分享:
分享到微信朋友圈
X
临床研究
基于IVIM-DWI定量参数的影像组学预测鼻咽癌短期疗效
戴干棉 武文渊 傅丽莉 李天生 羊倩羽 黄薇园 郭义昊 陈峰

DAI G M, WU W Y, FU L L, et al. Radiomics analysis based on IVIM-DWI quantitative parameters to predict the short-term therapeutic effect of nasopharyngeal carcinoma[J]. Chin J Magn Reson Imaging, 2023, 14(9): 56-62, 69.引用本文:戴干棉, 武文渊, 傅丽莉, 等. 基于IVIM-DWI定量参数的影像组学预测鼻咽癌短期疗效[J]. 磁共振成像, 2023, 14(9): 56-62, 69. DOI:10.12015/issn.1674-8034.2023.09.010.


[摘要] 目的 基于治疗前鼻咽癌体素内不相干运动扩散加权成像(intravoxel incoherent motion diffusion-weighted imaging, IVIM-DWI)影像组学定量特征,建立用于预测鼻咽癌短期疗效的预测模型。材料与方法 回顾性收集2019年1月至2021年8月期间于海南省人民医院放疗科接受治疗的首程病理确诊鼻咽癌患者80例。在治疗前均行头颈部MRI平扫+增强检查及11个b值(区间0~800 s/mm2)IVIM-DWI检查,在接受以放疗为主的综合治疗后每3个月进行头颈部常规MRI随访,依据实体肿瘤的治疗反应评价标准1.1版在治疗后6个月的MRI随访图像上行疗效评价,将患者分为完全缓解组(n=62)及非完全缓解组(n=18)。IVIM-DWI经双指数模型后处理计算得到真实扩散系数(true diffusion coefficient, D)、灌注相关扩散系数(perfusion related diffusion coefficient, D*)和灌注分数(perfusion fraction, f)的定量参数图像。使用ITK-SNAP软件在IVIM-DWI的S0图像上逐层勾画病灶的感兴趣区(region of interest, ROI),鼻咽癌常规及增强MRI图像作为定位参照。应用3D slicer软件分别在D、D*和f定量参数图像上相应的ROI区域提取影像组学特征,包括直方图特征、纹理特征以及基于形态学的组学特征。使用最小绝对收缩和选择算子算法筛选出与疗效高度相关的影像组学特征,采用逻辑回归方法分别构建基于D、D*、f和联合参数的影像组学预测模型,预测性能采用ROC曲线、曲线下面积(area under the curve, AUC)及校正曲线评估,并使用决策曲线分析(decision curve analysis, DCA)评价预测模型的临床实用性。10次十折交叉验证被使用于模型内部验证,计算平均敏感度及特异度。结果 共计851个影像组学特征被提取,经过特征筛选后,筛选出2个D值特征,构建的影像组学模型的敏感度为60.0%,特异度为79.6%,AUC值为0.734;筛选出2个f值特征,构建的影像组学模型敏感度为66.1%,特异度为76.3%,AUC值为0.747;筛选出1个D*值特征,构建的影像组学模型敏感度为76.1%,特异度为75.9%,AUC值为0.726;联合以上3种参数共5个影像组学特征,构建的联合影像组学模型敏感度为81.7%,特异度为80.6%,AUC值为0.827。校正曲线显示各模型均具有良好的拟合优度,DCA显示4种模型均具有良好的临床效益,而IVIM联合模型的临床效益最高。结论 基于IVIM-DWI参数建立的影像组学模型能够在治疗前较好地预测鼻咽癌患者的治疗反应性。其中,效能最高的是IVIM-DWI联合参数模型,该模型可以为患者的临床诊疗决策提供帮助。
[Abstract] Objective To establish a predictive model based on quantitative characteristics of intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) radiomics to predict short-term treatment efficacy of nasopharyngeal carcinoma before treatment.Materials and Methods A retrospective study was conducted to collect 80 patients with nasopharyngeal carcinoma diagnosed pathologically at the first visit who were treated in the Radiotherapy Department of Hainan Provincial People's Hospital from January 2019 to August 2021. Before treatment, all subjects underwent MRI plain scan + enhanced examination and 11 b-value (interval 0-800 s/mm2) IVIM-DWI examination. After receiving comprehensive treatment based on radiotherapy, routine MRI follow-up of head and neck was conducted every 3 months. Use MRI follow-up images taken 6 months after the end of treatment for efficacy evaluation. According to respond evaluation criteriain solid tumors, version 1.1 standard, the patients were divided into complete remission group (n=62) and incomplete remission group (n=18). The real diffusion coefficient (D), perfusion related diffusion coefficient (D*) and perfusion fraction (f) were obtained by post-processing IVIM-DWI with a double exponential model. Itk-snap was used to delineate the region of interest (ROI) of the lesion layer by layer on the S0 image of IVIM-DWI, and the conventional and enhanced MRI images of nasopharyngeal carcinoma were used as the positioning reference. 3D slicer software was used to extract radiomics features, including histogram features, texture features and morphology features, from the corresponding ROI regions of D, D* and f quantitative parameter images. The least absolute shrinkage and selection operator algorithm was used to screen out the radiomics features that were highly correlated with the treatment effect. Logistic regression was used to construct radiomics prediction models based on D, D*, f, and joint parameters, and predictive performance was evaluated using ROC curves, area under the curve (AUC), and calibration curves. Decision curve analysis (DCA) was used to evaluate the clinical utility of the prediction models. A 10-fold cross-validation was used for internal model validation, and the average sensitivity and specificity were calculated.Results A total of 851 radiomics features were extracted, and after feature selection, two D-value features were selected to construct a radiomics model with a sensitivity of 60.0%, specificity of 79.6%, and an AUC value of 0.734. Two f-value features were selected to construct a radiomics model, with a sensitivity of 66.1%, specificity of 76.3%, and an AUC value of 0.747. One D*-value feature was selected to construct a radiomics model, with a sensitivity of 76.1%, specificity of 75.9%, and an AUC value of 0.726. The sensitivity, specificity and AUC of the radiomics model based on the three types of IVIM-DWI radiomics features were 81.7%, 80.6% and 0.827 respectively. The calibration curves showed good goodness-of-fit for all models, and the DCA demonstrated good clinical utility for all four models, with the IVIM joint model showing the highest clinical benefit.Conclusions The radiomics model based on IVIM-DWI parameters can predict the therapeutic response of nasopharyngeal carcinoma patients before treatment. Among them, the most effective model is the IVIM-DWI combined parameter model, which can assist in clinical decision-making for patients.
[关键词] 鼻咽癌;治疗疗效;体素内不相干运动成像;影像组学;磁共振成像
[Keywords] nasopharyngeal carcinoma;therapeutic effect;intravoxel incoherent motion imaging;radiomics;magnetic resonance imaging

