分享:
分享到微信朋友圈
X
临床研究
基于磁共振T2WI纹理分析对直肠癌淋巴结转移的预测价值评估
李国强 柯炜炜 孙祥林 魏宇泽 卢再鸣

Cite this article as: Li GQ, Ke WW, Sun XL, et al. Predictive value of MRI T2WI texture analysis for lymph node metastasis in rectal cancer[J]. Chin J Magn Reson Imaging, 2022, 13(7): 42-47.本文引用格式:李国强, 柯炜炜, 孙祥林, 等. 基于磁共振T2WI纹理分析对直肠癌淋巴结转移的预测价值评估[J]. 磁共振成像, 2022, 13(7): 42-47. DOI:10.12015/issn.1674-8034.2022.07.008.


[摘要] 目的 构建基于直肠癌患者T2WI纹理特征参数和临床指标的联合预测模型评价直肠癌是否存在术前淋巴结转移(lymph node metastasis, LNM)。材料与方法 本研究回顾性分析了112例病理诊断为直肠癌且接受直肠癌根治性切除术及淋巴结清扫术的患者的T2WI图像、血清肿瘤标志物及基本临床资料。所有患者按7∶3比例随机分为训练组和验证组,分别用于预测模型的训练和验证。在T2WI图像上手动勾画直肠癌病灶和目标淋巴结的感兴趣区域(region of interest, ROI)。应用人工智能软件自动化提取ROI的纹理参数,并从中筛选出能够鉴别LNM的纹理参数。利用logistic回归分析分别构建基于肿瘤组织纹理参数和目标淋巴结纹理参数的影像组学预测模型、基于患者临床指标的临床预测模型以及纹理参数和临床指标相结合的联合预测模型。采用受试者工作特征(receiver operating characteristic, ROC)曲线下面积(area under the curve, AUC)来评估不同模型鉴别术前LNM的诊断效能。运用DeLong检验对比各预测模型AUC差异。通过决策曲线分析(decision curve analysis, DCA)对各预测模型的临床获益度进行评估。P<0.05为差异有统计学意义。结果 从每例患者的肿瘤组织及目标淋巴结T2WI图像中各提取出401个纹理特征,经筛选,肿瘤组织保留7个纹理参数,目标淋巴结保留6个纹理参数。根据上述纹理参数构建的目标淋巴结纹理分析预测模型训练组AUC值为0.881,敏感度为86.67%,特异度为81.25%;验证组AUC值为0.795,敏感度为92.31%,特异度为66.67%。肿瘤组织纹理分析预测模型训练组AUC值为0.844,敏感度为80.00%,特异度为79.17%;验证组AUC值为0.897,敏感度为84.62%,特异度为90.48%。最终结合纹理参数、目标淋巴结短径/长径、患者血清CA19-9水平构建的联合预测模型诊断效能明显优于其他模型,且差异存在统计学意义(训练组AUC值为0.978,敏感度和特异度分别为93.33%和91.67%,验证组AUC值为0.897敏感度为84.62%,特异度为90.48%,P<0.05)。结论 利用直肠T2WI纹理特征联合临床指标构建的模型可以有效地预测LNM,为临床个体化治疗提供帮助。
[Abstract] Objective To construct a prediction model based on T2WI texture features and clinical indicators to predict preoperative lymph node metastasis before rectal cancer.Materials and Methods This study retrospectively analyzed T2WI images, serum tumor markers and basic clinical data of 112 patients who underwent radical resection and lymph node dissection of rectal cancer because of pathological diagnosis of rectal cancer. All patients were randomly divided into training group and validation group with a ratio of 7∶3 to train and validate prediction models, respectively. Region of interest (ROI) of rectal cancer lesions and target lymph nodes were manually delineated on T2WI images. The texture parameters used to identify lymph node metastasis were automatically extracted using artificial intelligence software logistic regression analyses were used to construct two prediction models based on tumor tissue texture parameters and target lymph node texture parameters, a clinical prediction model based on patient clinical indicators, and a combined prediction model combining texture parameters and clinical indicators, respectively. The area under the receiver operating characteristic (AUCs) curves were used to evaluate the diagnostic performances of different models. The DeLong tests were used to compare the AUC differences between prediction models. The net clinical benefit of each prediction model was evaluated by decision curve analysis (DCA). Statistical significance was set at P<0.05.Results Four hundred and one texture features were extracted from the T2WI images of each ROI. After screening, 7 texture parameters of tumor tissue and 6 texture parameters of the target lymph node were selected for model building. The AUC of the target lymph node texture analysis prediction model in the training group was 0.881, with a sensitivity of 86.67% and a specificity of 81.25%; the AUC of the validation group was 0.795, with a sensitivity of 92.31% and specificity of 66.67%. The AUC of the tumor tissue texture analysis prediction model in the training group was 0.844, with a sensitivity of 80.00% and a specificity of 79.17%; the AUC of the validation group was 0.897, with a sensitivity of 84.62% and a specificity of 90.48%. The combined prediction model constructed by combining texture parameters, the ratio of short to long diameter of the target lymph nodes and the serum CA19-9 level of the patients gets the best performance among the models (AUC of the training group was 0.978 with the sensitivity and specificity were 93.33% and 91.67%, respectively, and the AUC of the validation group was 0.897 with the sensitivity was 84.62%, the specificity was 90.48%, P<0.05).Conclusions The texture features of rectal T2WI images combined with clinical indexes can be used to construct an effective model for predicting lymph node metastasis and provide help for clinical individualized treatment.
[关键词] 直肠癌;磁共振成像;纹理分析;淋巴结转移;预测
[Keywords] rectal cancer;magnetic resonance imaging;texture analysis;lymph node metastasis;prediction

