分享:
分享到微信朋友圈
X
特别关注
基于矢状位T2WI瘤内瘤周影像组学列线图术前预测ⅠB期和ⅡA期宫颈癌的研究
徐青 彭雪艳 郭长义 朱欣阳 贺朝

Cite this article as: XU Q, PENG X Y, GUO C Y, et al. Intra- and peritumoral sagittal T2WI radiomics nomogram for preoperative prediction of patients with stage ⅠB and stage ⅡA cervical cancer[J]. Chin J Magn Reson Imaging, 2024, 15(8): 46-51, 64.本文引用格式:徐青, 彭雪艳, 郭长义, 等. 基于矢状位T2WI瘤内瘤周影像组学列线图术前预测ⅠB期和ⅡA期宫颈癌的研究[J]. 磁共振成像, 2024, 15(8): 46-51, 64. DOI:10.12015/issn.1674-8034.2024.08.007.


[摘要] 目的 建立一个基于矢状位T2加权图像(T2 weighted imaging, T2WI)的瘤内结合不同瘤周区域影像组学特征和临床危险因素的列线图,评估其术前预测宫颈癌ⅠB期和ⅡA期的价值。材料与方法 回顾性分析120例两家医院术前接受MRI检查和根治性子宫切除术加盆腔淋巴清扫并经术后病理证实的ⅠB和ⅡA期宫颈癌患者的实验室检查资料和MRI图像,分为训练组和外部验证组,对矢状位T2WI图像瘤内感兴趣区(region of interest, ROI)和1~6 mm的瘤周环(ROI-1、ROI-2、ROI-3、ROI-4、ROI-5和ROI-6)分别提取影像组学特征,采用Pearson分析和最小绝对收缩和选择算子(the least absolute shrinkage and selection operator, LASSO)回归进行特征选择。以最佳者绘制受试者工作特征(receiver operating characteristic, ROC)曲线,构建0~6 mm瘤周影像组学模型,并筛选出最佳影像组学模型,基于以上最佳影像组学模型评分与临床独立危险因素构建联合模型,并绘制列线图,以校准曲线评估模型校准度,以决策曲线分析(decision curve analysis, DCA)评价模型的应用价值。结果 由瘤内结合瘤周3 mm区域得到6个有效特征建立的影像组学模型预测效能最佳,在训练组和外部验证组,其曲线下面积(area under the curve, AUC)分别为0.972和0.857。由肿瘤最大径、红细胞计数(red blood cell, RBC)的临床独立危险因素建立的临床模型预测效能次之,在训练组和外部验证组AUC分别为0.940和0.847。基于肿瘤最大径、RBC、瘤内结合瘤周3 mm的6个有效特征构建的列线图预测效能更稳定,在训练组和外部验证组AUC分别为0.952和0.939,经校准曲线和DCA分析,列线图的校准度较高,临床净收益较大。结论 基于T2WI瘤内和瘤周3 mm组学特征构建的影像组学列线图可以较好地术前预测ⅠB期和ⅡA期宫颈癌,对指导患者个体化治疗有着重要临床意义。
[Abstract] Objective A comprehensive nomogram based on radiomics signature and clinical risk factors in the intra-and peritumoral regions of T2 weighted imaging (T2WI) was developed for the prediction of ⅠB and ⅡA stage in cervical cancer.Materials and Methods A total of 120 patients with stage ⅠB and ⅡA cervical cancer, who underwent preoperative MRI and radical hysterectomy with systematic pelvic lymph node dissection, were analysed retrospectively from two hospitals, and then randomly divided into training (n=80) and external validation groups (n=40). Intra- and peritumoral features (0-6 mm) were extracted separately in T2WI and selected by the Pearson analysis and the least absolute shrinkage and selection operator (LASSO) regression. Radiomic models were built using the best selected features from different regions. Receiver operating characteristic (ROC) was drew and the prediction performance of multi-regional radiomic models was built. Finally, the optimal peritumoral region was selected and the nomogram was developed combining the optimal peritumoral radiomics signature and the most predictive clinical parameters. The calibration degree of the model was evaluated by calibration curve and the application value of the model was evaluated by decision curve analysis (DCA).Results Six effective radiomics features, selected from the peritumoral regions with 3 mm distances in the T2WI, had the best predictive performance, achieving an area under curve (AUC) of 0.972 and 0.857 in the training and validation groups, respectively. The prediction efficiency of the model based on the maximum diameter and red blood cell (RBC), which were the clinical independent risk factors, is next, achieving an AUC of 0.940 and 0.847 in the training and validation groups, respectively. The prediction efficiency of the nomogram based on the maximum diameter, red RBC and six effective radiomics features from the peritumoral regions with 3 mm distances was more stable, achieving an AUC of 0.952 and 0.939 in the training and validation groups, respectively. The nomogram, tested by calibration curve and DCA, had the higher calibration and greater net clinical benefit.Conclusions The nomogram that was developed by intra- and peritumoral regions with 3 mm distances radiomics was excellent for the preoperative prediction of ⅠB and ⅡA stage in cervical cancer. It is important clinical significance to guide the individual treatment of patients.
[关键词] 宫颈癌;子宫广泛性切除术;影像组学;磁共振成像;列线图;术前分期
[Keywords] cervical cancer;radiomics;radical hysterectomy;magnetic resonance imaging;nomogram;preoperative staging