戴干棉    武文渊    傅丽莉    李天生    羊倩羽    黄薇园    郭义昊    陈峰 *  

海南医学院附属海南医院(海南省人民医院)放射科,海口 570100

通信作者:陈峰,E-mail:fenger0802@163.com

作者贡献声明:陈峰负责设计本研究的方案,对稿件重要内容进行了修改;戴干棉起草和撰写稿件,获取、分析本研究的数据;武文渊、傅丽莉、李天生、羊倩羽、郭义昊获取、分析本研究的数据;黄薇园对稿件重要内容进行了修改;陈峰、黄薇园获得了国家自然科学基金项目资助。全体作者都同意发表后的修改稿,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: 国家自然科学基金 81971602,82260339
收稿日期:2023-02-02
接受日期:2023-09-06
中图分类号:R445.2  R739.62 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2023.09.010
引用本文:戴干棉, 武文渊, 傅丽莉, 等. 基于IVIM-DWI定量参数的影像组学预测鼻咽癌短期疗效[J]. 磁共振成像, 2023, 14(9): 56-62, 69. DOI:10.12015/issn.1674-8034.2023.09.010.

0 前言

       鼻咽癌是一种多基因调控疾病,在中国的头颈部恶性肿瘤中发病率为首位[1, 2, 3],且具有明显的区域性高发特性,以广东、海南发病率最高。较大多数其他癌症,鼻咽癌的预后通常要好,五年生存率可达到80%[4]。根治性放疗是其主要治疗手段,但仍有近15%的患者在治疗结束后短期肿瘤局部复发导致治疗失败,预后不佳[5, 6],因此提高鼻咽癌治疗疗效至关重要。肿瘤微环境与肿瘤治疗效果密切相关[7, 8, 9],fMRI如动态对比增强MRI(dynamic contrast-enhanced MRI, DCE-MRI)与扩散加权成像(diffusion weighted imaging, DWI)被用于评估鼻咽癌治疗敏感性[10, 11, 12],DCE-MRI可提供肿瘤组织灌注相关信息,但容易受到患者血压和心率的影响。体素内不相干运动DWI(intravoxel incoherent motion DWI, IVIM-DWI)可将血流微循环灌注效应及组织水分子真实扩散进行量化分析[13],然而在使用传统方法分析IVIM-DWI定量参数时,预测性能低于传统TNM分期[14],且对鼻咽癌患者的预后预测相关性较弱,稳定性低[15]。影像组学方法为解决此问题提供了新思路,基于IVIM-DWI的影像组学分析在宫颈癌患者中已经被证明具有可行性[16],而将其应用于鼻咽癌的研究目前尚为空白。因此本研究通过影像组学建模的方法回顾性分析80例鼻咽癌患者的IVIM-DWI定量参数组学特征,旨在开发出用于临床预测鼻咽癌患者的短期放化疗疗效的生物标志物,为提高鼻咽癌治疗疗效提供新的思路和方法。

1 材料与方法

1.1 研究对象

       回顾性分析2019年1月至2021年9月期间于海南省人民医院经鼻咽镜检查及病理确诊的首程鼻咽癌并在放疗科接受以根治性放疗为基础综合性治疗的患者资料,所有患者均在治疗前行头颈部常规MRI检查及IVIM-DWI检查,并在治疗结束后一年内每3个月进行鼻咽部MRI随访。本研究遵守《赫尔辛基宣言》,经海南省人民医院伦理委员会批准,免除受试者知情同意,批准文号:HYLL-2022-368。

       纳入标准:(1)首次确诊为鼻咽癌的患者且经活检病理以及鼻咽镜证实;(2)既往无任何肿瘤病史;(3)既往未行任何抗肿瘤治疗;(4)Karnofsky评分>80,得分越高表明患者更有可能接受彻底的治疗。排除标准:(1)头颈部曾接受过放疗或者化疗;(2)MRI图像质量差、伪影及干扰项目多,无法满足诊断或评价需求;(3)拒绝完成随访计划;(4)患者有治疗禁忌证或检查禁忌证。

