分享:
分享到微信朋友圈
X
特别关注
基于治疗前多参数MRI影像组学特征预测局部晚期宫颈癌患者新辅助化疗后脉管浸润
董林逍 刘金金 张月洁 杨紫涵 吴青霞 王梅云

Cite this article as: DONG L X, LIU J J, ZHANG Y J, et al. Prediction of lymphovascular space invasion in locally advanced cervical cancer patients after neoadjuvant chemotherapy based on pre-treatment multi-parameter MRI radiomics features[J]. Chin J Magn Reson Imaging, 2024, 15(8): 25-30, 45.本文引用格式:董林逍, 刘金金, 张月洁, 等. 基于治疗前多参数MRI影像组学特征预测局部晚期宫颈癌患者新辅助化疗后脉管浸润[J]. 磁共振成像, 2024, 15(8): 25-30, 45. DOI:10.12015/issn.1674-8034.2024.08.004.


[摘要] 目的 基于治疗前多参数磁共振成像(multi-parametric magnetic resonance imaging, mpMRI)影像组学特征构建模型预测局部晚期宫颈癌(locally advanced cervical cancer, LACC)新辅助化疗(neoadjuvant chemotherapy, NACT)后淋巴脉管间隙浸润(lymphovascular space invasion, LVSI)状态。材料与方法 回顾性分析了300例于2013年至2022年来自于河南省人民医院(训练集187人,LVSI阳性73人)和河南省肿瘤医院(验证集113人,LVSI阳性31人)接受NACT并行根治性子宫切除术LACC患者的临床及影像资料。于轴位弥散加权成像(axial diffusion-weighted imaging, Ax_DWI)、矢状位T2加权成像(sagittal T2-weighted imaging, Sag_T2WI)和矢状位对比增强T1加权成像(sagittal T1-weighted contrast-enhanced imaging, Sag_T1C)上勾画肿瘤感兴趣区(region of interest, ROI)并提取特征,利用递归特征消除算法与最小绝对值收缩与选择算法筛选影像组学特征。随后,基于逻辑回归分类器分别建立单序列模型,双序列模型及基于三序列组学特征的联合序列模型。使用受试者工作特征(receiver operating characteristic, ROC)曲线评估各模型性能,使用DeLong检验比较曲线下面积(area under the curve, AUC),通过决策曲线评估模型的临床价值。结果 在验证集中,基于Ax_DWI、Sag_T2WI及Sag_T1C构建的单序列模型的AUC分别为0.717 [95%置信区间(confidence interval, CI):0.605~0.829]、0.734(95% CI:0.633~0.836)和0.733(95% CI:0.626~0.841);基于Ax_DWI+Sag_T2WI、Ax_DWI+Sag_T1C及Sag_T2WI+Sag_T1C构建的双序列模型的AUC值分别为0.763(95% CI:0.660~0.866)、0.786(95% CI:0.692~0.881)与0.815(95% CI:0.731~0.899);联合序列模型的AUC值为0.829(95% CI:0.740~0.914),高于各单序列模型与双序列模型,但联合序列模型与Ax_DWI模型、Sag_T2W1模型及Ax_DWI+Sag_T2W1模型之间AUC差异无统计学意义(P=0.015~0.047)。决策曲线显示联合序列模型的临床净效益高于单序列模型与各双序列模型。结论 基于治疗前mpMRI影像组学特征构建的联合序列模型可有效预测LACC患者NACT后的LVSI状态。
[Abstract] Objective To develop a model utilizing radiomic features from pre-treatment multiparametric magnetic resonance imaging (mpMRI) to predict lymphovascular space invasion (LVSI) status after neoadjuvant chemotherapy (NACT) in locally advanced cervical cancer (LACC).Materials and Methods A retrospective analysis was conducted on clinical and imaging data of 300 patients with locally advanced cervical cancer (LACC) who underwent neoadjuvant chemotherapy (NACT) followed by radical hysterectomy. These patients were divided into a training set (187 patients, with 73 LVSI positive cases) from Henan Provincial People's Hospital and a validation set (113 patients, with 31 LVSI positive cases) from Henan Provincial Cancer Hospital. Tumor regions of interest (ROIs) were delineated on axial diffusion-weighted imaging (Ax_DWI), sagittal T2-weighted imaging (Sag_T2WI), and sagittal T1-weighted contrast-enhanced imaging (Sag_T1C), and features were extracted. Radiomic features were selected using recursive feature elimination (RFE) algorithm and least absolute shrinkage and selection operator (LASSO) algorithm. Subsequently, single-sequence models, dual-sequence models, and combined model based on three-sequence radiomic features were established using logistic regression classifiers. The performance of each model was evaluated using receiver operating characteristic (ROC) curves, with area under the curve (AUC) compared using the Delong test. Clinical utility was assessed using decision curves.Results In the validation set, the AUCs of the single-sequence models constructed based on Ax_DWI, Sag_T2WI, and Sag_T1C were 0.717 [95% confidence interval (CI): 0.605-0.829], 0.734 (95% CI: 0.633-0.836), and 0.733 (95% CI: 0.626-0.841) respectively. The AUCs of the dual-sequence models constructed based on Ax_DWI+Sag_T2WI, Ax_DWI+Sag_T1C, and Sag_T2WI+Sag_T1C were 0.763 (95% CI: 0.660-0.866), 0.786 (95% CI: 0.692-0.881), and 0.815 (95% CI: 0.731-0.899) respectively. The AUC of the combined model was 0.829 (95% CI: 0.740-0.914), which was higher than that of the single-sequence and dual-sequence models, however, the difference in AUC between the combined sequence model and the Ax_DWI model, Sag_T2WI model, as well as the Ax_DWI+Sag_T2WI model was not statistically significant (P=0.015-0.047). Decision curves showed that the clinical net benefit of the joint-sequence model was higher than that of the single-sequence and dual-sequence models.Conclusions The combined model constructed based on pre-treatment multiparametric MRI radiomic features can effectively predict the LVSI status after NACT in LACC patients based on pre-treatment mpMRI.
[关键词] 宫颈癌;淋巴脉管间隙浸润;磁共振成像;影像组学;新辅助化疗
[Keywords] cervical cancer;lymphovascular space invasion;magnetic resonance imaging;radiomics;neoadjuvant chemotherapy