徐青 1, 2   彭雪艳 3   郭长义 2   朱欣阳 2   贺朝 1, 2*  

1 陕西中医药大学医学技术学院,咸阳 712046

2 陕西中医药大学第二附属医院影像中心,咸阳 712046

3 咸阳市中心医院,咸阳 712046

通信作者:贺朝,E-mail:1753972278@qq.com

作者贡献声明:贺朝、郭长义设计本研究的方案,对稿件重要内容进行了修改,贺朝获得了咸阳市重点研发计划社会发展领域项目资助;徐青起草和撰写稿件,获取、分析和解释本研究的数据/文献;彭雪艳、朱欣阳获取分析和解释本研究的数据/文献,对稿件重要内容进行了修改,朱欣阳获得了陕西省科技厅自然科学基础研究计划青年项目和陕西省教育厅自然科学专项基金资助;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: 陕西省科技厅自然科学基础研究计划青年项目 2024JC-YBQN-0795 陕西省教育厅自然科学专项 21JK0593 咸阳市重点研发计划社会发展领域项目 L2023-ZDYF-SF-054
收稿日期:2024-01-27
接受日期:2024-05-13
中图分类号:R445.2  R737.33 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2024.08.007
本文引用格式:徐青, 彭雪艳, 郭长义, 等. 基于矢状位T2WI瘤内瘤周影像组学列线图术前预测ⅠB期和ⅡA期宫颈癌的研究[J]. 磁共振成像, 2024, 15(8): 46-51, 64. DOI:10.12015/issn.1674-8034.2024.08.007.

0 引言

       宫颈癌是全球女性中第四大最常见恶性肿瘤,对女性健康构成了严重威胁[1]。有研究表明,宫颈癌是20~39岁年轻女性癌症死亡的第二大原因[2],保留生育能力是早期宫颈癌患者,尤其是有生育意愿年轻宫颈癌患者的强烈需求。根据国际妇产科协会(Federation International of Gynecology and Obstetrics, FIGO)(2018版)指南[3],阴道受侵是早期宫颈癌分期的关键依据。美国国家综合癌症网络建议[4],ⅠB期、肿瘤大小≤2 cm、生殖年龄<45岁、无临床证据表明生育能力受损的宫颈癌患者,若成像检查也未发现转移证据,可以选择宫颈切除等治疗方式而保留生育能力。因此,准确判断早期宫颈癌分期,尤其是准确判断是否存在阴道受侵对早期宫颈癌患者选择保留生育能力的手术方式至关重要。影像组学从医学图像中提取高通量的定量特征反映肿瘤的异质性[5, 6]。近年来,有文献报道,瘤周区域也蕴含丰富肿瘤异质性信息[7, 8],在疾病精准分期[9, 10]、远处转移[11]及疗效评估[12]等方面表现出巨大优势。有研究表明[13, 14, 15],影像组学可以很好地术前预测ⅠA2,ⅠB和ⅡA期宫颈癌患者宫旁浸润和淋巴结转移情况,提高了宫颈癌分期诊断率,但尚未有研究报道影像组学在评估早期宫颈癌患者阴道受侵情况的潜在应用价值。本研究假设通过提取矢状位T2WI图像的瘤内瘤周组学特征,以更全面反映肿瘤异质性,同时将获取的组学特征与临床资料结合构建列线图,旨在进一步评估早期宫颈癌患者的阴道受侵情况,帮助临床提高ⅠB期和ⅡA期宫颈癌的诊断率,为患者选择适合的手术方式,保留生育能力。

1 材料与方法

1.1 研究对象

       本回顾性研究遵守《赫尔辛基宣言》,获得陕西中医药大学第二附属医院伦理审查委员会的批准(批准文号:LW2023022)和咸阳市中心医院的伦理审查委员会的批准(批准文号:2023-IRB-20),免除受试者知情同意。本研究遵循诊断试验准确性研究报告标准(Standards for Reporting of Diagnostic Accuracy, STARD)。

       本研究回顾性分析于2019年1月至2023年11月期间在陕西中医药大学第二附属医院和咸阳市中心医院确诊的120例早期宫颈癌患者病例,并从医院医疗记录系统收集了人口统计学特征、实验室检查和MRI图像。纳入标准:(1)病理确诊的ⅠB期和ⅡA期宫颈癌患者;(2)完整且可获得患者临床资料以及MRI图像;(3)患者行根治性子宫切除术和双侧盆腔淋巴结清扫术。排除标准:(1)患者同时有其他肿瘤病史;(2)术前接受新辅助化疗或放疗患者;(3)磁共振图像质量差或有严重运动伪影;(4)肿瘤过小(≤5 mm)在磁共振图像上不易勾画感兴趣区(region of interest, ROI)。受试者资料包括年龄、月经史、人乳头状瘤病毒(human papilloma virus, HPV)感染情况、肿瘤最大直径、鳞状细胞癌胚抗原(squamous cell carcinoma antigen, SCCA)、红细胞计数(red blood cell, RBC)、白细胞计数(white blood cell, WBC)和血小板计数(platelet, PLT)。陕西中医药大学第二附属医院病例作为训练组(共80例,包括45例ⅠB期患者和35例ⅡA期患者)和咸阳市中心医院病例作为外部验证组(共40例,包括25例ⅠB期患者和15例ⅡA期患者)。