1.2 扫描设备及方法

       入组患者在治疗前应用德国Siemens Magnetom Skyra 3.0 T MRI设备及头颈Head/Neck 20通道联合线圈完成以下MRI扫描序列:

       (1)质子加权鼻咽部轴位快速自旋回波序列:视野180 mm×180 mm,层厚4 mm,层间距1.2 mm,回波时间9 ms,重复时间625 ms,层数20层;(2)鼻咽部轴位T1加权自旋回波序列:视野180 mm×180 mm,层厚4 mm,层间距1.2 mm,回波时间30 ms,重复时间4070 ms,层数20层;(3)鼻咽部轴位T1加权增强扫描序列:视野180 mm×180 mm,层厚4 mm,层间距1.2 mm,回波时间30 ms,重复时间4070 ms,层数20层;(4)鼻咽部IVIM-DWI序列:视野240 mm×240 mm,层厚4.4 mm,回波时间75 ms,重复时间4170 ms,层数25层,体素2.6 mm×2.6 mm×2.6 mm,采用0、10、20、30、40、50、100、150、300、600、800 s/mm2共11个b值。

1.3 治疗方案

       放疗:放疗均采用三维调强适形放疗,放疗靶区经体格检查、影像学检查确定。肿瘤部位及侵犯范围被定义为肿瘤靶区(gross tumor volume, GTV),包括咽后转移淋巴结(GTVrpn)、颈部转移淋巴结(GTVnd)、鼻咽癌大体肿瘤体积(GTVnx)。放疗的总持续时间为43~54天,总剂量为68.2~72.6 Gy,分31~33次计划给量;1次/天,5天/周。

       化疗:同步化疗采用药物为单药顺铂或奈达铂。诱导化疗采用以铂类药物(奈达铂/顺铂)为基础并联合其他化疗药物的化疗方案(吉西他滨、紫杉醇、多西他赛、氟尿嘧啶)。每3周1次进行诱导化疗及同步化疗。

       其他治疗:部分患者于放疗期间每周使用一次尼妥珠单抗。

1.4 疗效评价

       采用肿瘤消退率(tumor response, TR)评估鼻咽癌治疗疗效,依据实体肿瘤的治疗反应评价标准(respond evaluation criteriain solid tumors, RECIST)1.1版,TR被定义为治疗前肿瘤最大径与放疗结束后6个月MRI随访图像上肿瘤最大径之差占治疗前肿瘤最大径的百分比,由影像科医师在鼻咽癌患者治疗前及治疗后6个月鼻咽部轴位T1增强、质子加权扫描图像上评估。在不同序列图像的肿瘤最大层面上勾画最大径线,分别计算并最终取TR的平均值。患者分组亦参照RECIST 1.1标准,根据随访TR情况将患者分为4类:完全缓解(complete relief, CR),指TR=100%;疾病进展(progression disease, PD),指病灶较治疗前增加20%;部分缓解(partial relief, PR),指TR≥30%;疾病稳定(stable disease, SD),指病灶缓解情况不符合PR/PD。最终将患者PR、PD和SD患者分为非完全缓解组(非CR组)以及完全缓解组(CR组)。

1.5 数据处理及图像特征提取方法

       根据双指数模型,使用MATLAB软件(version R2021b)采用分两步拟合的方式,计算出患者IVIM-DWI图像真实扩散系数(true diffusion coefficient, D)、灌注相关扩散系数(perfusion related diffusion coefficient, D*)和灌注分数(perfusion fraction, f)的定量参数图像[13]。由两名有2年MRI诊断经验的住院医师采用双盲法并使用ITK-SNAP(version 3.8)软件,参照常规MRI平扫及增强扫描图像,在IVIM-DWI的S0图像上沿肿瘤边缘手动逐次勾画感兴趣区(region of interest, ROI),包括整个鼻咽癌原发肿瘤病灶区域及肿瘤坏死区,尽量避开邻近伪影、骨质、软组织和空气,通过3D slicer软件分别提取出ROI区域相应的D、D*、f值参数图像中的特征以及这些参数图经过高斯-拉普拉斯滤波后图像中的特征。提取的特征包括一阶特征、灰度依赖矩阵(gray level dependence matrix, GLDM)、灰度大小区域矩阵(gray level size zonematrix, GLSZM)、灰度共生矩阵(gray level co-occurrence matrix, GLCM)、相邻灰度色调差异矩阵、灰度游程长度矩阵(gray level run length matrix, GLRLM),肿瘤原发灶的三维大小和形状等形态学特征,共7大类。提取的特征观察者间一致性采用组内相关系数(intra-class correlation coefficient, ICC)来评估,ICC值高于0.75的特征用于后续分析。

       采用最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)回归去除冗余及不相关特征,筛选出最佳的特征,并基于筛选出的特征使用logistic regression方法建立影像组学预测模型,ROC曲线及曲线下面积(area under the curve, AUC)被用来评价预测模型性能。为了验证模型结果并非过拟合得到,采用10次十折交叉验证对模型进行内部验证,计算10次重复验证的平均敏感度及特异度。

1.6 统计学分析

       全部的人口学和临床资料采用IBM SPSS(windows 22.0)统计学软件进行分析,非正态性数据使用Mann-Whitney U检验及卡方检验进行相应的统计分析,正态性数据使用独立样本t检验分析,数据分布的正态性使用Kolmogorov-Smirnov检验评估,采用DeLong检验来比较各参数模型的预测效能,P<0.05时差异有统计学意义。

