国际肿瘤学杂志

国际肿瘤学杂志››2019,Vol. 46››Issue (3): 141-146.doi:10.3760/cma.j.issn.1673-422X.2019.03.003

• 论著 •上一篇下一篇

CT纹理分析预测食管鳞状细胞癌放化疗 敏感性的研究

衣明慧1,4,5李振江2曹强3李宝生4,5

  1. 1济南大学 betway.com 医学与生命科学学院250022； 2山东大学附属山东省肿瘤医院放射物理技术室，济南250117； 3山东省医学影像与放疗工程技术研究中心，济南250117； 4山东大学附属山东省肿瘤医院放疗科，济南250117；5betway.com ，济南250062
  • 出版日期:2019-03-08发布日期:2019-05-17
  • 通讯作者:李宝生 E-mail:baoshli1963@163.com
  • 基金资助:

    国家自然科学基金（81874224）；山东省重点研发计划（2017CXZC1206）

Study of texture analysis based on CT images to predict radiochemotherapy sensitivity in patients with esophageal squamous cell carcinoma

Yi Minghui1,4,5, Li Zhenjiang2, Cao Qiang3, Li Baosheng4,5

  1. 1School of Medicine and Life Sciences, University of JinanShandong Academy of Medical Sciences, Jinan 250022, China; 2Department of Radiation Physics, Shandong Cancer Hospital Affiliated to Shandong University, Jinan 250117, China; 3Shandong Medical Imaging and Radiotherapy Engineering Center, Jinan 250117, China; 4Department of Radiation Therapy, Shandong Cancer Hospital Affiliated to Shandong University, Jinan 250117, China; 5Shandong Academy of Medical Sciences, Jinan 250062, China
  • Online:2019-03-08Published:2019-05-17
  • Contact:Li Baosheng E-mail:baoshli1963@163.com
  • Supported by:

    National Natural Science Foundation of China (81874224); Key Research and Development Program of Shandong Province of China (2017CXZC1206)

PDF

501

摘要：目的探索食管鳞状细胞癌（ESCC）患者的CT纹理特征与放化疗敏感性的关系。方法回顾性收集2014年1月1日至2017年12月31日山东大学附属山东省肿瘤医院收治的92例ESCC患者，通过评价治疗的敏感性，将患者分为敏感组（完全缓解+部分缓解）和抵抗组（疾病稳定+疾病进展）。从患者定位CT图像中提取纹理参数。采用traintestsplit将入组患者分为训练集（46例）和测试集（46例），训练集用来建立纹理参数预测敏感性的模型，测试集进行模型外部验证。结果训练集中敏感组患者31例，抵抗组患者15例，对放疗敏感的比例为67.4%。单因素分析结果显示灰度直方图参数中的偏度在敏感组和抵抗组患者中的差异有统计学意义（Z=2.097，P=0.036），曲线下面积（AUC）为0.692，95%CI为0.539～0.820，预测敏感性的阈值是≤-2.58。将灰度直方图参数中的偏度（Z=2.097，P=0.036）、灰度区域大小矩阵参数中的高灰区域优势（HGZE）（Z=1.722，P=0.085）和小区域高灰度优势（SZHGE）（Z=1.640，P=0.101）纳入二元logistic回归中，结果显示偏度为敏感性的独立影响因素(OR=0.558，95%CI为0.338~0.923，P=0.023 )。对二元logistic回归模型进行ROC曲线分析，结果显示AUC为0.718，95%CI为0.550～0.886，表明回归模型可以预测ESCC放化疗敏感性。用测试集对模型进行验证，AUC为0.706，敏感性为70.6%，特异性为69.0%，说明模型具有一定的预测能力。结论CT纹理分析能够在一定程度上预测ESCC患者放化疗的敏感性。

关键词:食管肿瘤,放射疗法,化学疗法,辅助,治疗效果,纹理分析

Abstract:ObjectiveTo investigate the relationship between texture features based on CT and radiochemotherapy sensitivity in patients with esophageal squamous cell carcinoma (ESCC). MethodsA total of 92 ESCC patients treated at Shandong Cancer Hospital Affiliated to Shandong University between January 2014 and December 2017 were retrospectively collected. All patients were divided into responders (complete response + partial response) and nonresponders (stable disease + progression disease) according to therapeutic sensitivity. The texture features were extracted from CT images for positioning. And the patients were divided into training set (46 patients) and test set (46 patients) using traintestsplit, training set for establishing predictive model and test set for model validation. ResultsThere were 31 responders and 15 nonresponders in the training set, and the portion of responders was 67.4%. Univariate analyses showed that the histogram matrix (HISTO)sknewess was significantly different between the two groups (Z=2.097, P=0.036) and the area under the curve (AUC) was 0.692 with 95%CI of 0.5390.820. Sknewess ≤-2.58 intended to be responders. Binary logistic regression of sknewess (Z=2.097, P=0.036) in HISTO, high gray level zone emphasis (HGZE) (Z=1.722, P=0.085) and small zone high gray level emphasis (SZHGE) (Z=1.640, P=0.101) in gray level zonelength matrix (GLZLM) showed that sknewess was the independent influence factor of sensitivity (OR=0.558, 95%CI: 0.3380.923, P=0.023), and the AUC of logistic regression model was 0.718 with 95%CI of 0.5500.886, which indicted that the model had the ability to predict treatment response of ESCC patients. The model was validated by using test set and the AUC was 0.706, and the sensitivity of the model was 70.6% while the specificity was 69.0%. It showed that the model had certain ability in predicting treatment response. ConclusionCT texture analysis can predict the radiochemotherapy sensitivity in patients with ESCC to some extent.

Key words:Esophageal neoplasms,Radiotherapy,Chemotherapy, adjuvant,Treatment effect,Texture analysis