李国强    柯炜炜    孙祥林    魏宇泽    卢再鸣 *  

中国医科大学附属盛京医院放射科,沈阳 110000

卢再鸣,E-mail:luzm@sj-hospital.org

作者利益冲突声明:全体作者均声明无利益冲突。


基金项目: 国家自然科学基金 81771947
收稿日期:2022-02-08
接受日期:2022-06-24
中图分类号:R445.2  R735.37 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2022.07.008
本文引用格式:李国强, 柯炜炜, 孙祥林, 等. 基于磁共振T2WI纹理分析对直肠癌淋巴结转移的预测价值评估[J]. 磁共振成像, 2022, 13(7): 42-47. DOI:10.12015/issn.1674-8034.2022.07.008.

       直肠癌是最常见的消化道肿瘤之一,其发病率逐年升高,是世界范围内第三大癌症死亡的原因,严重危害人类生命与健康[1, 2]。淋巴结转移(lymph node metastasis, LNM)是直肠癌最常见的转移方式之一,评估直肠癌患者淋巴结是否转移在治疗方案的制订及预后评估等方面起着至关重要的作用,由于术前淋巴结评估的不确定性,各个国家治疗直肠癌的方案也不尽相同[3, 4, 5, 6, 7],尽管传统影像学可从形态、大小等方面初步评估淋巴结性质,但结果仍存在偏差,目前尚未有一种能够在术前准确且无创地评估直肠癌淋巴转移情况的方法[8]

       纹理分析技术从医学图像中高通量提取定量信息,通过无创的方法更深层次的评价肿瘤内部结构,弥补了传统影像学肉眼观察的局限性,在各类肿瘤相关领域的研究上展现出了巨大的潜力[9, 10]。作为直肠MRI的常规序列,T2WI在显示病灶结构特征方面较T1WI更具优势,可操作性更好,同时扫描成本低于增强序列,且无需考虑个体循环差异及药物影响,在直肠癌患者随访中应用广泛,故本研究拟基于直肠癌患者术前T2WI序列图像的纹理特征,结合直肠患者术前临床指标构建可准确预测LNM的联合模型,从而为临床制订或调整治疗计划、评估患者预后提供更好的帮助。

1 材料与方法

1.1 研究对象

       本研究通过中国医科大学附属盛京医院医学伦理委员会批准,免除受试者知情同意,批准文号:2020PS416K。回顾性分析2015年1月至2020年12月在中国医科大学附属盛京医院同一台MRI设备上进行直肠扫描并通过病理确诊为直肠癌的患者临床资料及影像学资料,所有资料均通过医院病历系统获得。纳入标准:(1)经病理证实为直肠癌;(2)MRI检查前未接受任何放化疗,检查后进行直肠癌根治性切除术以及淋巴结节清扫;(3)病理及相关临床资料完整;(4)无其他恶性肿瘤病史。排除标准:(1)图像存在伪影,影响观察或难以勾画感兴趣区域(region of interest, ROI);(2)未进行手术或仅局部切除未能获取淋巴结病理情况。研究初始纳入172例患者,因手术前行放疗或化疗排除27例、图像存在伪影排除22例、未行根治性切除术及淋巴结清扫术排除11例。最终本次研究纳入112例患者,年龄(63±8)岁,其中男69例,女43例,LNM患者(LNM+组)43例,非LNM患者(LNM-组)69例。纳入对象被按照7∶3的比例随机分为训练组和验证组,训练组78例(LNM+:30例,LNM-:48例),验证组34例(LNM+:13例,LNM-:21例)。