2 结果

2.1 人口学及临床资料

       共收集99例鼻咽癌患者,排除无随访资料患者12例,MRI图像伪影过大患者7例,最终纳入80例患者,根据上述疗效评估标准,在放疗完成后6个月左右评估治疗疗效,将62例患者归入CR组,18例归入非CR组,图12显示了具有代表性的非CR组及CR组的鼻咽癌患者IVIM-DWI图像,IVIM-DWI原始图S0显示,治疗前两组患者鼻咽癌病灶均呈高信号,在D、D*、f参数伪彩图上均以蓝色为主,CR组鼻咽癌病灶的D、f值伪彩图颜色较非CR组更深,D*值伪彩图的绿色部分CR组较非CR组丰富。

       表1显示了患者的一般人口学资料及临床资料。非CR组患者的T分期、M分期、临床分期及年龄较CR组高,差异具有统计学意义(P<0.05)。

图1  男,33岁,非CR鼻咽癌患者。1A~1C分别为D值、D*值、f值参数伪彩图;1D为PD序列;1E为IVIM S0图像;1F为IVIM S0图像上病灶ROI勾画示意图。IVIM原始图S0显示治疗前非CR鼻咽癌病灶以高信号为主,累及左侧咽旁间隙、左侧腭帆张提肌,边界不清;f参数值伪彩图上病灶呈蓝绿信号混杂,以蓝色为主;病灶在D参数值伪彩图上为淡蓝色;D*值伪彩图上肿块呈蓝、绿、红混杂色,以深蓝色为主。CR:完全缓解;D:真实扩散系数;D*:灌注相关扩散系数;f:灌注分数;PD:疾病进展;IVIM:体素内不相干运动;ROI:感兴趣区。
Fig. 1  The text describes an image set of a 33-year-old male patient with non-CR nasopharyngeal carcinoma. 1A-1C: Pseudo-color maps of D value, D* value, and f value parameters respectively. 1D: A PD sequence image. 1E: An IVIM S0 image showing the pre-treatment non-CR lesion with predominantly high signal intensity, involving the left parapharyngeal space and left levator veli palatini muscle, with indistinct borders. 1F: An ROI on the IVIM S0 image showing the lesion, which appears as a mixture of blue and green signals on the pseudo-color map of f value parameter, with blue being the dominant color. The lesion appears as light blue on the pseudo-color map of D value parameter, and a mixture of deep blue, green, and red n the pseudo-color map of D* value parameter, with deep blue being the dominant color. CR: complete remission; D: true diffusion coefficient; D*: perfusion related diffusion coefficient; f: perfusion fraction; PD: progression disease; IVIM: intravoxel incoherent motion; ROI: region of interest.
图2  男,30岁,CR鼻咽癌患者。2A~2C分别为D值、D*值、f值参数伪彩图;2D为PD序列;2E为IVIM S0图像;2F为IVIM S0图像上病灶ROI勾画示意图。放化疗前CR组鼻咽癌病灶IVIM原始图S0在边缘以高信号为主,信号强度在中央逐渐减低,病灶向前累及左侧后鼻孔区;f参数值伪彩图上病灶呈深蓝色、绿色混杂,以深蓝色为主;D参数值伪彩图上病灶为深蓝色;D*参数值伪彩图上肿块呈蓝、绿、红混杂色,以深蓝色为主。CR:完全缓解;D:真实扩散系数;D*:灌注相关扩散系数;f:灌注分数;PD:疾病进展;IVIM:体素内不相干运动;ROI:感兴趣区。
Fig. 2  The text describes an image set of a 30-year-old male patient with CR nasopharyngeal carcinoma. 2A-2C: Pseudo-color maps of D value, D* value, and f value parameters respectively. 2D: A PD sequence image. 2E: An IVIM S0 image showing the pre-radiotherapy CR lesion with predominantly high signal intensity at the edges, decreasing in intensity towards the center, and involving the left posterior nasal cavity. 2F: An ROI on the IVIM S0 image showing the lesion, which appears as a mixture of deep blue and green signals on the pseudo-color map of f value parameter, with deep blue being the dominant color. The lesion appears as deep blue on the pseudo-color map of D value parameter, and a mixture of blue, green, and red on the pseudo-color map of D* value parameter, with deep blue being the dominant color. CR: complete remission; D: true diffusion coefficient; D*: perfusion related diffusion coefficient; f: perfusion fraction; PD: progression disease; IVIM: intravoxel incoherent motion; ROI: region of interest.
表1  鼻咽癌患者人口学资料及临床资料
Tab. 1  Demographic and clinical data of patients with nasopharyngeal carcinoma

2.2 影像组学特征提取及筛选

       按照上述方法、步骤及软件进行影像组学特征提取及筛选(图3)。共计851个影像组学特征被提取,其中一阶特征为162个、灰度依赖矩阵为126个、灰度大小区域矩阵为144个、灰度共生矩阵为216个、相邻灰度色调差异矩阵为45个、灰度游程长度矩阵为144个,肿瘤原发灶的三维大小和形状等形态学特征为14个。去除了ICC值<0.75的187个特征,共664个特征进入后续分析。采用LASSO筛选方法最终筛选出2个D值特征;1个D*值特征;2个f值特征(图4)。

图3  影像组学分析流程图。IVIM:体素内不相干运动;ROI:感兴趣区;LASSO:最小收缩算子算法;LR:逻辑回归。
Fig. 3  Flow chart of imaging radiomics analysis. IVIM: intravoxel incoherent motion; ROI: region of interest. LR: logistic regression. LASSO: least absolute shrinkage and selection operator algorithm.
图4  鼻咽癌患者CR组及非CR组IVIM定量参数影像组学特征。CR:完全缓解;IVIM:体素内不相干运动。
Fig. 4  Imaging radiomics features of IVIM quantitative parameters in nasopharyngeal carcinoma patients with CR and non-CR groups. CR: complete remission; IVIM: intravoxel incoherent motion.