1.2 扫描设备及参数

       MRI图像由两家医院不同MRI设备获得,为了避免图像信息丢失,我们直接从图像存档和通信系统(picture archiving and communication systems, PACS)中获取了医学数字成像和通信(digital imaging and communications in medicine, DICOM)图像,不进行任何压缩或降采样。陕西中医药大学第二附属医院病例在西门子3.0 T Skyar MRI上完成检查,使用8通道相控阵腹部线圈。扫描参数:矢状位T2WI序列,重复时间(repetition time ,TR)/回波时间(echo time, TE)4700/48 ms,视野(field of view, FOV)320 mm×320 mm,激发数(number of excitations, NEX)2,层厚5 mm,层间距1 mm。咸阳市中心医院病例在西门子3.0 T Skyar MRI和Philips Achieva 1.5 T MR上完成检查,使用8通道相控阵腹部线圈。3.0 T Skyar MRI扫描参数:矢状位T2WI序列,TR/TE 3966/86 ms,FOV 280 mm×280 mm,NEX 2,层厚4.5 mm,层间距1.125 mm;Philips Achieva 1.5 T MRI扫描参数:矢状T2WI序列,TR/TE 2710/90 ms,FOV 260 mm×143 mm,NEX 2,层厚5 mm,层间距1 mm。

1.3 影像组学分析

       影像组学分析流程包括四个步骤:图像分割、影像组学特征提取、特征选择和模型构建(图1)。

1.3.1 肿瘤图像分割

       使用联影智能科研平台(Version 20231115, http://urp.united-imaging.com/)[16]对矢状位T2WI图像中的肿瘤进行逐层手工分割。对于边界模糊地方,参考其他序列。ROI包括肿瘤内的出血、坏死或囊性区域,同时避免肿瘤组织附近的正常肌膜。利用ROI基于阈值去除宫颈内黏液部分。使用numpy库中的pad函数(Version 1.21.2, https://numpy.org/doc/stable/reference/generated/numpy.pad.html)将分割的原始ROI以1 mm的间隔扩张距离(最大6 mm),对膨胀到邻近器官的部分进行手动擦除。肿瘤分割和瘤周区域见图2。邀请一位具有8年丰富妇科盆腔MRI诊断经验的影像科副主任医师在不了解患者病理结果情况下勾画,邀请另一位具有10年丰富ROI分割结果验证经验的影像科副主任医师在不了解患者病理结果情况下验证。

1.3.2 图像预处理和影像组学特征提取

       采用联影智能科研平台上的最近邻插值算法将图像重采样到3 mm×3 mm×3 mm,以获得相同的体素间距。将预处理后的图像通过24个滤波器进行变换,最终在T2WI图像中每层肿瘤和瘤周区域提取了包括一阶特征、形状特征、纹理特征和小波特征在内的2264个组学特征。纹理特征包括灰度大小区域矩阵(gray level size zone matrix, GLSZM)、灰度共生矩阵(gray level cooccurrence matrix, GLCM)、邻接灰度差异矩阵(neighboring gray tone difference matrix, NGTDM)、灰度依赖矩阵(gray level dependent matrix, GLDM)和灰度运行长度矩阵(gray level run length matrix, GLRLM)。在对小波特征进行提取前,需要使用三维小波变换对原始图像进行分解,分解后得到8个新的图像,然后再提取一阶特征、形状特征和纹理特征。本研究中对每个特征进行Z-score标准化处理。

1.3.3 特征选择

       在所有影像组学特征中,通过两种方法进行特征选择。首先,采用Pearson相关系数计算每个影像组学特征在预测ⅠB和ⅡA期宫颈癌的P值,然后,将P<0.05的特征作为显著的预测因子。其次,使用最小绝对收缩和选择算子(the least absolute shrinkage and selection operator, LASSO)回归方法,根据最佳Alpha值选择特征。LASSO路径如图3所示。

图3  使用最小的绝对收缩和选择算子回归方法选择影像组学特征。
Fig. 3  Radiomics features are selected using the minimum absolute shrinkage and selection operator (LASSO) regression method.

1.3.4 临床模型建立

       本研究共纳入8个候选临床特征,包括年龄、月经情况、HPV、SCCA、RBC、WBC和PLT。将显著变量(P<0.05)纳入模型,建立预测宫颈癌分期的逻辑回归模型。

1.3.5 影像组学评分和联合模型的开发与验证

       通过特征选择筛选出最具预测价值的影像组学特征,构建逻辑回归模型,根据其截距和系数计算出影像组学评分(Radscore;公式1)。基于Radscore和显著临床特征建立联合模型。

       其中:wavelet_glszm_wavelet-hhh-GrayLevelNonUniformity为经高频带分解后灰度大小区域矩阵纹理特征;original_shape_Maximum2DDiameterSlice为最大2D直径(切片)形态特征;original_shape_Maximum3DDiameter为最大3D直径形态特征;original_shape_VoxelVolume为体素体积形态特征。

1.3.6 影像组学列线图的应用

       利用显著临床特征和Radscore构建预测宫颈癌分期列线图。绘制受试者工作特征(receiver operating characteristic, ROC)曲线、计算曲线下面积(area under the curve, AUC)、敏感度和特异度评估模型预测效能,利用DCA评价列线图的临床价值。

图1  影像组学分析流程图。
Fig. 1  The flowchart of the radiomics analysis.
图2  ⅠB期宫颈癌患者的T2WI图像和感兴趣区(ROI)的分割和膨胀过程,红色表示瘤内区域,彩色环表示瘤周区域,每个环代表1 mm宽的膨胀距离。
Fig. 2  Example of the image segmentation and expansion process of the T2WI images of stage ⅠB patient and ROI. The red region represents the tumoral region, colorful rings indicate peritumoral regions, with each ring represents one millimeters wide dilation.