1.2 MRI扫描方法

       所有患者术前应用美国GE公司3.0 T Signa HDx双梯度磁共振扫描仪进行检查,使用8通道体部相控阵线圈。患者术前4 h禁食水,48 h内不行任何其他对比增强影像检查,患者取仰卧位,平静呼吸。扫描范围从结直肠末段至肛门。T2WI序列具体扫描参数为:TR 2981 ms,TE 90 ms,FOV 240 mm×240 mm,NEX 2,矩阵224×224,层厚8 mm,层间距3 mm,翻转角15°。扫描时间为150 s左右。

1.3 观察指标

       术前记录直肠癌患者性别、年龄等基本临床信息,检测肿瘤标记物癌胚抗原(carcinoma embryonic antigen, CEA)及糖类抗原CA19-9水平并测量目标淋巴结在横断位及矢状位的T2WI序列图像上的最大径,计算出目标淋巴结短径/长径比值,汇总成为直肠癌患者临床指标。

       将直肠癌患者影像资料以DICOM形式导入A.K.软件(Artificial Intelligence Kit, GE Healthcare, China)进行图像标准化预处理,同时采用线性插值法进行重采样处理。将完成标准化预处理的图像导入至ITK-SNAP软件,由两名具有5年以上腹部MRI诊断经验的医师依据手术结果和病理资料勾画出直肠癌肿瘤及目标淋巴结所在层面的全部ROI。LNM+组淋巴结的勾画根据病理结果选取短径最大的转移性淋巴结作为研究对象,LNM-组则选取扫描平面内距离病灶最近且短径>3 mm的淋巴结进行勾画,勾画过程中避免肠内容物、血管等因素的干扰。勾画结果由另一名更高年资医师进行复核,勾画过程中如发生异议,则通过协商达成一致,勾画结果如图1所示。

       将分割的ROI文件再次导入A.K.软件,对ROI图像的一阶参数(如直方图参数)和高阶参数[灰度共生矩阵参数(gray-level co-occurrence matrix, GLCM)、行程矩阵参数(run-length matrix, RLM)]等纹理特征参数进行提取,最终从每例患者的肿瘤病灶和目标淋巴结T2WI序列图像中各提取出了401个纹理特征,纹理特征异常值用中位数代替。纹理特征筛选过程如下:第一步,运用ANOVA及Mann-Whitney U检验从肿瘤病灶纹理参数筛选出60个纹理参数,从目标淋巴结纹理参数中筛选出41个纹理参数;第二步,采用Spearman相关系数去除r≥0.9的高维特征冗余,进一步筛选出34个和20个纹理特征;第三步,采用十倍交叉验证的最小绝对收缩和选择算法(least absolute shrinkage and selection operator, LASSO)回归进一步降维(图2)。最终从肿瘤组织和目标淋巴结的401组数据中选分别取出7个和6个显著性较强的纹理特征,其中肿瘤组为:Correlation_AllDirection_offset1_SD、HaralickCorrelation_AllDirection_offset1_SD、GreyLevelNonuniformity_AllDirection_offset7_SD、HighGreyLevelRunEmphasis_AllDirection_offset7_SD、LongRunEmphasis_angle45_offset1、ShortRunEmphasis_AllDirection_offset7_SD、IntensityVariability;目标淋巴结组为:histogramEntropy、uniformity、GLCMEntropy_AllDirection_offset4、ClusterShade_AllDirection_offset1_SD、InverseDifferenceMoment_AllDirection_offset7_SD、HighGreyLevelRunEmphasis_AllDirection_offset1_SD。

