resistant prostate cancer using deep learning: A multicenter retrospective study[J/OL]. Front Oncol, 2024, 14: 1287995 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC10972942/. DOI: 10.3389/fonc.2024.1287995.
[10]
SUN N N, GE X L, LIU X S, et al. Histogram analysis of DCE-MRI for chemoradiotherapy response evaluation in locally advanced esophageal squamous cell carcinoma[J]. Radiol Med, 2020, 125(2): 165-176. DOI: 10.1007/s11547-019-01081-1.
[11]
PHAM T T, LINEY G, WONG K, et al. Multi-parametric magnetic resonance imaging assessment of whole tumour heterogeneity for chemoradiotherapy response prediction in rectal cancer[J]. Phys Imaging Radiat Oncol, 2021, 18: 26-33. DOI: 10.1016/j.phro.2021.03.003.
[12]
SKOTHEIM R I, BOGAARD M, CARM K T, et al. Prostate cancer: Molecular aspects, consequences, and opportunities of the multifocal nature[J/OL]. Biochim Biophys Acta Rev Cancer, 2024, 1879(2): 189080 [2024-10-21]. https://pubmed.ncbi.nlm.nih.gov/38272101/. DOI: 10.1016/j.bbcan.2024.189080.
[13]
ZHAO Y, ZHANG J, WANG N, et al. Intratumoral and peritumoral radiomics based on contrast-enhanced MRI for preoperatively predicting treatment response of transarterial chemoembolization in hepatocellular carcinoma[J/OL]. BMC Cancer, 2023, 23(1): 1026 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC10594790/. DOI: 10.1186/s12885-023-11491-0.
[14]
MOTTET N, VAN DEN BERGH R C N, BRIERS E, et al. EAU-EANM-ESTRO-ESUR-SIOG guidelines on prostate cancer-2020 update. part 1: screening, diagnosis, and local treatment with curative intent[J]. Eur Urol, 2021, 79(2): 243-262. DOI: 10.1016/j.eururo.2020.09.042.
[15]
ZWANENBURG A, VALLIÈRES M, ABDALAH M A, et al. The image biomarker standardization initiative: standardized quantitative radiomics for high-throughput image-based phenotyping[J]. Radiology, 2020, 295(2): 328-338. DOI: 10.1148/radiol.2020191145.
[16]
DESAI K, MCMANUS J M, SHARIFI N. Hormonal therapy for prostate cancer[J]. Endocr Rev, 2021, 42(3): 354-373. DOI: 10.1210/endrev/bnab002.
[17]
牛向南, 张雁钢. 晚期前列腺癌内分泌治疗的研究进展[J]. 现代肿瘤医学, 2024, 32(5): 973-977. DOI: 10.3969/j.issn.1672-4992.2024.05.033.
NIU X N, ZHANG Y G. Advances in endocrine therapy in advanced prostate cancer[J]. J Mod Oncol, 2024, 32(5): 973-977. DOI: 10.3969/j.issn.1672-4992.2024.05.033.
[18]
HUSSAIN M, GOLDMAN B, TANGEN C, et al. Prostate-specific antigen progression predicts overall survival in patients with metastatic prostate cancer: data from Southwest Oncology Group Trials 9346 (Intergroup Study 0162) and 9916[J]. J Clin Oncol, 2009, 27(15): 2450-2456. DOI: 10.1200/JCO.2008.19.9810.
[19]
YANKEELOV T E, GORE J C. Dynamic contrast enhanced magnetic resonance imaging in oncology: theory, data acquisition, analysis, and examples[J]. Curr Med Imaging Rev, 2009, 3(2): 91-107. DOI: 10.2174/157340507780619179.
[20]
PETRALIA G, SUMMERS P E, AGOSTINI A, et al. Dynamic contrast-enhanced MRI in oncology: how we do it[J]. Radiol Med, 2020, 125(12): 1288-1300. DOI: 10.1007/s11547-020-01220-z.
[21]
FENG X, CHEN X, PENG P, et al. Values of multiparametric and biparametric MRI in diagnosing clinically significant prostate cancer: a multivariate analysis[J/OL]. BMC Urol, 2024, 24(1): 40 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC10870467/. DOI: 10.1186/s12894-024-01411-0.
[22]
FUKUNAGA T, TAMADA T, KANOMATA N, et al. Quantitative diffusion-weighted imaging and dynamic contrast-enhanced MR imaging for assessment of tumor aggressiveness in prostate cancer at 3T[J]. Magn Reson Imaging, 2021, 83: 152-159. DOI: 10.1016/j.mri.2021.08.009.