1.4 统计学分析

       本研究中所有统计分析均在SPSS 26.0软件进行。采用Kolmogorov-Smirnov检验评估连续性变量的正态分布,符合正态分布的数据比较采用独立样本t检验,不符合正态分布的数据比较采用Mann-Whitney U检验,分类数据比较采用卡方检验。不同模型的ROC采用DeLong检验比较。P<0.05为差异有统计学意义。

2 结果

2.1 临床基线资料

       本研究纳入两个中心共120例早期宫颈癌患者病例,患者临床资料详见表1。在训练组和验证组中,ⅠB期和ⅡA期患者RBC和肿瘤最大径差异有统计学意义(P=0.001),HPV在训练组差异有统计学意义(P=0.014),而在验证组差异无统计学意义(P=0.128)。

表1  患者临床资料
Tab. 1  Clinical characteristics of the enrolled patients

2.2 影像组学预测效能

       基于不同瘤内瘤周筛选影像组学特征(表2),不同瘤周经筛选分别保留了3~10个不等特征,大部分是小波特征。基于瘤周3 mm共提取到6个组学特征,包括1个原始特征,5个小波特征,小波特征包括4个纹理特征,1个一阶特征。基于瘤周3 mm和瘤周5 mm的影像组学模型诊断效能在训练组和验证组表现均较佳,训练组AUC分别为0.972和0.975,验证组AUC分别为0.857和0.854,ROC曲线见图4。DeLong检验结果显示瘤内与不同瘤周ROC之间差异有统计学意义。

图4  T2WI瘤内结合瘤周3 mm和5 mm的影像组学特征的ROC曲线。4A:训练组;4B:验证组。ROC:受试者工作特征。
Fig. 4  The ROC curve of radiomics features of intra- and peritumoral 3 mm and 5 mm expansion in T2WI. 4A: Training group; 4B: Test group. ROC: receiver operating characteristic.
表2  T2WI图像瘤内瘤周组学特征预测ⅡA期性能
Tab. 2  Prediction performance of stage ⅡA of cervical cancer in the intra- and peritumoral features in T2WI images

2.3 模型预测的性能与验证

       不同模型预测ⅡA期宫颈癌诊断效能分析显示(表3):在训练组中,影像组学模型诊断效能最高,AUC为0.972,敏感度和特异度分别为82.8%和92.3%;在验证组中,联合模型诊断效能最佳,AUC为0.939,敏感度和特异度分别为77.8%和82.9%。临床模型在训练组和验证组诊断效能较差,AUC分别为0.940和0.847。DeLong检验结果显示联合模型与临床模型ROC差异有统计学意义(P=0.018),联合模型与影像组学模型ROC差异有统计学意义(P=0.025),临床模型与影像组学模型ROC差异无统计学意义(P=0.196)。

表3  各模型在训练组和验证组诊断效能比较
Tab. 3  Comparison of diagnostic performance between different models in the training and validation groups

2.4 列线图的增益结果

       基于瘤周3 mm的Radscore、肿瘤最大径和RBC构建列线图(图5),在训练组和验证组中,列线图在预测ⅡA期方面表现出了较好的性能,AUC分别为0.952和0.939。校准曲线(图6)和DCA(图7)显示在阈值概率41%到100%,列线图在预测ⅡA期宫颈癌方面显示更好的应用价值。

图5  基于RBC和肿瘤最大径的临床资料和Radscore构建的列线图。Radscore表示由瘤内结合瘤周3 mm提取的6个影像组学特征及其相应系数的线性和。每个预测变量的值可以根据列线图顶部的“point”转换为一个风险评分,将每个个体的风险得分相加后得到一个“Total point”,对应列线图底部的预测概率。RBC:红细胞计数;Radscore:影像组学评分。
Fig. 5  Nomogram was developed by integrating the clinical factors and combined Radscore in the training group. Radscore represents the linear sum of the six radiomic features of intratumoral plus peritumoral 3 mm expansion and their corresponding coefficients. The value of each predictor can be converted into a risk score according to the “Points” at the top of the nomogram. After adding up the individual risk score of these predictors in “Total Points”, the corresponding prediction probability at the bottom of the nomogram can be obtained. RBC: red blood cell; Radscore: radiomic score.
图6  影像组学列线图的校准曲线(训练组)。X轴表示模态图预测ⅠB和ⅡA期的概率,Y轴表示实际ⅠB和ⅡA期概率,绿色对角线表示完美预测,蓝色对角线表示列线图的实际预测。
图7  影像组学列线图的决策曲线分析(训练组)。X轴表示阈值概率,Y轴表示净收益。
Fig. 6  Calibration curve of the radiomics nomogram in the training group. X axis shows the predicted probability of stage ⅠB and ⅡA cervical cancer, Y axis represents the actual probability of stage ⅠB and ⅡA cervical cancer, green diagonal indicates the perfect prediction, blue diagonal indicates the actual prediction of the radiomics nomogram.
Fig. 7  Decision curve analysis of radiomics nomogram in the training group. X axis represents the threshold probability, Y axis represents the net yield.