图1  感兴趣区域(ROI)勾画示意图。1A~1B:男,69岁,术后病理诊断为直肠中分化腺癌伴多发淋巴结转移,图片红色区域分别为T2WI序列肿瘤病灶(1A)和确诊为转移性的淋巴结(1B)ROI勾画示意图。1C~1D:女,60岁,术后病理诊断为直肠中分化腺癌伴肠周淋巴结多发反应性增生,图片红色区域分别为T2WI序列肿瘤病灶(1C)和反应性增生淋巴结(1D)的ROI勾画示意图。
Fig. 1  Region of interest (ROI) delineation diagram. 1A, 1B: A 69-year-old male patient, with postoperative pathological diagnosis of rectal moderately differentiated adenocarcinoma with multiple lymph node metastasis. The red areas of the picture are the ROI delineations in T2WI sequence of the tumor lesions (1A) and the diagnosed metastatic lymph nodes (1B), respectively. 1C, 1D: A 60-year-old female patient, with postoperative pathological diagnosis of moderately differentiated adenocarcinoma of the rectum with multiple reactive hyperplasia of peri-intestinal lymph nodes, the red areas of the picture are the ROI delineations of the T2WI sequence tumor lesion (1C) and reactive hyperplastic lymph node (1D).
图2  LASSO回归纹理特征降维筛选曲线。2A:肿瘤组织组,Log(λ)=0.7;2B:目标淋巴结组,Log(λ)=0.8。
Fig. 2  LASSO dimensionality reduction regression analysis curve of texture features. 2A shows the tumor tissue group, Log (λ)=0.7; 2B is the target lymph node group, Log (λ)=0.8.

1.4 统计学方法

       采用SPSS(version 26.0 IBM, Armonk, NY)、MedCalc(Medcalc Softeware BVBA, Ostend, Belgium)及R软件[version 4.2.0(2022-04-22 ucrt)- “vigorous calisthenics”]进行数据处理。首先对两名放射医师勾画ROI的纹理特征参数及淋巴结测量值进行组内相关系数检验(intraclass correlation coefficient, ICC),以ICC≥0.75为标准保证两组数据具有一致性后,取两组数据平均值进行统计学分析。临床定量资料以均数±标准差(x¯±s)表示,若符合正态分布及方差齐性检验,则采用独立样本t检验比较数据间差异性,否则采用非参数检验;定性资料采用卡方检验进行对比分析。然后将具有显著差异的临床指标和纹理特征参数再进行线性回归分析,以方差膨胀因子(variance inflation factor, VIF)<5为标准,剔除存在多重共线性的参数,最后将保留的各参数纳入logistic回归分析,对比受试者工作特征(receiver operating characteristic, ROC)曲线下面积(area under the curve, AUC)值分析临床资料预测模型、纹理分析预测模型及联合预测模型的诊断效能。应用DeLong检验进一步比较各模型间AUC值。最后运用R软件进行决策曲线分析(decision curve analysis, DCA)评估预测模型在临床效用方面的净收益程度。置信区间设为95%,P<0.05为差异有统计学意义。

2 结果

2.1 临床指标

       如表1所示,LNM+组和LNM-组的直肠癌患者性别、年龄、肿瘤最大径数值差异无统计学意义(P>0.05),而直肠癌患者目标淋巴结短径/长径、肿瘤标记物CEA和糖类抗原CA19-9水平在两组之间的差异有统计学意义(P<0.05)。

表1  患者临床特征
Tab. 1  Clinical characteristics of the patients

2.2 预测模型的诊断表现

       logistic回归分析结果显示,目标淋巴结短径/长径[OR=40.503;95%置信区间(confidence interval, CI):5.063~324.001;P<0.05]、血清CA19-9水平(OR=1.044;95% CI:1.001~1.088;P<0.05)均为直肠癌LNM的独立预测因子,而肿瘤标记物CEA水平(OR=1.029;95% CI:0.994~1.065,P>0.05)在此过程中被排除,不作为独立预测因子。将目标淋巴结短径/长径与血清CA19-9水平相结合构建的临床资料预测模型训练组AUC值(图3A)为0.802(95% CI:0.696~0.884),敏感度为70.00%,特异度为81.25%;验证组AUC值(图3B)为0.696(95% CI:0.515~0.841),敏感度为92.31%,特异度为52.38%。以Youden指数计算的最佳临界值为0.434。

       基于目标淋巴结纹理特征参数构建的预测模型训练组AUC值(图3C)为0.881(95% CI:0.788~0.943),敏感度为86.67%,特异度为81.25%;验证组AUC值(图3D)为0.795(95% CI:0.711~0.894),敏感度为92.31%,特异度为66.67%。这表明基于淋巴结T2WI纹理特征构建的预测模型在术前诊断直肠癌LNM方面具有良好效能。

       基于肿瘤组织纹理构建的预测模型训练组AUC值(图3E)为0.844(95% CI:0.745~0.917),敏感度为80.00%,特异度为79.17%;验证组AUC值(图3F)为0.897(95% CI:0.745~0.975),敏感度为84.62%,特异度为90.48%。