2.3 影像组学模型建立及验证

       上述所选出的特征被使用于预测模型的建立,10次十折交叉验证方法对数据进行重复验证,使用ROC曲线对模型预测性能进行评估,结果D值模型的敏感度为60.0%,特异度为79.6%,AUC值为0.734;f值模型的敏感度为66.1%,特异度为76.3%,AUC值为0.747;D*值模型的敏感度为76.1%,特异度为75.9%,AUC值为0.726;联合以上3种参数共5个影像组学特征,最终构建的IVIM影像组学联合参数模型的敏感度为81.7%,特异度为80.6%,AUC值为0.827(表2)校准曲线显示D、D*、f值三种模型的预测值接近45度标准线的趋势类似,表明这三类模型的一致性相似,而IVIM联合模型的预测值接近45度标准线的趋势最明显(图5)。临床决策曲线分析还表明4种模型具有良好的临床效益,而IVIM联合模型的临床效益最高(图6)。

       DeLong检验显示IVIM联合参数模型的预测效能较D值模型、D*值模型、f值模型高,差异具有统计学意义(P<0.05),而D值、D*值、f值模型相互之间的预测效能差异无统计学意义(表3)。IVIM联合参数模型的计算公式见式(1)

       结果表明,基于IVIM-DWI定量参数D、D*、f值及联合三种参数影像学特征构建的logistic回归模型在预测鼻咽癌患者的短期治疗疗效上具有良好的预测效能,而联合三种参数影像学特征构建的logistic回归模型的预测效能最高(图7)。

       +0.2282×(D-GLSZM-Zone Entropy)

       +1.0682×(D-GLDM-Dependence Variance)

       -0.9486×(f-GLSZM-Small Area Low Gray Level Emphasis)

图5  D、D*、f及IVIM联合参数模型校正曲线。5A、5B、5C、5D图分别为D、D*、f及IVIM联合参数模型的校正曲线,平均绝对误差分别为0.042、0.042、0.043、0.039,表明各模型的拟合优度佳。D:真实扩散系数;D*:灌注相关扩散系数;f:灌注分数;IVIM:体素内不相干运动。
Fig. 5  D, D*, f, and IVIM joint parameter model correction curves. 5A, 5B, 5C, 5D show the calibration curves of the D value, D* value, f value, and IVIM combined parameter models, with mean absolute errors of 0.042, 0.042, 0.043, and 0.039, respectively, indicating good fitting performances of each model. D: true diffusion coefficient; D*: perfusion related diffusion coefficient; f: perfusion fraction; IVIM: intravoxel incoherent motion.
图6  D、D*、f及IVIM联合参数模型临床决策分析曲线图。LR_D、LR_f、LR_D*、LR_IVIM分别表示4种影像组学模型。LR:逻辑回归;D:真实扩散系数;D*:灌注相关扩散系数;f:灌注分数;IVIM:体素内不相干运动。
Fig. 6  The clinical decision analysis curves for the D value, D* value, f value, and IVIM combined parameters model. LR_D, LR_f, LR_D*, and LR_IVIM represent four radiomics models. LR: logistic regression; D: true diffusion coefficient; D*: perfusion related diffusion coefficient; f: perfusion fraction; IVIM: intravoxel incoherent motion.
图7  D、D*、f及IVIM联合参数模型ROC曲线,用于预测治疗反应,其中IVIM联合参数模型曲线下面积最大,为0.827。LR_D、LR_f、LR_D*、LR_IVIM分别表示4种影像组学模型。LR:逻辑回归;D:真实扩散系数;D*:灌注相关扩散系数;f:灌注分数;IVIM:体素内不相干运动。
Fig. 7  The ROC curves of the D value, D* value, f value, and IVIM combined parameter models for predicting treatment response, with the IVIM combined parameter model having the largest area under the curve of 0.827. LR_D, LR_f, LR_D*, and LR_IVIM represent four radiomics models. LR: logistic regression; D: true diffusion coefficient; D*: perfusion related diffusion coefficient; f: perfusion fraction; IVIM: intravoxel incoherent motion.
表2  鼻咽癌患者IVIM定量参数特征模型的预测性能
Tab. 2  Predictive performance of quantitative parameter models of IVIM in nasopharyngeal carcinoma patients
表3  鼻咽癌患者IVIM定量参数特征模型的预测性能统计学对比
Tab. 3  Comparative statistical analysis of predictive performance for quantitative parameter models of IVIM in nasopharyngeal carcinoma patients