董林逍 1   刘金金 2   张月洁 1   杨紫涵 2   吴青霞 1, 2*   王梅云 1, 2, 3  

1 河南大学人民医院(河南省人民医院)放射科,郑州 450003

2 郑州大学人民医院(河南省人民医院)放射科,郑州 450003

3 河南省科学院医工融合研究所,脑科学与类脑智能实验室,郑州 450003

通信作者:吴青霞,E-mail:qxwu@zzu.edu.cn

作者贡献声明:吴青霞设计本研究的方案;吴青霞、王梅云参与分析、解释本研究的重要数据,同时对稿件重要内容进行了修改;董林逍起草和撰写稿件,搜集、分析和解释本研究的数据;刘金金、杨紫涵、张月洁搜集、分析或解释本研究的数据,对稿件重要内容进行了修改;吴青霞获得了国家自然科学基金(编号:82001783)资助,王梅云获得了国家重点研发计划重点专项、国家自然科学基金(编号:82371934)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: 国家自然科学基金项目 82001783,82371934 国家重点研发计划重点专项 2023YFC2414200
收稿日期:2023-12-31
接受日期:2024-03-21
中图分类号:R445.2  R737.33 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2024.08.004
本文引用格式:董林逍, 刘金金, 张月洁, 等. 基于治疗前多参数MRI影像组学特征预测局部晚期宫颈癌患者新辅助化疗后脉管浸润[J]. 磁共振成像, 2024, 15(8): 25-30, 45. DOI:10.12015/issn.1674-8034.2024.08.004.

0 引言

       宫颈癌是我国最常见的女性生殖系统恶性肿瘤,每年新发和死亡病例分别约13.1万和5.3万[1, 2]。宫颈癌的IB3-IVA期被定义为局部晚期宫颈癌(locally advanced cervical cancer, LACC),大约37%的宫颈癌患者在诊断时即处于LACC阶段[3]。美国国家综合癌症网络(national comprehensive cancer network, NCCN)指南推荐对LACC采用同步根治性放化疗(concurrent chemotherapy and radiotherapy, CCRT)加近距离放疗作为首选治疗方案[4]。尽管目前的证据尚不支持新辅助化疗(neoadjuvant chemotherapy, NACT)使LACC患者长期生存获益,但它有利于减少放疗所带来的消化及泌尿生殖系统并发症[5]。因此,在临床实践中仍有应用[6, 7, 8, 9]。鉴于此,筛选能从CCRT中受益的人群显得尤为重要。

       脉管浸润又称为淋巴脉管间隙浸润(lymphovascular space invasion, LVSI),是指在淋巴管或血管内发现至少一簇肿瘤细胞[10]。根据“Sedlis标准”,LVSI是宫颈癌生存的中等风险因素。对于具有两种及以上中等风险因素的患者,应考虑进行辅助放疗或化疗[11]。因而,利用NACT前的图像预测NACT后是否存在LVSI,有助于避免多种治疗手段所带来的并发症。