3 讨论

       本研究利用瘤内瘤周影像组学特征建立组学预测模型,并将预测效能最佳的瘤内瘤周3 mm的组学特征结合临床资料构建列线图,探究其在术前预测ⅠB期和ⅡA期宫颈癌阴道受侵的潜在价值。结果显示,在训练组和验证组中,列线图的AUC分别为0.952和0.939,具有较好的预测效能。因此,本研究揭示了瘤周组学特征能够为影像组学预测早期宫颈癌分期提供额外的价值,不仅为临床医生帮助有生育意愿的早期宫颈癌患者选择宫颈切除等保留生育能力的治疗方式提供依据,对患者手术及后续放化疗方案的制订也起着指导性作用[17, 18]

3.1 不同检查方法在评估宫颈癌分期的应用价值分析

       临床对阴道是否受侵的诊断,妇科检查最常用通过临床医生肉眼评估肿瘤的大小和观察穹隆变浅或消失来判定阴道是否受侵,但对于轻度糜烂、充血、炎症样改变的病灶,仅凭阴道妇科检查不能判断阴道壁是否受侵[19, 20]。MRI具有软组织分辨率高、无辐射等优点[21, 22]。可通过宫颈自身信号变化了解肿瘤侵犯宫颈程度,及其是否侵至阴道或宫体,为临床宫颈癌分期提供依据,但和妇科检查一样,对于阴道微小病灶和黏膜的表浅浸润,其诊断效率和敏感性并不高,ZHANG等[23]报道了4729例ⅠB、ⅡA期患者影像评估分期准确率仅为83%、60%。超声是一种无创、方便的成像技术,在评估宫颈癌患者方面也具有重要的作用,但不能三维地显示宫颈及阴道正常结构和可疑病变,在检测宫颈癌阴道侵袭方面的效能低于MRI[24],因此较少被用于诊断阴道病变。阴道镜具有图像放大和醋酸试验、碘试验的图像特点,能提高临床对阴道病变的检出率和诊断的准确率[25, 26],但当宫颈或阴道出血、感染时,阴道、宫颈暴露不充分,会影响阴道镜检查结果的判读。与以往检查方式的不同,本研究利用影像组学技术,创新性地将瘤周组学特征纳入预测模型中,结果表明,瘤周组学对提高宫颈癌分期的诊断率具有潜在价值。

3.2 瘤内瘤周组学技术预测宫颈癌分期的效能分析

       影像组学是一种临床诊断肿瘤良恶性和分期的非侵入性方法[27, 28],近年来,成为临床研究的热点,近期有研究报道瘤周区域也提供了大量肿瘤异质性信息[11, 29]。XU等[30]研究发现基于预处理乳腺癌动态增强磁共振成像中瘤内瘤周的影像组学纹理分析成功地预测了新辅助化疗的病理完全反应,证实瘤内联合瘤周的影像组学模型效能优于单区域,可以提高预测准确率,增加模型的稳定性。在本研究中,从不同瘤周提取了170~267个不等有效影像组学特征,比仅从瘤内提取出的94个有效影像组学特征多得多,表明瘤周含有更多肿瘤异质性信息。其中来自瘤周3 mm和5 mm影像组学特征具有最好预测价值。这与HUANG等[31]研究的结果相符。从病理角度分析,肿瘤细胞倾向于从原发肿瘤迁移到瘤周区域,从而导致MRI图像上病灶形态学改变[32]。另外,不同的瘤周特征经选择保留了3~10个不等组学特征,大部分是小波特征。易芹芹等[33]在使用MRI影像组学模型预测直肠癌化疗效果时发现,超过90%的特征为小波特征,这与本试验的结果相一致。小波特征是从8个空间域反映肿瘤信息[34, 35, 36],这可能也是影像医生难以通过肉眼来确定阴道是否受侵犯的原因。基于以上研究证明经小波变换后的组学特征更能反映肿瘤的异质性。

3.3 列线图临床应用价值分析

       列线图是一种对提高疾病诊断准确率有用的辅助工具[37],利用列线图预测肿瘤恶性程度和预后评价是影像组学技术发展的一个持续趋势[38]。本研究利用Radscore与临床资料开发并验证了一个预测宫颈癌分期的个体化列线图。结果显示,在训练组和验证组中均表现出了良好的预测效能,AUC分别为0.952和0.939,而基于肿瘤最大径和RBC建立的临床模型在训练组和验证组中AUC仅为0.940和0.847,表明列线图比单独临床模型具有更好的预测效果,这与LI等[39]的研究相似,他们认为MRI影像组学列线图在术前预测宫颈癌淋巴—血管间隙浸润具有很好应用价值。同时DCA结果表明,在大量阈值概率范围内,列线图对宫颈癌分期的预测具有良好的净效益。最终,本研究通过将影像组学特征与术前临床危险因素相结合开发列线图,为临床医生提供了一个便捷的术前评估早期宫颈癌分期的视觉工具[40, 41, 42],可以为患者术前生成一个个体化的宫颈癌分期的预测概率。