3 讨论

       本研究基于鼻咽癌患者fMRI IVIM-DWI的定量参数图像,提取影像组学特征并采用logistic回归构建预测模型,用于预测鼻咽癌短期治疗疗效,使用ROC曲线对模型进行评估。结果显示各定量参数(D、D*、f)预测模型具有较好的预测效能,而IVIM-DWI联合参数的模型的预测效能高于单一参数模型。

3.1 影像组学分析预测鼻咽癌疗效的价值

       影像组学作为一种重要的分析方法,可以深入表征肿瘤的异质性,反映肿瘤的细胞微环境、增殖、血管生成、缺氧和坏死[17],这些因素与较差的治疗反应和肿瘤预后不良明显相关[18, 19]。因此,基于MRI的放射组学方法可能有助于通过表征肿瘤异质性来预测肿瘤治疗反应[20, 21]。本研究结果表明,LASSO选择的用于模型构建的放射组学特征均为纹理特征,纹理特征在预测鼻咽癌患者的治疗反应方面优于一阶特征或体积/形状特征,与既往研究一致[11]

       BORDRON等[22]使用影像组学模型预测直肠癌放化疗疗效时,联合多参数模型的预测效能高于单一参数模型,与本研究类似,表明多参数模型较单参数模型更能全面反映肿瘤内部异质性。WANG等[23]基于常规T1增强建立的影像组学预测模型效能AUC达0.852(95%置信区间:0.847~0.857),稍高于本研究模型效能,这可能是因为fMRI IVIM-DWI分辨率低,在鼻咽部病灶显示效果差,且受颅底磁敏感伪影、吞咽运动干扰较常规MRI明显。而IVIM-DWI能提供肿瘤组织灌注信息,更全面地反映肿瘤异质性,优势于常规MRI。效能更佳、更具有稳定性的预测模型有望通过将常规MRI及功能MRI图像融合分析获得。

3.2 IVIM-MRI预测鼻咽癌疗效的价值

       根据IVIM-DWI理论,D值属于扩散相关系数,反映组织内水分子的纯扩散,与细胞密度呈反相关,与组织内的细胞坏死、囊变呈正相关。坏死区域常伴有缺氧和组织酸中毒,这可引起肿瘤对放化疗的抵抗。既往研究表明,D值高的肿瘤短期预后较差[10, 11, 24-25],而治疗前低D值表明对放化疗敏感[10, 25],表明扩散相关系数D值是预测鼻咽癌预后的生物标记物,与本研究结果类似。

       D*值及f值作为灌注相关系数,反映鼻咽癌组织内部血液灌注状态,肿瘤内毛细血管网分布与肿瘤灌注参数正相关。既往研究表明,分级较高的肿瘤灌注参数高于低级别肿瘤[26, 27, 28],而肿瘤灌注参数的增加与肿瘤分级[29]和不良预后[30, 31]密切相关,这可能是由于越高,肿瘤实体的级别越高,瘤体内毛细血管网越丰富,从而易肿瘤早期转移引起不良预后。而在鼻咽癌的研究中,采用常规方法分析IVIM-DWI灌注相关系数时,部分研究的结论相悖。鼻咽癌相关的既往研究[32, 33, 34]表明,D*值及f值高的患者预后及疗效更佳;而QAMAR等[14]的研究显示f值与鼻咽癌患者的短期及长期预后均不相关,而较低的D*值与较差的预后相关。因此IVIM-DWI灌注相关系数的稳定性和可重复性较差。而影像组学分析方法在研究IVIM-MRI灌注系数时可以改善预测稳定性,在ZHANG等[35]的研究中,使用灌注系数D*值及f值构建的影像组学模型在预测局部晚期宫颈癌疗效中具有良好及稳定的预测效能,与本研究结果类似。

3.3 局限性

       本研究尚存在一些不足:首先,本研究所选择样本集中TNM分期和放化疗治疗的多样性未被考虑,这可能会影响结果,因此有必要对更大的患者群体进行进一步研究,预测按TNM分期或治疗方案分层的不同患者队列的治疗反应;其次,本研究使用单个数据集进行内部验证,这可能会导致结果的误差,若能引进多个数据集或多中心数据进行外部验证,可能会进一步提高模型的稳定性;第三,本研究仅基于患者的IVIM-DWI影像组学特征构建预测模型,未引入临床特征分析。对于这些问题还需要增加临床变量、结合影像基因组学及扩大样本量等方法进一步研究。

4 结论

       综上所述,基于IVIM-DWI定量参数的影像组学模型可用于预测鼻咽癌的短期疗效,联合IVIM参数模型的预测性能较单一参数模型的预测效能更高,可以为患者个体化诊疗提供帮助,指导患者的治疗方案选择。