       影像组学可从医学图像中高通量地提取影像特征进行定量分析,进而量化肿瘤异质性。近年来,这一技术备受关注[12, 13, 14, 15]。研究报道指出,影像组学已被应用于子宫内膜癌、乳腺癌和胃癌的LVSI预测[16, 17, 18, 19],以及LACC患者NACT后总体生存率的预测[20]。值得注意的是,尚未有关于NACT后LVSI预测的相关报道。因此,本研究拟探讨基于治疗前多参数磁共振成像(multi-parametric magnetic resonance imaging, mpMRI)构建影像组学模型来预测LACC患者NACT后LVSI状态。

1 材料与方法

1.1 研究对象

       本研究遵守《赫尔辛基宣言》,经河南省人民医院(伦理注册号:2021150)及河南省肿瘤医院(伦理注册号:20150315)医学伦理委员会批准,免除患者知情同意。回顾性分析2013年1月至2022年6月接受NACT与根治性子宫切除术的患者。纳入标准:(1)NACT治疗前1周内行盆腔mpMRI平扫及增强扫描;(2)患者接受1~3个周期NACT且术后病理结果完整;(3)NACT前未进行其他治疗。排除标准:(1)MRI伪影大,影响观察;(2)宫颈癌病灶在各序列难以判断。

       本研究最终纳入300例患者临床及影像资料进行模型的构建与验证。其中187人自于河南省人民医院(LVSI阳性73人)作为训练集用于模型构建;113人来自河南省肿瘤医院(LVSI阳性31人)作为验证集用于验证模型效能。

1.2 MRI图像采集

       所有mpMRI检查在1.5 T(signa HDxt, GE Healthcare)与3.0 T(discoveryTM MR 750, GE healthcare; MAGNETOM TrioTim and MAGNETOM Skyra, Siemens Healthineers)磁共振设备上进行。扫描方案包括轴位弥散加权成像(axial diffusion-weighted imaging, Ax_DWI)、矢状位T2加权成像(sagittal T2-weighted imaging, Sag_T2WI)和矢状位对比增强T1加权成像(sagittal T1-weighted contrast-enhanced imaging, Sag_T1C)三个序列,其中Ax_DWI图像均包含低b值0 s/mm2与高b值800 s/mm2。详细参数见表1

表1  MRI检查序列及扫描参数
Tab. 1  MRI examination parameters for each sequence

1.3 NACT方案

       NACT方案包括1至3个周期的含铂静脉化疗,每隔3周进行一次。最后一次NACT结束3周后行根治性子宫切除术和盆腔淋巴结清扫术。如果在术前影像学检查或者术中发现腹膜后淋巴结肿大,则进行腹主动脉旁淋巴结清扫术。

1.4 感兴趣区勾画

       使用Itk-snap(version 4.0.0,http://www.itksnap.org)手动分割病灶感兴趣区(region of interest, ROI)。由两名放射科医生于Ax_DWI(b值为800 s/mm2)、Sag_T2WI和Sag_T1C图像中沿肿瘤边缘逐层手动勾画ROI,并重建生成三维容积感兴趣区,如图1

图1  女,60岁,病理证实为宫颈癌(FIGO分期:IB3期)。轴位弥散加权成像(Ax_DWI;1A)、矢状位对比增强T1加权成像(Sag_T1C;1B)、矢状位T2加权成像(Sag_T2WI;1C)图像,以及对应的感兴趣区(ROI)勾画示意图(1D、1E、1F)、轴位单层宫颈癌病灶ROI(1G)、矢状位单层宫颈癌病灶ROI(1H)、基于ROI融合的三维宫颈癌ROI图像(1I)。
Fig. 1  A 60-year-old female patient with confirmed cervical cancer (FIGO stage: IB3). Illustrates the schematic of regions of interest (ROI) delineated on multiparametric magnetic resonance imaging (mpMRI). 1A: Axial diffusionweighted imaging (Ax_DWI); 1B: Sagittal T1-weighted contrast-enhanced imaging (Sag_T1C); 1C: Sagittal T2-weighted imaging (Sag_T2WI); 1D: Ax_DWI ROI schematics; 1E: Sag_T1C ROI schematics; 1F: Sag_T 2WI ROI schematics; 1G: Single-layer ROI of axial cancer lesion; 1H: Single-layer ROI of sagittal cancer lesion; 1I: Volume of interest in cervical cancer based on fusion of single-layer ROI.