[23]
WU X, REINIKAINEN P, KAPANEN M, et al. Monitoring radiotherapy induced tissue changes in localized prostate cancer by multi-parametric magnetic resonance imaging (MP-MRI)[J]. Diagn Interv Imaging, 2019, 100(11): 699-708. DOI: 10.1016/j.diii.2019.06.003.
[24]
HÖTKER A M, MAZAHERI Y, ZHENG J T, et al. Prostate Cancer: assessing the effects of androgen-deprivation therapy using quantitative diffusion-weighted and dynamic contrast-enhanced MRI[J]. Eur Radiol, 2015, 25(9): 2665-2672. DOI: 10.1007/s00330-015-3688-1.
[25]
PIGNOT G, MAILLET D, GROSS E, et al. Systemic treatments for high-risk localized prostate cancer[J]. Nat Rev Urol, 2018, 15(8): 498-510. DOI: 10.1038/s41585-018-0017-x.
[26]
GE R B, WANG Z W, CHENG L. Tumor microenvironment heterogeneity an important mediator of prostate cancer progression and therapeutic resistance[J/OL]. NPJ Precis Oncol, 2022, 6(1): 31 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC9068628/. DOI: 10.1038/s41698-022-00272-w.
[27]
YAN Y, SHAO L Z, LIU Z Y, et al. Deep learning with quantitative features of magnetic resonance images to predict biochemical recurrence of radical prostatectomy: A multi-center study[J/OL]. Cancers, 2021, 13(12): 3098 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC8234539/. DOI: 10.3390/cancers13123098.
[28]
LI L, SHIRADKAR R, LEO P, et al. A novel imaging based Nomogram for predicting post-surgical biochemical recurrence and adverse pathology of prostate cancer from pre-operative bi-parametric MRI[J/OL]. EBioMedicine, 2021, 63: 103163 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC7744939/. DOI: 10.1016/j.ebiom.2020.103163.
[29]
SOLLINI M, ANTUNOVIC L, CHITI A, et al. Towards clinical application of image mining: a systematic review on artificial intelligence and radiomics[J]. Eur J Nucl Med Mol Imaging, 2019, 46(13): 2656-2672. DOI: 10.1007/s00259-019-04372-x.
[30]
NAGASAKA K, SATAKE H, ISHIGAKI S, et al. Histogram analysis of quantitative pharmacokinetic parameters on DCE-MRI: correlations with prognostic factors and molecular subtypes in breast cancer[J]. Breast Cancer, 2019, 26(1): 113-124. DOI: 10.1007/s12282-018-0899-8.
[31]
VAN NIEKERK C G, VAN DER LAAK J A W M, HAMBROCK T, et al. Correlation between dynamic contrast-enhanced MRI and quantitative histopathologic microvascular parameters in organ-confined prostate cancer[J]. Eur Radiol, 2014, 24(10): 2597-2605. DOI: 10.1007/s00330-014-3301-z.
[32]
BORREN A, GROENENDAAL G, VAN DER GROEP P, et al. Expression of hypoxia-inducible factor-1α and-2α in whole-mount prostate histology: relation with dynamic contrast-enhanced MRI and Gleason score[J]. Oncol Rep, 2013, 29(6): 2249-2254. DOI: 10.3892/or.2013.2392.
[33]
MOHAMED O A A, TESEN H S, HANY M, et al. The role of hypoxia on prostate cancer progression and metastasis[J]. Mol Biol Rep, 2023, 50(4): 3873-3884. DOI: 10.1007/s11033-023-08251-5.
[34]
CHANG C B, LIN Y C, WONG Y C, et al. Quantitative dynamic contrast-enhanced magnetic resonance parameters could predict international society of urological pathology risk groups of prostate cancers on radical prostatectomy[J/OL]. Life, 2023, 13(9): 1944 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC10532885/. DOI: 10.3390/life13091944.
[35]
ZHANG A N, SONG J C, MA Z L, et al. Combined dynamic contrast-enhanced magnetic resonance imaging and diffusion-weighted imaging to predict neoadjuvant chemotherapy effect in FIGO stage IB2-IIA2 cervical cancers[J]. La Radiol Med, 2020, 125(12): 1233-1242. DOI: 10.1007/s11547-020-01214-x.
[36]
CALIMERI T, STEFFANONI S, GAGLIARDI F, et al. How we treat primary central nervous system lymphoma[J/OL]. ESMO Open, 2021, 6(4): 100213 [2024-10-21]. https://pubmed.ncbi.nlm.nih.gov/34271311/. DOI: 10.1016/j.esmoop.2021.100213.
[37]
TOFTS P S, KERMODE A G. Measurement of the blood-brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts[J]. Magn Reson Med, 1991, 17(2): 357-367. DOI: 10.1002/mrm.1910170208.