图3  单一预测模型训练组和验证组受试者工作特征(ROC)曲线。3A、3B:分别为临床预测模型训练组、验证组ROC曲线;3C、3D:分别为目标淋巴结纹理预测模型训练组、验证组ROC曲线;3E、3F:分别为肿瘤组织纹理预测模型训练组、验证组ROC曲线。
Fig. 3  Receiver operating characteristic (ROC) curves of training group and validation group of single prediction models. 3A, 3B: ROC curves of the clinical prediction model training group and validation group, respectively; 3C, 3D: ROC curves of the lymph node texture analysis model training group and validation group, respectively; 3E, 3F: ROC curves of the tumor tissue texture analysis model training group and validation group, respectively.

2.3 联合预测模型性能分析

       为进一步提高预测模型诊断效能,将临床指标与纹理特征结合构建联合模型。首先通过线性回归分析,以VIF>5为标准认为变量间存在多重共线性,排除了淋巴结纹理参数中的histogramEntropy、uniformity、GLCMEntropy_AllDirection_offset4,其余变量均被保留,调整后R2=0.627,初步验证联合预测模型具有良好诊断效能。进一步将保留的纹理参数及临床指标进行logistic回归分析,构建的联合预测模型训练组AUC值(图4A)为0.978(95% CI:0.917~0.998),敏感度和特异度分别为93.33%和91.67%,验证组AUC值(图4B)为0.897(95% CI:0.745~0.975),敏感度为84.62%,特异度为90.48%。经DeLong检验(表2),最终建立的联合预测模型的AUC大于各单一预测模型的AUC,差异具有统计学意义(P<0.05),而单一预测模型间的AUC值差异不具有统计学意义(P>0.05)。

图4  联合预测模型训练组受试者工作特征(ROC)曲线(4A)、联合预测模型验证组ROC曲线(4B)、预测模型训练组间ROC曲线对比分析(4C)和联合预测模型的校准曲线(4D)。
Fig. 4  The receiver operating characteristic (ROC) curves of combined prediction model of training group (4A) and validation group (4B), comparative analysis for ROC curves of training groups between prediction models (4C) and the calibration curve of combined prediction model (4D).
表2  淋巴结转移联合预测模型和单一预测模型间的DeLong检验结果
Tab. 2  DeLong test results between combined prediction model and single prediction model of lymph node metastasis

2.4 各预测模型的临床获益度评价

       DCA显示(图5),使用联合预测模型评估直肠癌患者LNM较各单一预测模型有更高的临床净获益率,这再次验证了联合预测模型较单一预测模型更具优势,并且临床影响曲线显示,联合预测模型的预测结果也更接近真实值。

图5  决策曲线分析图。5A:联合预测模型(黑色)的临床获益度高于各单一预测模型;5B~5E:分别为临床预测模型、肿瘤组织纹理分析模型、淋巴结纹理分析模型及联合预测模型的临床影响曲线,红色曲线表示在各阈概率下被模型划分为淋巴结转移的人数,蓝色曲线表示在各阈概率下淋巴结转移人数的真实数量。两条曲线越接近,说明预测模型越准确。
Fig. 5  Decision curve analysis (DCA) graph. 5A shows that the clinical benefit of the combined prediction model (black) is higher than that of each single prediction model; 5B-5E are the clinical impact curves of the clinical prediction model, the tumor tissue texture analysis model, the lymph node texture analysis model and the combined prediction model, respectively. The red curve represents the number of people who are classified as lymph node metastasis by the model under each threshold probability. The blue curve represents the true number of lymph node metastases at each threshold probability. The closer the two curves are, the more accurate the prediction model is.

3 讨论

       本研究通过纹理分析方法探索直肠癌肿瘤组织内部异质性与LNM之间的关系,以肿瘤组织为研究对象构建的预测模型一定程度上可以对直肠癌的疾病发展进行动态监测,即便是在传统影像诊断方法未能发现肿大淋巴结时,也可以通过肿瘤内部的结构变化预测潜在LNM的风险。而直肠癌目标淋巴结纹理分析预测模型,直接对淋巴结状态进行评估,提高了对肿大淋巴结是否发生转移的诊断精确性。临床指标预测模型通过淋巴结短径/长径比值反映淋巴结形态情况,将血清CEA和糖类抗原CA19-9水平纳入研究来反映血清学改变。最后将三者相结合构建的联合预测模型可通过对直肠癌患者进行多方面评估,来预测发生LNM的风险。作为直肠癌最常见的转移方式之一,LNM也与患者局部复发、总体生存率息息相关,是影响直肠癌预后的独立危险因素[11]。对于一些潜在的LNM患者而言,尽早进行淋巴结清扫术至关重要。实现在手术前通过对肿瘤内部信息的评估,准确高效地预测LNM的可能性,对临床精准化、个体化治疗有重要意义。本研究基于MRI图像的纹理特征,建立能够精准评估、预测直肠癌LNM敏感人群的模型,从而帮助临床做出更好的判断。