1.5 影像组学特征的提取与选择

       在特征提取前,将每幅图像的灰度量化为5个灰度级,并对每一次MRI扫描进行均值方差归一化,以获得图像强度的标准正态分布;利用重采样和插值,并使用神经网络的b样条曲面(b-spine)构造方法,将Sag_T1C重建为1 mm×1 mm×1 mm体素大小;将Ax_DWI和Sag_T2WI重建为1 mm×1 mm×5 mm体素大小。使用A.K.软件的Pyadiomics包(version 2.3.0)进行影像组学特征提取。在每个序列的ROI中各提取2264个特征,包括形状特征(Shape)、一阶统计特征(First-order)、灰度共生矩阵(gray level co-occurrence matrix, GLCM)、灰度尺寸区域矩阵(gray level size zone matrix, GLSZM)特征、灰度游程长度矩阵(gray level run length matrix, GLRLM)、灰度相关矩阵(gray level dependence matrix, GLDM)和邻域灰度差矩阵(neighboring gray tone difference matrix, NGTDM)。

       使用Z-score标准化方法对特征进行归一化,将数据转换为标准化强度范围,降低特征的量纲差异。为了减少机器学习模型的过拟合,首先采用递归特征消除算法(recursive feature elimination, RFE),其次采用最小绝对值收缩与选择算法(least absolute shrinkage and selection operator, LASSO)降维,并在训练队列中进行10倍交叉验证,以获得最优参数并过滤影像组学特征(图2)。

图2  最小绝对收缩与选择算子(LASSO)算法降维特征筛选,LASSO模型中调节参数(λ)的选择。X轴代表λ的值,Y轴代表验证集曲线下面积(AUC)值,展示了不同λ值下LASSO模型AUC的变化情况。
Fig. 2  illustrates the feature selection using the least absolute shrinkage and selection operator (LASSO) algorithm and the selection of tuning parameter (λ) in the LASSO model. The X-axis represents the values of λ, while the Y-axis represents the values of the area under the curve (AUC) on the validation set. The plot demonstrates the variation of the AUC of the LASSO model with different values of λ.

1.6 影像组学模型建立与评价

       首先,使用逻辑回归分类器,利用从Ax_DWI、Sag_T2WI和Sag_T1C筛选出的影像组学特征分别建立相应序列的单序列模型。随后,将三序列所提取出的特征进行两两融合,分别构建基于Ax_DWI+Sag_T2WI、Ax_DWI+Sag_T1C与Sag_T2WI+Sag_T1C的双序列模型。最后,融合所提取的三个序列的组学特征构建三模态的联合序列模型。模型的构建和验证均在开源软件PyRadiomics(version 3.0.1, http://github.com/Radiomics/pyradiomics)中进行。

       通过受试者工作特征(receiver operating characteristic, ROC)曲线下面积(area under the curve, AUC)、敏感度、特异度、阴性预测值(negative predictive value, NPV)、阳性预测值(positive predictive value, PPV)和准确率来评价每个模型的效能。

1.7 统计学分析

       基于SPSS(version 27.0, IBM Corp., USA, https://www.ibm.com/cn-zh/spss)、R语言(version 4.2.0, R Foundation for Statistical Computing, Austria, https://www.r-project.org/)进行统计学分析。分类变量的组间比较使用Pearson卡方检验或Fisher精确检验,连续变量使用Mann-Whitney U秩和检验。采用DeLong检验比较不同模型间AUC的差异。运用决策曲线分析(decision curve analysis, DCA)计算不同模型在不同阈值概率下的净收益。使用校准曲线和Hosmer-Lemeshow拟合优度检验评估联合序列模型的模型拟合度。P<0.05认为差异具有统计学意义。

2 结果

2.1 研究对象特征

       在训练集与验证集中,不同LVSI状态患者的年龄、组织学类型及NACT周期差异均无统计学意义(P>0.05),入组患者的一般资料见表2。单因素及多因素逻辑回归显示年龄、NACT周期、组织学类型与LVSI间差异均无统计学意义(P>0.05)。

表2  不同LVSI状态组患者的临床及病理特征
Tab. 2  Clinical and pathological characteristics of patients in different lymph node status groups

2.2 影像组学预测模型与效能评估

       每个序列各筛选出10个与宫颈癌LVSI相关的影像组学特征,三个序列共包括4个一阶特征与6个GLCM、12个GLZSM、2个GLRLM和6个GLDM特征。

       各影像组学预测模型诊断效能分析发现(图3表3),在训练集中,基于Ax_DWI、Sag_T2WI和Sag_T1C的单序列组学模型的AUC分别为0.773 [95%置信区间(confidence interval, CI):0.704~0.842]、0.779(95% CI:0.712~0.847)和0.793(95% CI:0.725~0.862),各模型之间AUC差异均无统计学意义(P=0.698~0.899);在验证集中,基于Ax_DWI、Sag_T2WI和Sag_T1C的单序列组学模型的AUC分别为0.717(95% CI:0.605~0.829)、0.734(95% CI:0.633~0.836)和0.733(95% CI:0.626~0.841),各模型之间AUC差异均无统计学意义(P=0.803~0.990)。