[38]
TOFTS P S. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging[J]. J Magn Reson Imaging, 1997, 7(1): 91-101. DOI: 10.1002/jmri.1880070113.
[39]
SHUKLA-DAVE A, LEE N Y, JANSEN J F A, et al. Dynamic contrast-enhanced magnetic resonance imaging as a predictor of outcome in head-and-neck squamous cell carcinoma patients with nodal metastases[J]. Int J Radiat Oncol Biol Phys, 2012, 82(5): 1837-1844. DOI: 10.1016/j.ijrobp.2011.03.006.
[40]
BOZDAĞ M, ER A L, ÇINKOOĞLU A. Histogram analysis of ADC maps for differentiating brain metastases from different histological types of lung cancers[J]. Can Assoc Radiol J, 2021, 72(2): 271-278. DOI: 10.1177/0846537120933837.
[41]
LI C M, YU L, JIANG Y W, et al. The histogram analysis of intravoxel incoherent motion-kurtosis model in the diagnosis and grading of prostate cancer - A preliminary study[J/OL]. Front Oncol, 2021, 11: 604428 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC8579734/. DOI: 10.3389/fonc.2021.604428.
[42]
VAUPEL P. Hypoxia and aggressive tumor phenotype: implications for therapy and prognosis[J]. 2008, 13(Suppl 3): 21-26. DOI: 10.1634/theoncologist.13-S3-21.
[43]
陈世雄, 陈航榕. 乏氧响应型纳米药物的设计、合成及其抗肿瘤作用研究[J]. 药学进展, 2021, 45(4): 254-262. DOI: 1001-5094(2021)04-0254-09.
CHEN S X, CHEN H R. Design, synthesis and anti-tumor activity of hypoxia-responsive nanomedicines[J]. Prog Pharm Sci, 2021, 45(4): 254-262. DOI: 1001-5094(2021)04-0254-09.
[44]
BRISTOW R G, BERLIN A, DAL PRA A. An arranged marriage for precision medicine: hypoxia and genomic assays in localized prostate cancer radiotherapy[J/OL]. Br J Radiol, 2014, 87(1035): 20130753 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC4064607/. DOI: 10.1259/bjr.20130753.
[45]
OTO A, YANG C, KAYHAN A, et al. Diffusion-weighted and dynamic contrast-enhanced MRI of prostate cancer: correlation of quantitative MR parameters with Gleason score and tumor angiogenesis[J]. AJR Am J Roentgenol, 2011, 197(6): 1382-1390. DOI: 10.2214/AJR.11.6861.
[46]
DONALDSON S B, WEST C M L, DAVIDSON S E, et al. A comparison of tracer kinetic models for T1-weighted dynamic contrast-enhanced MRI: application in carcinoma of the cervix[J]. Magn Reson Med, 2010, 63(3): 691-700. DOI: 10.1002/mrm.22217.
[47]
FENG D C, XIONG Q, WEI Q, et al. Cellular landscape of tumour microenvironment in prostate cancer[J]. Immunology, 2023, 168(2): 199-202. DOI: 10.1111/imm.13456.
[48]
FANG B W, LU Y, LI X M, et al. Targeting the tumor microenvironment, a new therapeutic approach for prostate cancer[J/OL]. Prostate Cancer Prostatic Dis, 2024 [2024-10-21]. https://pubmed.ncbi.nlm.nih.gov/38565910/. DOI: 10.1038/s41391-024-00825-z.
[49]
MUSTAFA M, RASS H ABU, YAHYA M, et al. Primary metastatic prostate cancer between prognosis or adequate/proper medical therapy[J/OL]. World J Surg Oncol, 2021, 19(1): 5 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC7783967/. DOI: 10.1186/s12957-020-02111-3.
[50]
刘靖, 陈珂, 刘菊, 等. 超声造影早期预测前列腺癌经内分泌治疗后进展为去势抵抗性前列腺癌的价值[J]. 放射学实践, 2021, 36(9): 1127-1132. DOI: 10.13609/j.cnki.1000-0313.2021.09.010.
LIU J, CHEN K, LIU J, et al. Value of contrast-enhanced ultrasound in early predicting the progression of castration-resistant prostate cancer after endocrine therapy in patients with prostate cancer[J]. Radiol Pract, 2021, 36(9): 1127-1132. DOI: 10.13609/j.cnki.1000-0313.2021.09.010.
[51]
GASINSKA A, JASZCZYNSKI J, RYCHLIK U, et al. Prognostic significance of serum PSA level and telomerase, VEGF and GLUT-1 protein expression for the biochemical recurrence in prostate cancer patients after radical prostatectomy[J]. Pathol Oncol Res, 2020, 26(2): 1049-1056. DOI: 10.1007/s12253-019-00659-4.