[1]
ZHANG Z M, BAO Y, ZHOU L X, et al. Preliminary application of endonasopharyngeal ultrasound-guided transnasopharyngeal needle aspiration in the diagnosis of submucosal nasopharyngeal carcinoma[J]. Zhonghua Er Bi Yan Hou Tou Jing Wai Ke Za Zhi, 2019, 54(1): 46-49. DOI: 10.3760/cma.j.issn.1673-0860.2019.01.010.
[2]
MASTERSON L, AGALATO E, PEARSON C. Image-guided sinus surgery: practical and financial experiences from a UK centre 2001-2009[J]. J Laryngol Otol, 2012, 126(12): 1224-1230. DOI: 10.1017/S002221511200223X.
[3]
卢玉强, 胡婷, 林初阳, 等. 广东省四会市鼻咽癌筛查方案应用的长期效果评价[J]. 中国肿瘤, 2018, 27(11): 842-846. DOI: 10.11735/j.issn.1004-0242.2018.11.A005.
LU Y Q, HU T, LIN C Y, et al. Evaluation of nasopharyngeal carcinoma screening scheme in Sihui city, Guangdong Province[J]. China Cancer, 2018, 27(11): 842-846. DOI: 10.11735/j.issn.1004-0242.2018.11.A005.
[4]
LEE A W, NG W T, CHAN L L, et al. Evolution of treatment for nasopharyngeal cancer: success and setback in the intensity-modulated radiotherapy era[J]. Radiother Oncol, 2014, 110(3): 377-384. DOI: 10.1016/j.radonc.2014.02.003.
[5]
LIN G W, WANG L X, JI M, et al. The use of MR imaging to detect residual versus recurrent nasopharyngeal carcinoma following treatment with radiation therapy[J]. Eur J Radiol, 2013, 82(12): 2240-2246. DOI: 10.1016/j.ejrad.2013.09.014.
[6]
HUANG Y Y, CAO X, CAI Z C, et al. Short-term efficacy and long-term survival of nasopharyngeal carcinoma patients with radiographically visible residual disease following observation or additional intervention: a real-world study in China[J]. Laryngoscope Investig Otolaryngol, 2022, 7(6): 1881-1892. DOI: 10.1002/lio2.980.
[7]
VAUPEL P, MAYER A. Hypoxia in cancer: significance and impact on clinical outcome[J]. Cancer Metastasis Rev, 2007, 26(2): 225-239. DOI: 10.1007/s10555-007-9055-1.
[8]
DZOBO K, SENTHEBANE D A, DANDARA C. The tumor microenvironment in tumorigenesis and therapy resistance revisited[J/OL]. Cancers, 2023, 15(2): 376[2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/36672326/. DOI: 10.3390/cancers15020376.
[9]
ZHU Y F, LI X Y, WANG L, et al. Metabolic reprogramming and crosstalk of cancer-related fibroblasts and immune cells in the tumor microenvironment[J/OL]. Front Endocrinol, 2022, 13: 988295 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/36046791/. DOI: 10.3389/fendo.2022.988295.
[10]
XIAO Y P, PAN J J, CHEN Y B, et al. Intravoxel incoherent motion-magnetic resonance imaging as an early predictor of treatment response to neoadjuvant chemotherapy in locoregionally advanced nasopharyngeal carcinoma[J/OL]. Medicine, 2015, 94(24): e973 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/26091468/. DOI: 10.1097/MD.0000000000000973.
[11]
QIN Y H, YU X P, HOU J, et al. Predicting chemoradiotherapy response of nasopharyngeal carcinoma using texture features based on intravoxel incoherent motion diffusion-weighted imaging[J/OL]. Medicine, 2018, 97(30): e11676 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/30045324/. DOI: 10.1097/MD.0000000000011676.
[12]
MUI A W L, LEE A W M, LEE V H F, et al. Prognostic and therapeutic evaluation of nasopharyngeal carcinoma by dynamic contrast-enhanced (DCE), diffusion-weighted (DW) magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS)[J/OL]. Magn Reson Imaging, 2021, 83: 50-56 [2023-02-19]. https://pubmed.ncbi.nlm.nih.gov/34246785/. DOI: 10.1016/j.mri.2021.07.003.
[13]
BIHAN D L, TURNER R. The capillary network: a link between IVIM and classical perfusion[J]. Magn Reson Med, 1992, 27(1): 171-178. DOI: 10.1002/mrm.1910270116.
[14]
QAMAR S, KING A D, AI Q H, et al. Pre-treatment intravoxel incoherent motion diffusion-weighted imaging predicts treatment outcome in nasopharyngeal carcinoma[J/OL]. Eur J Radiol, 2020, 129: 109127 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/32563165/. DOI: 10.1016/j.ejrad.2020.109127.
[15]
NOIJ D P, MARTENS R M, MARCUS J T, et al. Intravoxel incoherent motion magnetic resonance imaging in head and neck cancer: a systematic review of the diagnostic and prognostic value[J/OL]. Oral Oncol, 2017, 68: 81-91 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/28438298/. DOI: 10.1016/j.oraloncology.2017.03.016.
[16]
CHEN H, HE Y Y, ZHAO C C, et al. Reproducibility of radiomics features derived from intravoxel incoherent motion diffusion-weighted MRI of cervical cancer[J]. Acta Radiol, 2021, 62(5): 679-686. DOI: 10.1177/0284185120934471.
[17]
GANESHAN B, GOH V, MANDEVILLE H C, et al. Non-small cell lung cancer: histopathologic correlates for texture parameters at CT[J]. Radiology, 2013, 266(1): 326-336. DOI: 10.1148/radiol.12112428.
[18]
SAGAERT X, VANSTAPEL A, VERBEEK S. Tumor heterogeneity in colorectal cancer: what do we know so far?[J]. Pathobiology, 2018, 85(1/2): 72-84. DOI: 10.1159/000486721.