3.4 本研究局限性及展望

       第一,虽然本研究是多中心研究,但考虑到受试者数量有限,可能存在模型的欠拟合现象,计划在后续的研究中加入更多数据进一步验证模型的预测效能和稳定性;第二,由于术前增强MRI图像和扩散加权成像(diffusion-weighted imaging, DWI)序列的图像有限,本研究只对平扫图像进行了组学分析,这可能低估了组学分析的临床价值,后续研究将纳入更多增强图像和DWI序列图像加以完善;第三,保留子宫的手术不是临床常规方法,标准也存在一定争议,在未来,影像组学模型和列线图对选择子宫保留患者的实际应用价值仍需进一步验证;第四,本研究中所有ROI均是由一位诊断经验丰富的影像科医生手动分割和另一名验证经验丰富的影像科医生验证,没有做一致性分析,在后续研究中进行一致性分析和进一步探索机器学习的宫颈肿瘤全自动分割来提供更精确的数据。

4 结论

       综上所述,基于临床资料和瘤内瘤周组学特征建立的列线图,可以很好的术前预测ⅠB期和ⅡA期宫颈癌,帮助影像科医师提高宫颈癌分期的诊断准确率,为患者制订更精准的个体化治疗方案。

[1]
SUNG H, FERLAY J, SIEGEL R L, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA Cancer J Clin, 2021, 71(3): 209-249. DOI: 10.3322/caac.21660.
[2]
ZHANG X J, LV Z F, XU X X, et al. Comparison of adenocarcinoma and adenosquamous carcinoma prognoses in Chinese patients with FIGO stage IB-ⅡA cervical cancer following radical surgery[J]. BMC Cancer, 2020, 20(1): 664. DOI: 10.1186/s12885-020-07148-x.
[3]
SIEGEL R L, MILLER K D, JEMAL A. Cancer statistics, 2018[J]. CA Cancer J Clin, 2018, 68(1): 7-30. DOI: 10.3322/caac.21442.
[4]
周晖, 刘昀昀, 罗铭, 等. 《2023 NCCN子宫颈癌临床实践指南(第1版)》解读[J]. 中国实用妇科与产科杂志, 2023, 39(2): 189-196. DOI: 10.19538/j.fk2023020115.
ZHOU H, LIU Y Y, LUO M, et al. Interpretation of 2023 NCCN clinical practice guidelines for cervical cancer (1st edition)[J]. Chin J Pract Gynecol Obstet, 2023, 39(2): 189-196. DOI: 10.19538/j.fk2023020115.
[5]
COSTALES A, MICHENER C, ESCOBAR-RODRIGUEZ P F. Radical trachelectomy for early stage cervical cancer[J]. Curr Treat Options Oncol, 2018, 19(12): 63-75. DOI: 10.1007/s11864-018-0591-4.
[6]
MATSUO K, CHEN L, MANDELBAUM R S, et al. Trachelectomy for reproductive-aged women with early-stage cervical cancer: minimally invasive surgery versus laparotomy[J/OL]. Am J Obstet Gynecol, 2019, 220(5): 469.e1-469.e13 [2024-04-08]. http://www.sciencedirect.com/science/journal/00029378. DOI: 10.1016/j.ajog.2019.02.038.
[7]
WU M L, WU J W, HUANG L C, et al. Comparison of contrast-enhanced ultrasonography and magnetic resonance imaging in the evaluation of tumor size and local invasion of surgically treated cervical cancer[J]. Abdom Radiol, 2022, 47(8): 2928-2936. DOI: 10.1007/s00261-022-03558-6.
[8]
SHI J X, DONG Y, JIANG W Y, et al. MRI-based peritumoral radiomics analysis for preoperative prediction of lymph node metastasis in early-stage cervical cancer: a multi-center study[J]. Magn Reson Imaging, 2022, 88: 1-8. DOI: 10.1016/j.mri.2021.12.008.
[9]
ZHANG Y J, WU C, DU J L, et al. Prediction of recurrence risk factors in patients with early-stage cervical cancers by nomogram based on MRI handcrafted radiomics features and deep learning features: a dual-center study[J]. Abdom Radiol, 2024, 49(1): 258-270. DOI: 10.1007/s00261-023-04125-3.
[10]
XU X, ZHANG H L, LIU Q P, et al. Radiomic analysis of contrast-enhanced CT predicts microvascular invasion and outcome in hepatocellular carcinoma[J]. J Hepatol, 2019, 70(6): 1133-1144. DOI: 10.1016/j.jhep.2019.02.023.
[11]
XIAO M L, FU L, WEI Y, et al. Intratumoral and peritumoral MRI radiomics nomogram for predicting parametrial invasion in patients with early-stage cervical adenocarcinoma and adenosquamous carcinoma[J]. Eur Radiol, 2024, 34(2): 852-862. DOI: 10.1007/s00330-023-10042-2.
[12]
ZHANG X F, WU H Y, LIANG X W, et al. Deep-learning-based radiomics of intratumoral and peritumoral MRI images to predict the pathological features of adjuvant radiotherapy in early-stage cervical squamous cell carcinoma[J]. BMC Womens Health, 2024, 24(1): 182-194. DOI: 10.1186/s12905-024-03001-6.
[13]
XIA X M, LI D D, DU W, et al. Radiomics based on nomogram predict pelvic lymphnode metastasis in early-stage cervical cancer[J]. Diagnostics, 2022, 12(10): 2446-2458. DOI: 10.3390/diagnostics12102446.
[14]
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.
[15]