[52]
YANG D D, MAHAL B A, MURALIDHAR V, et al. Androgen deprivation therapy and overall survival for gleason 8 versus gleason 9-10 prostate cancer[J]. Eur Urol, 2019, 75(1): 35-41. DOI: 10.1016/j.eururo.2018.08.033.
[53]
PALLARES J, ROJO F, IRIARTE J, et al. Study of microvessel density and the expression of the angiogenic factors VEGF, bFGF and the receptors Flt-1 and FLK-1 in benign, premalignant and malignant prostate tissues[J]. Histol Histopathol, 2006, 21(8): 857-865. DOI: 10.14670/HH-21.857.
[54]
EGEVAD L, MICOLI C, DELAHUNT B, et al. Prognosis of Gleason score 8 prostatic adenocarcinoma in needle biopsies: a nationwide population-based study[J]. Virchows Arch, 2024, 484(6): 995-1003. DOI: 10.1007/s00428-024-03810-y.
[55]
EGEVAD L, MICOLI C, SAMARATUNGA H, et al. Prognosis of gleason score 9-10 prostatic adenocarcinoma in needle biopsies: a nationwide population-based study[J]. Eur Urol Oncol, 2024, 7(2): 213-221. DOI: 10.1016/j.euo.2023.11.002.
[56]
JANG W S, YOON C Y, KIM K H, et al. Prognostic significance of vas deferens invasion after radical prostatectomy in patients with pathological stage T3b prostate cancer[J]. Ann Surg Oncol, 2017, 24(4): 1143-1149. DOI: 10.1245/s10434-016-5653-x.
[57]
HEERING M, BERG K D, BRASSO K, et al. Radical prostatectomy in Denmark: Survival analysis and temporal trends in clinicopathological parameters with up to 20 years of follow-up[J]. Surg Oncol, 2017, 26(1): 21-27. DOI: 10.1016/j.suronc.2016.12.002.
[58]
YAXLEY W, DELAHUNT B, YAXLEY J, et al. Prognostic significance of percentage Gleason grade 5 prostatic adenocarcinoma in needle biopsies from patients treated by radical prostatectomy[J]. Pathology, 2022, 54(6): 694-699. DOI: 10.1016/j.pathol.2022.02.009.
[59]
WENZEL M, WÜRNSCHIMMEL C, CHIERIGO F, et al. Pattern of biopsy gleason grade group 5 (4 + 5 vs 5 + 4 vs 5 + 5) predicts survival after radical prostatectomy or external beam radiation therapy[J]. Eur Urol Focus, 2022, 8(3): 710-717. DOI: 10.1016/j.euf.2021.04.011.
[60]
MIAN B M, FEUSTEL P J, AZIZ A, et al. Complications following transrectal and transperineal prostate biopsy: results of the ProBE-PC randomized clinical trial[J]. J Urol, 2024, 211(2): 205-213. DOI: 10.1097/JU.0000000000003788.
[61]
ALHAMDANI Z, POPPENBEEK S, BOLTON D, et al. Do alpha blockers reduce the risk of urinary retention post-transperineal prostate biopsy? A systematic narrative review[J/OL]. World J Urol, 2024, 42(1): 332 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11101363/. DOI: 10.1007/s00345-024-05001-5.
[62]
HU C H, QIAO X M, HUANG R P, et al. Development and validation of a multimodality model based on whole-slide imaging and biparametric MRI for predicting postoperative biochemical recurrence in prostate cancer[J/OL]. Radiol Imaging Cancer, 2024, 6(3): e230143 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC11148661/. DOI: 10.1148/rycan.230143.
[63]
VIS A N, MEIJER D, ROBERTS M J, et al. Development and external validation of a novel nomogram to predict the probability of pelvic lymph-node metastases in prostate cancer patients using magnetic resonance imaging and molecular imaging with prostate-specific membrane antigen positron emission tomography[J]. Eur Urol Oncol, 2023, 6(6): 553-563. DOI: 10.1016/j.euo.2023.03.010.
[64]
ZAMBOGLOU C, PEEKEN J C, JANBAIN A, et al. Development and validation of a multi-institutional nomogram of outcomes for PSMA-PET-based salvage radiotherapy for recurrent prostate cancer[J/OL]. JAMA Netw Open, 2023, 6(5): e2314748 [2024-10-21]. https://pmc.ncbi.nlm.nih.gov/articles/PMC10208140/. DOI: 10.1001/jamanetworkopen.2023.14748.

上一篇 基于MRI影像组学预测结直肠癌KRAS基因突变的系统回顾和Meta分析
下一篇 ADCmean联合PSAD对PI-RADS≥3分临床显著性前列腺癌的预测价值
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2