       近年来,专家和学者们运用影像组学方法从直肠癌病理学、基因组学、治疗反应和临床结果等诸多方面进行了大量的研究。于丹丹等[12]的一项研究表明,通过常规3.0 T MRI影像诊断方式对直肠癌术前分期的预测准确性并不理想,仅为0.57。而顾洪卫等[13]的研究通过DWI序列结合DCE-MRI检查的方法,构建的直肠癌LNM预测模型的AUC值达到了0.93。本研究通过T2WI单一序列对肿瘤组织和淋巴结的双重分析联合淋巴结形态改变及血清学变化构建的预测模型AUC值为0.978,表现出了更高的诊断效能。Yang等[14]在探究直肠癌T2WI序列直方图特征与LNM关系的研究中发现直肠癌T2WI图像像素的Skewness是预测LNM的独立危险因素,AUC值为0.75,其研究仅通过一个纹理特征参数构建的单一预测模型诊断效能低于本研究。

       本研究通过纹理分析技术筛选出的10个纹理参数,从不同方面反映图像内部结构信息,其中Correlation_AllDirection_offset1_SD、HaralickCorrelation_AllDirection_offset1_SD属于GLCM参数,二者可反映像素间灰度的相似度;ClusterShade_AllDirection_offset4_SD是集群阴影参数,也是GLCM的一类特征,可以度量矩阵的偏斜度,其值越高,表示灰度差异性越大,图像对称性越低;RLM纹理参数GreyLevelNonuniformity_AllDirection_offset7_SD、HighGreyLevelRunEmphasis_AllDirection_offset7_SD、HighGreyLevelRunEmphasis_AllDirection_offset1_SD可通过灰度值相似性反映病灶内部成分复杂程度,其数值越大,ROI像素间灰度相似度越低,病灶越不均匀;而LongRunEmphasis_angle45_offset1、ShortRunEmphasis_AllDirection_offset7_SD可反映图像平滑或粗糙程度;InverseDifferenceMoment_AllDirection_offset7_SD反映的是局部同质性;直方图参数IntensityVariability则能反映出ROI纹理的均匀程度[15, 16, 17]。这些参数的改变与直肠癌LNM具有相关性。

       本研究发现血清CA19-9水平是预测直肠癌LNM的独立危险因素,其可能原因是CA19-9可以活化与肿瘤增殖和转移密切相关的HIF-1和STAT3等信号通路,同时CA19-9高表达也会促进miRNA-192和miRNA-23b-3p等促癌性miRNA的表达水平,从而促进直肠癌细胞增殖转移[18, 19, 20]。而血清CEA水平似乎与直肠癌LNM的关系并不密切,这与Chen等[21]和李燕等[22]的研究结果相似。但既往研究发现CEA在肿瘤免疫逃逸过程中也有重要作用,CEA与程序性死亡抗体1(programmed death ligand 1, PD-L1)之间的相互调节,使得CEA高表达患者体内PD-L1水平升高,细胞免疫抑制,T淋巴细胞对肿瘤杀伤能力减低,从而促进癌细胞的发生和发展[23]。所以血清CEA水平与直肠癌LNM之间的关系值得进一步研究和探索。

       本研究有以下几点不足:首先,本研究纳入样本量较少,来源单一,缺乏外部验证;其次,人工勾画ROI尽管努力确保勾画准确性,但仍存在不可避免的误差;最后,这次研究只分析了直肠MRI T2WI序列的纹理特征,研究内容较单一,在下一步研究中应纳入其他常规序列和增强序列图像,同时尽可能多地结合临床资料和生化指标,进行综合评估,不断完善直肠癌术前LNM预测模型,提高诊断准确性。

       综上所述,本研究初步验证了基于T2WI纹理分析联合临床指标构建的预测模型能够在术前有效评估直肠癌患者淋巴结状态,为临床诊断直肠癌LNM提供帮助,从而为选择最佳的个体化治疗方案提供支持。