       基于Ax_DWI+Sag_T2WI、Ax_DWI+Sag_T1C以及Sag_T2WI+Sag_T1C构建的双序列模型在训练集上AUC分别为0.847(95% CI:0.791~0.903)、0.871(95% CI:0.819~0.923)与0.850(95% CI:0.795~0.905),在验证集上AUC分别为0.763(95% CI:0.660~0.866)、0.786(95% CI:0.692~0.881)与0.815(95% CI:0.731~0.899),但各模型之间AUC差异均无统计学意义(P=0.351~0.632)。

       基于三个序列构建的联合序列模型,在训练集AUC为0.913(95% CI:0.871~0.955),在验证集AUC为0.829(95% CI:0.740~0.914),优于各单序列模型与双序列模型,但联合序列模型与Ax_DWI模型、Sag_T2W1模型及Ax_DWI+Sag_T2W1模型之间AUC差异无统计学意义(P=0.015~0.047)。

图3  各模型的受试者工作特征曲线。3A:训练集;3B:验证集。Ax_DWI:轴位弥散加权成像;Sag_T2WI:矢状位T2加权成像;Sag_T1C:矢状位对比增强T1加权成像。
Fig. 3  The receiver operating characteristic (ROC) curves of each model. 3A: Training set; 3B: Validation set. Ax_DWI: axial diffusion-weighted imaging; Sag_T2WI: sagittal T1-weighted imaging; Sag_T1C: sagittal T1-weighted contrast-enhanced imaging.
表3  各MRI影像组学模型对LACC脉管浸润的预测效能
Tab. 3  The predictive efficacy of each MRI radiomics models for LVSI in LACC

2.3 模型的临床应用价值

       验证集中各预测模型的DCA(图4)结果显示,在风险阈值概率为40%~90%之间时,联合序列模型的临床效益优于各单序列模型与双序列模型。校准曲线提示联合序列模型预测结果与实际观测值之间良好的一致性,在训练集和验证集中Hosmer-Lemeshow拟合优度检验的P值分别为0.671和0.882(图5)。

图4  验证集中各模型的DCA曲线。决策曲线显示如果风险阈值处于40%~90%之间时,则用联合序列模型预测治疗后LVSI状态可获得更大的收益。
图5  验证集中联合序列模型的校准曲线。校准图显示,实际概率和预测概率之间预测准确性良好。DCA:决策曲线分析;LVSI:淋巴脉管间隙浸润;Ax_DWI:轴位弥散加权成像;Sag_T2WI:矢状位T2加权成像;Sag_T1C:矢位对比增强T1加权成像。
Fig. 4  The DCA curve of each model in validation. The decision curve indicates that if the risk threshold lies between 40% to 90%, using the Combined model for predicting post-treatment LVSI status yields greater profit.
Fig. 5  The calibration curve of the combined model in validation. The calibration plot revealed good predictive accuracy between the actual probability and predicted probability. DCA: decision curve analysis; LVSI: lymphovascular space invasion; Ax_DWI: axial diffusion-weighted imaging; Sag_T2WI: sagittal T1-weighted imaging; Sag_T1C: sagittal T1-weighted contrast-enhanced imaging.

3 讨论

       本研究首次建立基于NACT前的mpMRI影像数据的多个单序列模型、双序列模型和联合序列模型,这些模型可被用于预测LACC患者NACT后LVSI的状态。其中联合模型的预测价值最高,提示基线mpMRI组学特征有望帮助LACC患者个体化选择治疗方案,避免过度诊疗并改善患者预后。

3.1 影像组学特征对于预测NACT后LVSI的价值

       影像组学特征包括一阶特征、形态特征和纹理特征,不同特征维度代表不同的肿瘤信息[21, 22]。一阶特征显示灰度的整体分布情况,是灰度的统计学特征;纹理特征则显示灰度变化情况,反映肿瘤内异质性特点。本研究最终筛选出4个一阶特征与26个纹理特征,其中4个一阶特征包括2个T2WI序列特征,提示T2信号强度可能与宫颈癌LVSI相关;其余特征则均为纹理特征,提示瘤体内异质性特点可以用于预测LVSI[23, 24]。在本研究中,我们未能提取出形状特征,可能是由于图像的层厚或重采样方法不同[25]。另外,不同的图像分割方法会引起ROI的边界信息发生变化,从而导致形状特征无法被筛选出[26]