ZHU Y, LU M, DAI K J, et al. Comparison between 360° 3-D transvaginal ultrasound and magnetic resonance imaging for assessment of vaginal invasion in cervical cancer: a preliminary report[J]. Ultrasound Med Biol, 2021, 47(8): 2250-2257. DOI: 10.1016/j.ultrasmedbio.2021.04.013.
[16]
WU J J, XIA Y W, WANG X C, et al. uRP: an integrated research platform for one-stop analysis of medical images[J/OL]. Front Radiol, 2023, 3: 1153784 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/37492386/. DOI: 10.3389/fradi.2023.1153784.
[17]
KONG T W, KIM J, SON J H, et al. Preoperative nomogram for prediction of microscopic parametrial infiltration in patients with FIGO stage IB cervical cancer treated with radical hysterectomy[J]. Gynecol Oncol, 2016, 142(1): 109-114. DOI: 10.1016/j.ygyno.2016.05.010.
[18]
SHRESTHA P, POUDYAL B, YADOLLAHI S, et al. A systematic review on the use of artificial intelligence in gynecologic imaging - Background, state of the art, and future directions[J]. Gynecol Oncol, 2022, 166(3): 596-605. DOI: 10.1016/j.ygyno.2022.07.024.
[19]
ELKADY R M. Radiomics analysis in evaluation of cervical cancer: a further step on the road[J]. Acad Radiol, 2022, 29(8): 1141-1142. DOI: 10.1016/j.acra.2022.02.018.
[20]
WU Q X, WANG S, CHEN X, et al. Radiomics analysis of magnetic resonance imaging improves diagnostic performance of lymph node metastasis in patients with cervical cancer[J]. Radiother Oncol, 2019, 138: 141-148. DOI: 10.1016/j.radonc.2019.04.035.
[21]
林宝金, 龙先凤, 吴朝霞, 等. 基于MR-T2WI的深度学习与影像组学联合临床特征预测宫颈癌淋巴脉管间隙浸润[J]. 磁共振成像, 2024, 15(3): 130-136. DOI: 10.12015/issn.1674-8034.2024.03.021.
LIN B J, LONG X F, WU Z X, et al. Predicting lymph-vascular space invasion in cervical cancer based on MR-T2WI with deep learning and radiomic features combined with clinical features[J]. Chin J Magn Reson Imag, 2024, 15(3): 130-136. DOI: 10.12015/issn.1674-8034.2024.03.021.
[22]
BAE H S, KIM Y J, LIM M C, et al. Predictors of radiation field failure after definitive chemoradiation in patients with locally advanced cervical cancer[J]. Int J Gynecol Cancer, 2016, 26(4): 737-742. DOI: 10.1097/IGC.0000000000000662.
[23]
ZHANG Z H, WAN X J, LEI X Y, et al. Intra- and peri-tumoral MRI radiomics features for preoperative lymph node metastasis prediction in early-stage cervical cancer[J/OL]. Insights Imaging, 2023, 14(1): 65 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/37060378/. DOI: 10.1186/s13244-023-01405-w.
[24]
ZHENG R R, CAI M T, LAN L, et al. An MRI-based radiomics signature and clinical characteristics for survival prediction in early-stage cervical cancer[J/OL]. Br J Radiol, 2022, 95(1129): 20210838 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/34797703/. DOI: 10.1259/bjr.20210838.
[25]
KIDO A, NAKAMOTO Y. Implications of the new FIGO staging and the role of imaging in cervical cancer[J/OL]. Br J Radiol, 2021, 94(1125): 20201342 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/33989030/. DOI: 10.1259/bjr.20201342.
[26]
JHA A K, MITHUN S, SHERKHANE U B, et al. Systematic review and meta-analysis of prediction models used in cervical cancer[J/OL]. Artif Intell Med, 2023, 139: 102549 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/37100501/. DOI: 10.1016/j.artmed.2023.102549.
[27]
SONG M F, LIN J, SONG F Z, et al. The value of MR-based radiomics in identifying residual disease in patients with carcinoma in situ after cervical conization[J/OL]. Sci Rep, 2020, 10(1): 19890 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/33199785/. DOI: 10.1038/s41598-020-76853-1.
[28]
陈志军, 邓雪莉, 陈慧. 术前外周血SII、NLR、PLR对宫颈癌淋巴结转移的评估价值[J]. 中国妇幼健康研究, 2023, 34(4): 33-39. DOI: 10.3969/j.issn.1673-5293.2023.04.006.
CHEN Z J, DENG X L, CHEN H. Predictive value of peripheral blood SⅡ, NLR and PLR before operation for lymph node metastasis in cervical cancer[J]. Chin J Woman Child Health Res, 2023, 34(4): 33-39. DOI: 10.3969/j.issn.1673-5293.2023.04.006.
[29]
