基于MRI影像组学特征因素预测II~IVa期鼻咽癌患者复发转移风险及辅助化疗受益的临床意义

Clinical significance of predicting the risk of recurrence and metastasis and the benefits of adjuvant chemotherapy in stage II-IVa nasopharyngeal carcinoma patients based on MRI radiomics features

目的旨在建立一个MRI影像组学和临床因素模型,以预测原发性鼻咽癌(NPC)患者复发转移的风险,并验证其对辅助化疗(adjuvant chemotherapy,AC)疗效的预测作用。方法回顾性分析2018年2月~2021年10月在深圳市龙岗区耳鼻咽喉医院诊断为II~IVa期NPC患者135例。在我院接受标准同步放化疗后,部分患者接受诱导化疗和/或基于顺铂/奈达铂的辅助化疗。利用PyRadiomics平台提取增强MRI序列的成像特征。使用最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)算法过滤与复发或转移相关的特征,在训练队列中通过Cox多变量分析构建临床影像组学模型(clinical radiomics models,CRM),并在验证队列中进行验证。根据模型的Rad评分中位数,将所有患者分为高风险组和低风险组。Kaplan-Meier生存曲线用于比较高风险组和低风险组辅助化疗患者的3年无复发转移生存期(recurrence and metastasis free survival,RMFS)。结果共提取960个成像特征。CRM由9个特征(6个影像学特征和3个临床因素)构成。在训练队列中,3年RMFS的CRM曲线下面积(AUC)为0.867(P<0.001),敏感性和特异性分别为90.32%和79.66%;在验证队列中,AUC为0.836(P<0.001),敏感性和特异性分别为100.0%和71.43%。高风险组和低风险组3年RMFS分别为42.86%(27/63)和94.44%(68/72)(Log rank=50.818,P<0.001)。在CRM高风险患者中,接受辅助化疗的患者3年RMFS明显优于未接受患者(Log rank=6.204,P=0.013)。结论基于3个临床因素和6个MRI影像学特征的CRM为预测NPC的预后结局提供一种无创的方法,可能有助于指导临床辅助化疗的治疗决策,但还需要进一步外部验证。

OBJECTIVE To develop a magnetic resonance(MRI)imaging radiomics and clinical factor model to predict recurrence and metastasis in patients with primary stage II-IVa nasopharyngeal carcinoma(NPC)and to validate its predictive effect on adjuvant chemotherapy(AC)outcomes.METHODS A retrospective analysis was performed on 135 patients with stage II to IVa NPC diagnosed in Longgang Otolaryngology Hospital of Shenzhen City from February 2018 to October 2021.After receiving standard synchronous radiotherapy and chemotherapy at our hospital,some patients received induction chemotherapy and/or AC based on cisplatin/nedaplatin.The imaging features of enhanced MRI sequences were extracted using PyRadiomics platform.Using the least absolute shrinkage and selection operator(LASSO)algorithm to filter features associated with recurrence or metastasis,a clinical radiomics model(CRM)was constructed by Cox multivariate analysis in a training cohort and validated in a validation cohort.All patients were divided into high-risk and low-risk groups based on the model's median Rad score.Kaplan-Meier survival curves were used to compare 3-year recurrence or metastasis free survival(RMFS)in patients with AC in high-risk group and low risk-group.RESULTS A total of 960 imaging features were extracted.The CRM consists of 9 features(6 imaging features and 3 clinical factors).In the training cohort,the area under the CRM curve(AUC)of 3-year RMFS was 0.867(P<0.001),and the sensitivity and specificity were 90.32%and 79.66%,respectively.In the validation cohort,the AUC was 0.836(P<0.001)and the sensitivity and specificity were 100.0%and 71.43%,respectively.The 3-year RMFS in high-risk and low-risk groups was 42.86%(27/63)and 94.44%(68/72)(log rank=50.818,P<0.001),respectively.Among CRM high-risk patients,3-year RMFS was significantly better in patients who received AC than those who did not(log rank=6.204,P=0.013).CONCLUSION CRM based on 3 clinical factors and 6 MRI features provides a non-invasive method for predicting the prognosis of NPC,which may help guide treatment decisions for clinical adjuvant chemotherapy,but further external verification is needed.