[19]
SPADARELLA G, CALARESO G, GARANZINI E, et al. MRI based radiomics in nasopharyngeal cancer: systematic review and perspectives using radiomic quality score (RQS) assessment[J/OL]. Eur J Radiol, 2021, 140: 109744 [2023-02-01]. https://www.ejradiology.com/article/S0720-048X(21)00225-4/fulltext. DOI: 10.1016/j.ejrad.2021.109744.
[20]
KIM M J, CHOI Y, SUNG Y E, et al. Early risk-assessment of patients with nasopharyngeal carcinoma: the added prognostic value of MR-based radiomics[J/OL]. Transl Oncol, 2021, 14(10): 101180 [2023-02-19]. https://pubmed.ncbi.nlm.nih.gov/34274801/. DOI: 10.1016/j.tranon.2021.101180.
[21]
XU H, LIU J K, HUANG Y, et al. MRI-based radiomics as response predictor to radiochemotherapy for metastatic cervical lymph node in nasopharyngeal carcinoma[J/OL]. Br J Radiol, 2021, 94(1122): 20201212 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/33882240/. DOI: 10.1259/bjr.20201212.
[22]
BORDRON A, RIO E, BADIC B, et al. External validation of a radiomics model for the prediction of complete response to neoadjuvant chemoradiotherapy in rectal cancer[J/OL]. Cancers, 2022,14(4): 1079 [2023-02-01]. https://www.mdpi.com/2072-6694/14/4/1079. DOI: 10.3390/cancers14041079.
[23]
WANG G Y, HE L, YUAN C, et al. Pretreatment MR imaging radiomics signatures for response prediction to induction chemotherapy in patients with nasopharyngeal carcinoma[J/OL]. Eur J Radiol, 2018, 98: 100-106 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/29279146/. DOI: 10.1016/j.ejrad.2017.11.007.
[25]
YU X P, HOU J, LI F P, et al. Intravoxel incoherent motion MRI for predicting early response to induction chemotherapy and chemoradiotherapy in patients with nasopharyngeal carcinoma[J]. J Magn Reson Imaging, 2016, 43(5): 1179-1190. DOI: 10.1002/jmri.25075.
[26]
ZONARI P, BARALDI P, CRISI G. Multimodal MRI in the characterization of glial neoplasms: the combined role of single-voxel MR spectroscopy, diffusion imaging and echo-planar perfusion imaging[J]. Neuroradiology, 2007, 49(10): 795-803. DOI: 10.1007/s00234-007-0253-x.
[27]
TRAN D, NGUYEN D H, NGUYEN H K, et al. Diagnostic performance of MRI perfusion and spectroscopy for brainstem glioma grading[J]. Eur Rev Med Pharmacol Sci, 2022, 26(21): 7938-7948. DOI: 10.26355/eurrev_202211_30145.
[28]
ZHANG L, YANG L Q, WEN L, et al. Noninvasively evaluating the grading of glioma by multiparametric magnetic resonance imaging[J/OL]. Acad Radiol, 2021, 28(5): e137-e146 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/32417035/. DOI: 10.1016/j.acra.2020.03.035.
[29]
ZHU Q Q, ZHU W R, WU J T, et al. Comparative study of conventional diffusion-weighted imaging and introvoxel incoherent motion in assessment of pathological grade of clear cell renal cell carcinoma[J/OL]. Br J Radiol, 2022, 95(1133): 20210485 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/35442093/. DOI: 10.1259/bjr.20210485.
[30]
BISDAS S, KIRKPATRICK M, GIGLIO P, et al. Cerebral blood volume measurements by perfusion-weighted MR imaging in gliomas: ready for prime time in predicting short-term outcome and recurrent disease?[J]. AJNR Am J Neuroradiol, 2009, 30(4): 681-688. DOI: 10.3174/ajnr.A1465.
[31]
ZHU L N, WU J, ZHANG H, et al. The value of intravoxel incoherent motion imaging in predicting the survival of patients with astrocytoma[J]. Acta Radiol, 2021, 62(3): 423-429. DOI: 10.1177/0284185120926907.
[32]
QIN Y H, CHEN C, CHEN H T, et al. The value of intravoxel incoherent motion model-based diffusion-weighted imaging for predicting long-term outcomes in nasopharyngeal carcinoma[J/OL]. Front Oncol, 2022, 12: 902819 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/36531009/. DOI: 10.3389/fonc.2022.902819.
[33]
CHEN W B, ZHANG B, LIANG L, et al. To predict the radiosensitivity of nasopharyngeal carcinoma using intravoxel incoherent motion MRI at 3.0 T[J]. Oncotarget, 2017, 8(32): 53740-53750. DOI: 10.18632/oncotarget.17367.
[34]
LU L Y, LI Y H, LI W B. The role of intravoxel incoherent motion MRI in predicting early treatment response to chemoradiation for metastatic lymph nodes in nasopharyngeal carcinoma[J]. Adv Ther, 2016, 33(7): 1158-1168. DOI: 10.1007/s12325-016-0352-3.
[35]
ZHANG Y, ZHANG K Y, JIA H D, et al. IVIM-DWI and MRI-based radiomics in cervical cancer: prediction of concurrent chemoradiotherapy sensitivity in combination with clinical prognostic factors[J/OL]. Magn Reson Imaging, 2022, 91: 37-44 [2023-02-01]. https://pubmed.ncbi.nlm.nih.gov/35568271/. DOI: 10.1016/j.mri.2022.05.005.

上一篇 基于影像组学的机器学习模型鉴别孤立性纤维性肿瘤与血管瘤型脑膜瘤
下一篇 单、双指数模型扩散加权成像及动脉自旋标记预测复发性鼻咽癌近期疗效的应用价值
  
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