MRI影像组学预测三阴性乳腺癌患者发生放疗抵抗的应用研究

Application of MRI radiomics predicting radiotherapy resistance of patients with triple-negative breast cancer

摘要

目的 探讨磁共振成像(MRI)影像组学预测三阴性乳腺癌(TNBC)发生放射治疗(放疗)抵抗的价值,开发并验证放疗抵抗的预测模型,为乳腺癌患者放疗抵抗提供基础。方法 选择2021年8月至2022年10月于四川省简阳市人民医院行TNBC手术治疗的396例患者为研究对象,随机分为训练集277例和验证集119例,根据是否放疗抵抗分为放疗抵抗组和放疗敏感组.探究放疗抗性的影响因素,采用了Logistic回归分析方法;为了确定特征变量的稳定性和线性关系,运用了组内相关系数(ICC)并进行线性相关性检验.通过构建LASSO回归模型结合K折交叉验证策略,筛选出最优特征集.利用支持向量机(SVM)技术构建了影像组学标签,Wilcoxon检验差异性.进一步应用Logistic回归方法开发了一个直观的列线图预测模型,并通过Bootstrap重采样方法进行内部验证,最终利用受试者工作特征曲线(ROC曲线)分析来评估该模型的区分效能,校准曲线评估准确度,决策曲线评估临床收益。结果 Logistic回归结果显示雄激素受体(AR)、肿瘤蛋白p53(p53)、细胞角蛋白5/6(CK5/6)、增殖细胞核抗原(Ki-67)、血小板/淋巴细胞比率(PLR)、中性粒/淋巴细胞比率(NLR)是TNBC发生放疗抵抗的独立危险因素(OR值分别为1.364、2.587、3.054、1.486、2.187、3.520,P<0.05),癌胚抗原(CEA)、糖类抗原153(CA153)是TNBC发生放疗抵抗的保护因素(OR值分别为0.795、0.256,P<0.05);筛选出影像组学标签10个特征,在放疗抵抗组和放疗敏感组中差异性显著(P<0.05);训练集中列线图模型(AUC:0.797,95%CI:0.712~0.859)的灵敏度为70.09%,特异度为70.52%,准确度为71.03%,模型拟合效果显著,相较于单独的影像学模型和临床特征模型,本研究所构建的模型在净收益方面展现出了显著的优势。结论 MRI扫描后提取了10个较优特征,结合临床数据与动态增强扫描影像组学特征所构建的列线图预测模型,在实际临床应用中展现出了良好的可行性和实用性,临床决策指导价值较高.

Abstract

Objective To explore value of magnetic resonance imaging (MRI) radiomics for predicting radiation therapy (radiotherapy) resistance in triple-negative breast cancer (TNBC), develop and verify prediction model of radiotherapy resistance, so as to provide a basis for basic and clinical application researches on radiotherapy resistance in breast cancer patients. Methods 396 patients who underwent surgery for TNBC at Jianyang Municipal People's Hospital in Sichuan province from August 2021 to October 2022 were selected as study subjects. The patients were randomly divided into training set (277 cases) and validation set (119 cases). According to response to radiotherapy, they were categorized into radiotherapy-resistant group and radiotherapy-sensitive group. Logistic regression analysis was used to identify factors affecting radiotherapy resistance. The intra-class correlation coefficient (ICC) and linear correlation tests were employed to determine stability of characteristic variables and linear relationships among them. LASSO regression model was established, combined with K-fold cross-validation, optimal feature subset was identified out. Support vector machine (SVM) algorithm technique was utilized to construct radiomic labels, and Wilcoxon test was applied to find out differences among them. Further, Logistic regression was used to create a visual column chart (nomogram) model, Bootstrap self-sampling was used to validate the model, receiver operating characteristic (ROC) curve was used to evaluate distinguishing efficiency of the model, calibration curve was used to assess accuracy of the model, and decision curve was used to evaluate clinical benefit of the model. Results Logistic regression analysis shown that androgen receptor (AR) [odds ratio (OR)=1.364], tumor protein 53 (p53) (OR=2.587), cytokeratin 5/6 (CK5/6) (OR=3.054), proliferation cell nuclear antigen (Ki-67) (OR=1.486), platelet-to-lymphocyte ratio (PLR) (OR=2.187), and neutrophil-to-lymphocyte ratio (NLR) (OR=3.520) were independent risk factors for radiation resistance in TNBC (all P<0.05). Conversely, carcinoembryonic antigen (CEA) (OR=0.795) and cancer antigen 15-3 (CA15-3) (OR=0.256) were protective factors against radiotherapy in TNBC (both P<0.05). Totally, 10 radiomic features were screened out, there were significant differences between the radiation-resistant group and the radiation-sensitive group (all P<0.05). The sensitivity of the nomogram model in the training set was 70.09% (AUC: 0.797, 95%CI: 0.712-0.859), and the specificity was 70.52%, and the overall accuracy was 71.03%. The model demonstrated a significant fitting effect, and its net benefit was considerably higher than that of the imaging model or the clinical feature model alone. Conclusion After MRI scan, 10 optimal features are extracted. The nomogram model constructed by dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) combined with clinical data has good clinical feasibility and practicality, and high guiding value for clinical decision.