[1]
Stoffel EM, Murphy CC. Epidemiology and mechanisms of the increasing incidence of colon and rectal cancers in young adults[J]. Gastroenterology, 2020, 158(2): 341-353. DOI: 10.1053/j.gastro.2019.07.055.
[2]
Horvat N, Carlos Tavares Rocha C, Clemente Oliveira B, et al. MRI of rectal cancer: tumor staging, imaging techniques, and management[J]. Radiographics, 2019, 39(2): 367-387. DOI: 10.1148/rg.2019180114.
[3]
Nakanishi R, Akiyoshi T, Toda S, et al. Radiomics approach outperforms diameter criteria for predicting pathological lateral lymph node metastasis after neoadjuvant (chemo)radiotherapy in advanced low rectal cancer[J]. Ann Surg Oncol, 2020, 27(11): 4273-4283. DOI: 10.1245/s10434-020-08974-w.
[4]
Benson AB, Venook AP, Al-Hawary MM, et al. NCCN guidelines insights: rectal cancer, version 6.2020[J]. J Natl Compr Canc Netw, 2020, 18(7): 806-815. DOI: 10.6004/jnccn.2020.0032.
[5]
Glynne-Jones R, Wyrwicz L, Tiret E, et al. Rectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up[J]. Ann Oncol, 2017, 28(suppl_4): iv22-iv40. DOI: 10.1093/annonc/mdx224.
[6]
中华人民共和国国家卫生健康委员会. 中国结直肠癌诊疗规范(2020年版)[J]. 中华外科杂志, 2020, 58(08): 561-585. DOI: 10.3760/cma.j.cn112139-20200518-00390.
National Health Commission of the People's Republic of China. Chinese protocol of diagnosis and treatment of colorectal cancer (2020 edition)[J]. Zhonghua Wai Ke Za Zhi, 2020, 58(8): 561-585. DOI: 10.3760/cma.j.cn112139-20200518-00390.
[7]
Peacock O, Chang GJ. The landmark series: management of lateral lymph nodes in locally advanced rectal cancer[J]. Ann Surg Oncol, 2020, 27(8): 2723-2731. DOI: 10.1245/s10434-020-08639-8.
[8]
Xu HS, Zhao WY, Guo WB, et al. Prediction model combining clinical and MR data for diagnosis of lymph node metastasis in patients with rectal cancer[J]. J Magn Reson Imaging, 2021, 53(3): 874-883. DOI: 10.1002/jmri.27369.
[9]
Wilson R, Devaraj A. Radiomics of pulmonary nodules and lung cancer[J]. Transl Lung Cancer Res, 2017, 6(1): 86-91. DOI: 10.21037/tlcr.2017.01.04.
[10]
Liu YJ, Fan HJ, Dong D, et al. Computed tomography-based radiomic model at node level for the prediction of normal-sized lymph node metastasis in cervical cancer[J]. Transl Oncol, 2021, 14(8): 101113. DOI: 10.1016/j.tranon.2021.101113.
[11]
李梦蕾, 张敬, 淡一波, 等. 术前预测结直肠癌淋巴结转移的临床-影像组学列线图的建立和验证[J]. 中国癌症杂志, 2020, 30(1): 49-56. DOI: 10.19401/j.cnki.1007-3639.2020.01.006.
Li ML, Zhang J, Dan YB, et al. Development and validation of a clinical-radiomics nomogram for preoperative prediction of lymph node metastasis in colorectal cancer[J]. China Oncol, 2020, 30(1): 49-56. DOI: 10.19401/j.cnki.1007-3639.2020.01.006.
[12]
于丹丹, 王湄, 许礼平, 等. 3T MRI在直肠癌术前分期中的准确性和临床价值研究[J]. 中国肿瘤外科杂志, 2021, 13(5): 454-457. DOI: 10.3969/j.issn.1674-4136.2021.05.008.
Yu DD, Wang M, Xu LP, et al. Clinical value of 3T MRI for the preoperative staging of rectal cancer[J]. Chin J Surg Oncol, 2021, 13(5): 454-457. DOI: 10.3969/j.issn.1674-4136.2021.05.008.
[13]
顾洪卫, 郑信兵, 陆鸣建. DWI联合DCE-MRI方法在评估直肠癌区域性淋巴结转移中的作用研究[J]. 医学理论与实践, 2021, 34(20): 3503-3505, 3531. DOI: 10.19381/j.issn.1001-7585.2021.20.004.