3.2 MRI的影像组学评价

       MRI各序列图像中,T2WI显示宫颈癌病灶位置、形态及浸润范围的效果较好,而T1C可重点显示病变强化程度、提高病灶与周围正常组织的对比度,因此探讨不同序列对于肿瘤病理特征的预测价值有助于患者扫描方案的选择。相关研究证明基于T2WI序列的模型可作为预测宫颈癌LVSI的工具[27, 28]。LI等利用基于T1C序列构建的列线图来预测宫颈癌LVSI,在训练集与验证集中AUC分别达到了0.754与0.727。本研究显示Sag_T2WI组学模型相较于Ax_DWI和Sag_T1C模型具有更高的敏感度与AUC,但各序列模型之间AUC相差不大。根据欧洲泌尿生殖系统放射学会指南,MRI对于宫颈癌的分期、治疗反应与复发评估具有很高的准确性,特别是T2WI和DWI序列,而对比增强MRI的价值仍需进一步研究[29]。本研究建立了Sag_T1C模型与基于各序列所构建的Ax_DWI+Sag_T1C模型、Sag_T2WI+Sag_T1C模型与联合序列模型,最终联合序列模型在验证集上取得最高的AUC(0.829)。这表明,虽然单独应用Sag_T1C序列的价值有限,但是Sag_T1C结合其他序列构建的联合模型则明显提高了LVSI预测效能。因此,T1C序列与其他序列的联合应用对于LVSI的预测价值值得进一步研究。

       mpMRI可以从不同方面反映病变的特征,包括肿瘤强化程度、细胞密度和血管分布,这些特征的补充价值会提高影像组学模型的校准与区分能力[30, 31]。先前对病理学确认为宫颈癌的患者进行的研究表明,基于mpMRI的临床-影像组学联合模型在预测LVSI方面取得了高AUC值(训练集和测试集均>0.88),这优于临床模型和单一序列模型[32, 33]。本研究对象为NACT后手术的LACC患者,基于mpMRI的影像组学联合模型也得到了相似的结论,提示基于mpMRI建立的影像组学模型可以作为预测LACC患者NACT后LVSI的工具。其他研究显示组织学类型可在一定程度上预测宫颈癌LVSI状态[34],本研究中LVSI阳性与LVSI阴性患者间组织学类型差异无统计学意义,可能与样本量小有关。

3.3 本研究的局限性

       本研究有一定的局限性。首先,本研究是一项回顾性研究,且样本量仍然较小,未来需要大样本的前瞻性研究;其次,未纳入NACT后MRI影像,因此无法根据NACT后影像预测NACT后LVSI状态;最后,本研究纳入数据来源跨度达10年,mpMRI数据来源不同设备、不同参数,数据的同质化较难保证,可能会对结果造成一定影响。

