摘要
目的:探讨基于临床变量结合多参数磁共振成像(MRI)影像组学特征构建的列线图模型对乳腺癌化疗患者预后的预测价值。方法:选取2020年1月~2021年12月青岛市中心医院接受化疗的150例乳腺癌患者作为研究对象。所有患者化疗前均行乳腺MRI检查,利用分割软件手动勾画多参数乳腺MRI肿瘤感兴趣区,提取最佳影像组学特征。对所有患者随访2年,根据疾病复发、转移及死亡等预后情况分成预后良好组、预后不良组,并将患者按7:3比例分为训练集(n=105)和测试集(n=45)。收集患者的临床资料及多参数MRI影像组学特征,通过单因素和多因素Logistic回归分析乳腺癌化疗患者预后的影响因素,基于Logistic回归分析分别构建临床模型、影像组学模型及联合列线图模型并绘制列线图,采用受试者工作曲线分析3种模型对乳腺癌化疗患者预后的预测价值。结果:150例乳腺癌化疗患者中复发转移33例、死亡19例,共52例纳入预后不良组。Logistic回归分析显示,TNM分期、雌激素受体(ER)及孕激素受体(PR)是影响乳腺癌化疗患者预后的临床预测因子(P<0.05)。训练集中预后良好、预后不良患者分别有69、36例,测试集中预后良好、预后不良患者分别有29、16例;通过最小绝对收缩和选择算法分别从TIRM、DWI、DCE-MRI序列中筛选出16个与乳腺癌化疗患者预后相关的组学特征并计算影像组学评分,且训练集、测试集中预后良好患者的影像组学评分均高于预后不良组(P<0.05)。ROC曲线结果显示,影像组学模型、临床模型及联合列线图模型在训练集的曲线下面积(AUC)为0.853、0.741、0.923(P<0.05);影像组学模型、临床模型及联合列线图模型在测试集的AUC为0.797、0.749、0.896(P<0.05)。结论:基于化疗前的影像组学评分、TNM分期、ER及PR是乳腺癌化疗患者预后的独立预测因子,而临床-多参数MRI影像组学的列线图可有效提高乳腺癌化疗患者预后的预测效能。
Objective To investigate the predictive value of a nomogram model constructed with clinical variables and multiparametric magnetic resonance imaging(MRI)radiomics features in prognosis of patients undergoing chemotherapy for breast cancer.Methods A total of 150 breast cancer patients who received chemotherapy in Qingdao Central Hospital from January 2020 to December 2021 were enrolled for the study.All patients underwent breast MRI examination before chemotherapy,and the optimal radiomics features were extracted from the tumor regions of interest on multiparametric MRI image which were manually outlined using segmentation software.After 2-year follow-up,the patients were divided into good prognosis group and poor prognosis group according to whether there were disease recurrence,metastasis and death,and they were divided into training set(n=105)and test set(n=45)in a ratio of 7:3.The clinical data and multiparametric MRI radiomics features were collected,and the factors affecting the prognosis of breast cancer patients receiving chemotherapy were analyzed by univariate and multivariate Logistic regression analyses.Three models,namely clinical model,radiomics model,and clinical-multiparametric MRI radiomics-based nomogram model were constructed,and their predictive value in the prognosis of breast cancer patients treated with chemotherapy was analyzed with receiver operating characteristic curves.Results Out of 150 cases of breast cancer treated with chemotherapy, 52 cases were included in poor prognosis group, with33 cases of recurrence and/or metastases and 19 cases of deaths. Logistic regression analysis identified TNM stage, estrogenreceptor and progesterone receptor as clinical predictors for prognosis of breast cancer patients treated with chemotherapy (P<0.05). There were 69 and 36 patients with good and poor prognoses in training set, and 29 and 16 patients with good and poorprognoses in test set. Through least absolute shrinkage and selection operator algorithm, 16 radiomics features related to theprognosis of breast cancer patients treated with chemotherapy were screened out from TIRM, DWI, and DCE-MRIsequences, and their radiomics scores were calculated. Compared with patients with poor prognosis, those with goodprognosis in training set and test set had higher radiomics scores (P<0.05). ROC curve analysis showed that the area underthe curve of radiomics model, clinical model, and nomogram model were 0.853, 0.741, 0.923 in training set (P<0.05), and0.797, 0.749, 0.896 in test set (P<0.05), respectively. Conclusion Based on pre-chemotherapy radiomics scores, TNM stage,estrogen receptor and progesterone receptor are identified as independent predictors for prognosis of breast cancer patientstreated with chemotherapy, and the clinical-multiparametric MRI radiomics-based nomogram can effectively improve thepredictive efficacy of prognosis in breast cancer patients treated with chemotherapy.
作者
侯红
张亚琳
孙华静
慕强
方堃
HOU Hong;ZHANG Yalin;SUN Huajing;MU Qiang;FANG Kun(Department of Breast Surgery,Qingdao Central Hospital,Qingdao 266042,China)
出处
《中国医学物理学杂志》
CSCD
2024年第9期1115-1121,共7页
Chinese Journal of Medical Physics
基金
山东省自然科学基金青年项目(ZR2022QH055)。
关键词
乳腺癌
磁共振成像
影像组学
列线图
化疗
预后
breast cancer
magnetic resonance imaging
radiomics
nomogram
chemotherapy
prognosis