Gu HW, Zheng XB, Lu MJ. Research of the regional lymph node metastasis in rectal cancer DWI combined with DCE-MRI[J]. J Med Theory Pract, 2021, 34(20): 3503-3505, 3531. DOI: 10.19381/j.issn.1001-7585.2021.20.004.
[14]
Yang LQ, Liu D, Fang X, et al. Rectal cancer: can T2WI histogram of the primary tumor help predict the existence of lymph node metastasis?[J]. Eur Radiol, 2019, 29(12): 6469-6476. DOI: 10.1007/s00330-019-06328-z.
[15]
Shayesteh SP, Alikhassi A, Farhan F, et al. Prediction of response to neoadjuvant chemoradiotherapy by MRI-based machine learning texture analysis in rectal cancer patients[J]. J Gastrointest Cancer, 2020, 51(2): 601-609. DOI: 10.1007/s12029-019-00291-0.
[16]
Lubner MG, Smith AD, Sandrasegaran K, et al. CT texture analysis: definitions, applications, biologic correlates, and challenges[J]. Radiographics, 2017, 37(5): 1483-1503. DOI: 10.1148/rg.2017170056.
[17]
孙祥林, 由英宁, 赵相轩, 等. 基于T2WI的纹理分析在预测肝癌经导管肝动脉化疗栓塞术后早期复发的价值[J]. 磁共振成像, 2021, 12(8): 22-26, 32. DOI: 10.12015/issn.1674-8034.2021.08.005.
Sun XL, You YN, Zhao XX, et al. Prediction of the early recurrence of HCC patients after TACE surgery based on T2 weighted image texture analysis[J]. Chin J Magn Reson Imaging, 2021, 12(8): 22-26, 32. DOI: 10.12015/issn.1674-8034.2021.08.005.
[18]
Stojkovic Lalosevic M, Stankovic S, Stojkovic M, et al. Can preoperative CEA and CA19-9 serum concentrations suggest metastatic disease in colorectal cancer patients?[J]. Hell J Nucl Med, 2017, 20(1): 41-45. DOI: 10.1967/s002449910505.
[19]
韩万斌, 贾守洪. 胰腺癌根治性术患者血清CA19-9、PIVKA-Ⅱ、VCAM-1水平及其预后的临床价值[J]. 实用癌症杂志, 2020, 35(11): 1821-1825. DOI: 10.3969/j.issn.1001-5930.2020.11.021.
Han WB, Jia SH. Levels of serum CA19-9, PIVKA-Ⅱand VCAM-1 and clinical value of prognosis in patients with radical surgery of pancreatic cancer[J]. Pract J Cancer, 2020, 35(11): 1821-1825. DOI: 10.3969/j.issn.1001-5930.2020.11.021.
[20]
许卫娜, 王婷. 患者血清CEA、CA724、CA199水平对胰腺癌的评估价值[J]. 实用癌症杂志, 2019, 34(5): 747-749. DOI: 10.3969/j.issn.1001-5930.2019.05.015.
Xu WN, Wang T. Evaluation value of serum CEA, CA724 and CA199 levels in patients with pancreatic cancer[J]. Pract J Cancer, 2019, 34(5): 747-749. DOI: 10.3969/j.issn.1001-5930.2019.05.015.
[21]
Chen DY, Wu XB, Xia M, et al. Upregulated exosomic miR-23b-3p plays regulatory roles in the progression of pancreatic cancer[J]. Oncol Rep, 2017, 38(4): 2182-2188. DOI: 10.3892/or.2017.5919.
[22]
李燕, 张凯, 王小平, 等. CT纹理分析对于术前预测直肠癌淋巴结转移的初步研究[J]. 临床放射学杂志, 2021, 40(5): 930-934. DOI: 10.13437/j.cnki.jcr.2021.05.021.
Li Y, Zhang K, Wang XP, et al. Texture analysis of preoperative CT images for prediction of lymph node metastasis: a preliminary study in patients with rectal cancer[J]. J Clin Radiol, 2021, 40(5): 930-934. DOI: 10.13437/j.cnki.jcr.2021.05.021.
[23]
Macherla S, Laks S, Naqash AR, et al. Emerging role of immune checkpoint blockade in pancreatic cancer[J]. Int J Mol Sci, 2018, 19(11): E3505. DOI: 10.3390/ijms19113505.

上一篇 慢性肝病背景下磁共振弹性成像肝脏硬度值的可重复性和稳定性研究
下一篇 全病灶ADC直方图分析在直肠癌组织学分级的临床应用研究
  
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