4 结论

       综上所述,基于治疗前mpMRI影像组学特征构建的联合序列模型可有效预测LACC患者NACT后的LVSI状态,可为临床制订个体化治疗方案提供依据。

[1]
刘萍. 中国大陆13年宫颈癌临床流行病学大数据评价[J]. 中国实用妇科与产科杂志, 2018, 34(1): 41-45. DOI: 10.19538/j.fk2018010111.
LIU P. Big data evaluation of the clinical epidemiology of cervicalcancer in China's mainland[J]. Chin J Pract Gynecol Obstet, 2018, 34(1): 41-45. DOI: 10.19538/j.fk2018010111.
[2]
BRAY F, FERLAY J, SOERJOMATARAM I, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA Cancer J Clin, 2018, 68(6): 394-424. DOI: 10.3322/caac.21492.
[3]
MONK B J, TAN D S P, CHAGÜI J D H, et al. Proportions and incidence of locally advanced cervical cancer: a global systematic literature review[J]. Int J Gynecol Cancer, 2022, 32(12): 1531-1539. DOI: 10.1136/ijgc-2022-003801.
[4]
KATO S, OHNO T, THEPHAMONGKHOL K, et al. Long-term follow-up results of a multi-institutional phase 2 study of concurrent chemoradiation therapy for locally advanced cervical cancer in east and Southeast Asia[J]. Int J Radiat Oncol Biol Phys, 2013, 87(1): 100-105. DOI: 10.1016/j.ijrobp.2013.04.053.
[5]
GUPTA S, MAHESHWARI A, PARAB P, et al. Neoadjuvant chemotherapy followed by radical surgery versus concomitant chemotherapy and radiotherapy in patients with stage IB2, IIA, or IIB squamous cervical cancer: a randomized controlled trial[J]. J Clin Oncol, 2018, 36(16): 1548-1555. DOI: 10.1200/JCO.2017.75.9985.
[6]
MIRIYALA R, MAHANTSHETTY U, MAHESHWARI A, et al. Neoadjuvant chemotherapy followed by surgery in cervical cancer: past, present and future[J]. Int J Gynecol Cancer, 2022, 32(3): 260-265. DOI: 10.1136/ijgc-2021-002531.
[7]
MOUSAVI A, GILANI M M, AKHAVAN S, et al. The outcome of locally advanced cervical cancer in patients treated with neoadjuvant chemotherapy followed by radical hysterectomy and primary surgery[J]. Iran J Med Sci, 2021, 46(5): 355-363. DOI: 10.30476/ijms.2020.81973.0.
[8]
CHEN H J, LIANG C, ZHANG L, et al. Clinical efficacy of modified preoperative neoadjuvant chemotherapy in the treatment of locally advanced (stage IB2 to IIB) cervical cancer: randomized study[J]. Gynecol Oncol, 2008, 110(3): 308-315. DOI: 10.1016/j.ygyno.2008.05.026.
[9]
BENSON R, PATHY S, KUMAR L, et al. Locally advanced cervical cancer-neoadjuvant chemotherapy followed by concurrent chemoradiation and targeted therapy as maintenance: a phase II study[J]. J Cancer Res Ther, 2019, 15(6): 1359-1364. DOI: 10.4103/jcrt.JCRT_39_18.
[10]
KIKUCHI E, MARGULIS V, KARAKIEWICZ P I, et al. Lymphovascular invasion predicts clinical outcomes in patients with node-negative upper tract urothelial carcinoma[J]. J Clin Oncol, 2009, 27(4): 612-618. DOI: 10.1200/JCO.2008.17.2361.
[11]
ABU-RUSTUM N R, YASHAR C M, BEAN S, et al. NCCN guidelines insights: cervical cancer, version 1.2020[J]. J Natl Compr Canc Netw, 2020, 18(6): 660-666. DOI: 10.6004/jnccn.2020.0027.
[12]
LI Z C, LI H L, WANG S Y, et al. MR-based radiomics nomogram of cervical cancer in prediction of the lymph-vascular space invasion preoperatively[J]. J Magn Reson Imaging, 2019, 49(5): 1420-1426. DOI: 10.1002/jmri.26531.
[13]
LAMBIN P, RIOS-VELAZQUEZ E, LEIJENAAR R, et al. Radiomics: extracting more information from medical images using advanced feature analysis[J]. Eur J Cancer, 2012, 48(4): 441-446. DOI: 10.1016/j.ejca.2011.11.036.
[14]
YIP S S F, AERTS H J W L. Applications and limitations of radiomics[J]. Phys Med Biol, 2016, 61(13): R150-R166. DOI: 10.1088/0031-9155/61/13/R150.
[15]
GILLIES R J, KINAHAN P E, HRICAK H. Radiomics: images are more than pictures, they are data[J]. Radiology, 2016, 278(2): 563-577. DOI: 10.1148/radiol.2015151169.
[16]
LUO Y, MEI D D, GONG J S, et al. Multiparametric MRI-based radiomics nomogram for predicting lymphovascular space invasion in endometrial carcinoma[J]. J Magn Reson Imaging, 2020, 52(4): 1257-1262. DOI: 10.1002/jmri.27142.
[17]
LIU X F, YAN B C, LI Y, et al. Radiomics feature as a preoperative predictive of lymphovascular invasion in early-stage endometrial cancer: a multicenter study[J/OL]. Front Oncol, 2022, 12: 966529 [2024-03-12]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9433783/pdf/fonc-12-966529.pdf. DOI: 10.3389/fonc.2022.966529.
[18]
MENG L W, DONG D, CHEN X, et al. 2D and 3D CT radiomic features performance comparison in characterization of gastric cancer: a multi-center study[J]. IEEE J Biomed Health Inform, 2021, 25(3): 755-763. DOI: 10.1109/JBHI.2020.3002805.