关键词

MRI影像组学 / 三阴性乳腺癌 / 放疗抵抗 / 列线图模型 / 免疫组化 / 血液检测

Key words

magnetic resonance imaging radiomics / triple-negative breast cancer / radiotherapy resistance / nomogram model / immunohistochemistry / blood test

参考文献

[1] Sung H, Ferlay J, Siegel RL, 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.

[2] Pusztai L, Foldi J, Dhawan A, et al. Changing frameworks in treatment sequencing of triple-negative and HER2-positive, early-stage breast cancers[J]. Lancet Oncol, 2019, 20(7): e390-e396.

[3] Yin L, Duan JJ, Bian X W, et al. Triple-negative breast cancer molecular subtyping and treatment progress[J]. Breast Cancer Res, 2020, 22(1): 1-13.

[4] Pesapane F, Rotili A, Agazzi G M, et al. Recent radiomics advancements in breast cancer: Lessons and pitfalls for the next future[J]. Curr Oncol, 2021, 28(4): 2351-2372.

[5] 吴天斌, 张有健, 林桂涵, 等. 基于DCE-MRI影像组学和常规MRI特征的列线图预测乳腺癌淋巴血管侵犯[J]. 温州医科大学学报, 2022, 52(11): 882-888.

[6] Liu W, Zheng M, Zhang R, et al. RNF126-Mediated MRE11 ubiquitination activates the DNA damage response and confers resistance of triple-negative breast cancer to radiotherapy[J]. Adv Sci (Weinh), 2023, 10(5): e2203884.

[7] Bai X, Ni J, Beretov J, et al. Triple-negative breast cancer therapeutic resistance: Where is the Achilles' heel? [J]. Cancer Lett, 2021, 28(497): 100-111.

[8] 陈莉杰. Ki67在三阴性乳腺癌中的表达及其与临床预后相关性的分析[J]. 中国妇幼健康研究, 2020, 31(10): 1415-1418.

[9] Montero A, Ciérvide R, García-Aranda M, et al. Postmastectomy radiation therapy in early breast cancer: Utility or futility? [J]. Crit Rev Oncol Hematol, 2020, 147: 102887.

[10] Kierans SJ, Taylor CT. Regulation of glycolysis by the hypoxia-inducible factor (HIF): implications for cellular physiology[J]. J Physiol, 2021, 599(1): 23-37.

[11] Manjunath M, Choudhary B. Triple-negative breast cancer: a run-through of features, classification and current therapies[J]. Oncol Lett, 2021, 22(1): 512.

[12] Conti A, Duggento A, Indovina I, et al. Radiomics in breast cancer classification and prediction[J]. Semin Cancer Biol, 2021, 72: 238-250.

[13] Ngai V, Tai JCJ, Taj S, et al. Non-invasive predictors of axillary lymph node burden in breast cancer: a single-institution retrospective analysis[J]. Breast Cancer Res Treat, 2022, 195(2): 161-169.

[14] Xiong J, Zuo W, Wu Y, et al. Ultrasonography and clinicopathological features of breast cancer in predicting axillary lymph node metastases[J]. BMC Cancer, 2022, 22(1): 1155.

[15] Thompson JL, Wright G P. The role of breast MRI in newly diagnosed breast cancer: An evidence-based review[J]. Am J Surg, 2021, 221(3): 525-528.

[16] Szep M, Pintican R, Boca B, et al. Multiparametric MRI features of breast cancer molecular subtypes[J]. Medicina (Kaunas), 2022, 58(12): 1716.

[17] 夏旭东, 段成洲, 李铭, 等. 基于MRI影像组学列线图预测乳腺癌腋窝淋巴结转移[J]. 磁共振成像, 2022, 13(1): 118-122.

[18] Surov A, Kim JY, Aiello M, et al. Associations between dynamic contrast enhanced magnetic resonance imaging and clinically relevant histopathological features in breast cancer: a multicenter analysis[J]. In Vivo, 2022, 36(1): 398-408.

[19] Lee J, Kim SH, Kang BJ. Prognostic factors of disease recurrence in breast cancer using quantitative and qualitative magnetic resonance imaging (MRI) parameters.[J]. Sci Rep, 2020, 10(1): 7598.

[20] Wang Q, Li B, Liu Z, et al. Prediction model of axillary lymph node status using automated breast ultrasound (ABUS) and ki-67 status in early-stage breast cancer[J]. BMC Cancer, 2022, 22(1): 929.

[21] 梁春雨, 马伶玉, 徐志渊, 等. 磁共振成像(MRI)及肿瘤分子标记物对乳腺癌新辅助化疗腋窝淋巴结疗效的评估与预测[J]. 中国CT和MRI杂志, 2022, 20(3): 67-70.