BRAMAN N M, ETESAMI M, PRASANNA P, et al. Intratumoral and peritumoral radiomics for the pretreatment prediction of pathological complete response to neoadjuvant chemotherapy based on breast DCE-MRI[J/OL]. Breast Cancer Res, 2017, 19(1): 57 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/28521821/. DOI: 10.1186/s13058-017-0846-1.
[30]
XU H, WANG A, ZHANG C, et al. Intra- and peritumoral MRI radiomics assisted in predicting radiochemotherapy response in metastatic cervical lymph nodes of nasopharyngeal cancer[J/OL]. BMC Med Imaging, 2023, 23(1): 66 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/37254101/. DOI: 10.1186/s12880-023-01026-1.
[31]
HUANG Q H, DENG B D, WANG Y C, et al. Reduced field-of-view DWI-derived clinical-radiomics model for the prediction of stage in cervical cancer[J/OL]. Insights Imaging, 2023, 14(1): 18 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/36701003/. DOI: 10.1186/s13244-022-01346-w.
[32]
WANG T, GAO T T, GUO H, et al. Preoperative prediction of parametrial invasion in early-stage cervical cancer with MRI-based radiomics nomogram[J]. Eur Radiol, 2020, 30(6): 3585-3593. DOI: 10.1007/s00330-019-06655-1.
[33]
易芹芹, 周宙, 罗燕, 等. 基于术前MRI影像组学及临床特征的早期宫颈癌中危因素预测模型构建[J]. 磁共振成像, 2022, 13(4): 124-127, 136. DOI: 10.12015/issn.1674-8034.2022.04.024.
YI Q Q, ZHOU Z, LUO Y, et al. Construction of prediction model of intermediate risk factors for early cervical cancer based on preoperative MRI radiomics and clinical features[J]. Chin J Magn Reson Imag, 2022, 13(4): 124-127, 136. DOI: 10.12015/issn.1674-8034.2022.04.024.
[34]
GUIDA F, KIDMAN R, FERLAY J, et al. Global and regional estimates of orphans attributed to maternal cancer mortality in 2020[J]. Nat Med, 2022, 28(12): 2563-2572. DOI: 10.1038/s41591-022-02109-2.
[35]
TERZIC M, MAKHADIYEVA D, BILA J, et al. Reproductive and obstetric outcomes after fertility-sparing treatments for cervical cancer: current approach and future directions[J/OL]. J Clin Med, 2023, 12(7): 2614 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/37048696/. DOI: 10.3390/jcm12072614.
[36]
朱彦玲, 章文华, 刘蓓蓓, 等. 阴道镜检查判断子宫颈癌阴道受侵的初步探讨[J]. 中华妇产科杂志, 2020, 55(5): 322-326. DOI: 10.3760/cma.j.cn112141-20200211-00093.
ZHU Y L, ZHANG W H, LIU B B, et al. Preliminary study of the clinical value of colposcopy in diagnosing vagina invasion in cervical cancer[J]. Chin J Obstet Gynecol, 2020, 55(5): 322-326. DOI: 10.3760/cma.j.cn112141-20200211-00093.
[37]
YU W K, XU H E, CHEN F J, et al. Development and validation of a radiomics-based nomogram for the prediction of postoperative malnutrition in stage IB1-ⅡA2 cervical carcinoma[J/OL]. Front Nutr, 2023, 10: 1113588 [2024-04-22]. https://pubmed.ncbi.nlm.nih.gov/36819703/. DOI: 10.3389/fnut.2023.1113588.
[38]
SI L P, ZHONG J Y, HUO J Y, et al. Deep learning in knee imaging: a systematic review utilizing a Checklist for Artificial Intelligence in Medical Imaging (CLAIM)[J]. Eur Radiol, 2022, 32(2): 1353-1361. DOI: 10.1007/s00330-021-08190-4.
[39]
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.
[40]
RASPAGLIESI F, DITTO A, QUATTRONE P, et al. Prognostic factors in microinvasive cervical squamous cell cancer: long-term results[J]. Int J Gynecol Cancer, 2005, 15(1): 88-93. DOI: 10.1111/j.1048-891x.2005.15009.x.
[41]
乔志伟, 王安娜, 王纯雁, 等. 磁共振成像检查对早期宫颈癌分期价值研究[J]. 中国实用妇科与产科杂志, 2015, 31(3): 247-250. DOI: 10.7504/fk2015020117.
QIAO Z W, WANG A N, WANG C Y, et al. The value of MRI in staging of cervical cancer[J]. Chin J Pract Gynecol Obstet, 2015, 31(3): 247-250. DOI: 10.7504/fk2015020117.
[42]
NANTHAMONGKOLKUL K, HANPRASERTPONG J. Longer waiting times for early stage cervical cancer patients undergoing radical hysterectomy are associated with diminished long-term overall survival[J]. J Gynecol Oncol, 2015, 26(4): 262-269. DOI: 10.3802/jgo.2015.26.4.262.

上一篇 酰胺质子转移成像与动态对比增强MRI评估宫颈癌神经侵犯的价值
下一篇 阿尔茨海默病患者大脑形态学及结构协变网络的改变
  
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