[19]
LIU Z S, FENG B, LI C L, et al. Preoperative prediction of lymphovascular invasion in invasive breast cancer with dynamic contrast-enhanced-MRI-based radiomics[J]. J Magn Reson Imaging, 2019, 50(3): 847-857. DOI: 10.1002/jmri.26688.
[20]
AUTORINO R, GUI B, PANZA G, et al. Radiomics-based prediction of two-year clinical outcome in locally advanced cervical cancer patients undergoing neoadjuvant chemoradiotherapy[J]. La Radiol Med, 2022, 127(5): 498-506. DOI: 10.1007/s11547-022-01482-9.
[21]
CRANDALL J P, FRAUM T J, LEE M, et al. Repeatability of 18F-FDG PET radiomic features in cervical cancer[J]. J Nucl Med, 2021, 62(5): 707-715. DOI: 10.2967/jnumed.120.247999.
[22]
YAP F Y, VARGHESE B A, CEN S Y, et al. Shape and texture-based radiomics signature on CT effectively discriminates benign from malignant renal masses[J]. Eur Radiol, 2021, 31(2): 1011-1021. DOI: 10.1007/s00330-020-07158-0.
[23]
KIDOH M, SHINODA K, KITAJIMA M, et al. Deep learning based noise reduction for brain MR imaging: tests on phantoms and healthy volunteers[J]. Magn Reson Med Sci, 2020, 19(3): 195-206. DOI: 10.2463/mrms.mp.2019-0018.
[24]
TIAN C W, FEI L K, ZHENG W X, et al. Deep learning on image denoising: an overview[J]. Neural Netw, 2020, 131: 251-275. DOI: 10.1016/j.neunet.2020.07.025.
[25]
ZHAO B S, TAN Y Q, TSAI W Y, et al. Reproducibility of radiomics for deciphering tumor phenotype with imaging[J/OL]. Sci Rep, 2016, 6: 23428 [2024-03-12]. https://www.nature.com/articles/srep23428.pdf. DOI: 10.1038/srep23428.
[26]
TRAVERSO A, WEE L, DEKKER A, et al. Repeatability and reproducibility of radiomic features: a systematic review[J]. Int J Radiat Oncol, 2018, 102(4): 1143-1158. DOI: 10.1016/j.ijrobp.2018.05.053.
[27]
周小玲, 赖华, 文曦琳, 等. T2WI-FS序列影像组学诊断宫颈癌转移及脉管间隙浸润的价值[J]. 磁共振成像, 2021, 12(7): 69-71, 76. DOI: 10.12015/issn.1674-8034.2021.07.014.
ZHOU X L, LAI H, WEN X L, et al. Value of T2WI-FS based radiomics features in the diagnosis of cervical cancer metastasis and lymph vascular space invasion[J]. Chin J Magn Reson Imag, 2021, 12(7): 69-71, 76. DOI: 10.12015/issn.1674-8034.2021.07.014.
[28]
DU W, WANG Y, LI D D, et al. Preoperative prediction of lymphovascular space invasion in cervical cancer with radiomics-based nomogram[J/OL]. Front Oncol, 2021, 11: 637794 [2024-03-12]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8311659/pdf/fonc-11-637794.pdf. DOI: 10.3389/fonc.2021.637794.
[29]
AVESANI G, PERAZZOLO A, AMERIGHI A, et al. The utility of contrast-enhanced magnetic resonance imaging in uterine cervical cancer: a systematic review[J/OL]. Life, 2023, 13(6): 1368 [2024-03-12]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10303560/pdf/life-13-01368.pdf. DOI: 10.3390/life13061368.
[30]
HIMOTO Y, FUJIMOTO K, KIDO A, et al. Assessment of the early predictive power of quantitative magnetic resonance imaging parameters during neoadjuvant chemotherapy for uterine cervical cancer[J]. Int J Gynecol Cancer, 2014, 24(4): 751-757. DOI: 10.1097/IGC.0000000000000124.
[31]
FENG Y S, LIU H, DING Y Y, et al. Combined dynamic DCE-MRI and diffusion-weighted imaging to evaluate the effect of neoadjuvant chemotherapy in cervical cancer[J]. Tumori, 2020, 106(2): 155-164. DOI: 10.1177/0300891619886656.
[32]
HUANG G, CUI Y Q, WANG P, et al. Multi-parametric magnetic resonance imaging-based radiomics analysis of cervical cancer for preoperative prediction of lymphovascular space invasion[J/OL]. Front Oncol, 2022, 11: 663370 [2024-03-12]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8790703/pdf/fonc-11-663370.pdf. DOI: 10.3389/fonc.2021.663370.
[33]
WU Y, WANG S X, CHEN Y Q, et al. A multicenter study on preoperative assessment of lymphovascular space invasion in early-stage cervical cancer based on multimodal MR radiomics[J]. J Magn Reson Imaging, 2023, 58(5): 1638-1648. DOI: 10.1002/jmri.28676.
[34]
周滢, 姜继勇. 早期宫颈癌淋巴脉管间隙浸润与临床病理因素及预后的关系[J]. 实用妇产科杂志, 2018, 34(3): 203-207.
ZHOU Y, JIANG J Y. Relationship among lymph vascular space invasion with clinicopathological factors and prognosis in early cervical cancer[J]. J Pract Obstet Gynecol, 2018, 34(3): 203-207.

上一篇 治疗前多参数MRI影像组学特征预测晚期宫颈鳞癌患者新辅助化疗后淋巴结转移
下一篇 基于APTw的影像组学术前预测宫颈癌淋巴血管间隙侵犯
  
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