基于治疗前磁共振成像的放射组学预测鼻咽癌远处转移的初步研究
Radiomics based on pretreatment MRI for predicting distant metastasis of nasopharyngeal carcinoma: A preliminary study.
作者信息
Jiang Tingting, Tan Yalan, Nan Shuaimin, Wang Fang, Chen Wujie, Wei Yuguo, Liu Tongxin, Qin Weifeng, Lu Fangxiao, Jiang Feng, Jiang Haitao
机构信息
Department of Radiology, Cancer Hospital of the University of Chinese Academy of Sciences (Zhejiang Cancer Hospital), Hangzhou, China.
Institute of Basic Medicine and Cancer (IBMC), Chinese Academy of Sciences, Hangzhou, China.
出版信息
Front Oncol. 2022 Aug 9;12:975881. doi: 10.3389/fonc.2022.975881. eCollection 2022.
OBJECTIVE
To explore the feasibility of predicting distant metastasis (DM) of nasopharyngeal carcinoma (NPC) patients based on MRI radiomics model.
METHODS
A total of 146 patients with NPC pathologically confirmed, who did not exhibit DM before treatment, were retrospectively reviewed and followed up for at least one year to analyze the DM risk of the disease. The MRI images of these patients including T2WI and CE-T1WI sequences were extracted. The cases were randomly divided into training group (n=116) and validation group (n=30). The images were filtered before radiomics feature extraction. The least absolute shrinkage and selection operator (LASSO) regression was used to develop the dimension of texture parameters and the logistic regression was used to construct the prediction model. The ROC curve and calibration curve were used to evaluate the predictive performance of the model, and the area under curve (AUC), accuracy, sensitivity, and specificity were calculated.
RESULTS
72 patients had DM and 74 patients had no DM. The AUC, accuracy, sensitivity and specificity of the model were 0. 80 (95% CI: 0.720. 88), 75.0%, 76.8%, 73.3%. and0.70 (95% CI: 0.510.90), 66.7%, 72.7%, 63.2% in training group and validation group, respectively.
CONCLUSION
The radiomics model based on logistic regression algorithm has application potential for evaluating the DM risk of patients with NPC.
目的
探讨基于MRI影像组学模型预测鼻咽癌（NPC）患者远处转移（DM）的可行性。
方法
回顾性分析146例经病理确诊且治疗前未出现DM的NPC患者，随访至少1年以分析疾病的DM风险。提取这些患者包括T2WI和CE-T1WI序列的MRI图像。病例随机分为训练组（n = 116）和验证组（n = 30）。在进行影像组学特征提取前对图像进行滤波。采用最小绝对收缩和选择算子（LASSO）回归确定纹理参数维度，采用逻辑回归构建预测模型。采用ROC曲线和校准曲线评估模型的预测性能，并计算曲线下面积（AUC）、准确性、敏感性和特异性。
结果
72例患者发生DM，74例患者未发生DM。训练组和验证组模型的AUC、准确性、敏感性和特异性分别为0.80（95%CI：0.720.88）、75.0%、76.8%、73.3%和0.70（95%CI：0.510.90）、66.7%、72.7%、63.2%。
结论
基于逻辑回归算法的影像组学模型在评估NPC患者DM风险方面具有应用潜力。
Zotero 插件