CT增强扫描灰度直方图纹理分析技术对肺癌患者术前恶性程度评估及对预后的预测价值  

The Value of Gray Histogram Texture Analysis Technique of Enhanced CT Scanning in Evaluating Preoperative Malignancy Degree and Predicting Prognosis of Patients with Lung Cancer

作　　者：毛宇[1] 黎明[1] 乔文婷 郭剑峰 李容波 白艳艳[2] MAO Yu;LI Ming;QIAO Wen-ting;GUO Jian-feng;LI Rong-bo;BAI Yan-yan(Department of Thoracic Surgery,Huhhot First Hospital,Hohhot 010030,Inner Mongolia,China;Department of Anesthesiology,Huhhot First Hospital,Hohhot 010030,Inner Mongolia,China)

机构地区：[1]呼和浩特市第一医院胸外科,内蒙古呼和浩特010030 [2]呼和浩特市第一医院麻醉科,内蒙古呼和浩特010030

出　　处：《中国CT和MRI杂志》2024年第5期79-82,共4页Chinese Journal of CT and MRI

基　　金：国家自然基金(81960577);呼和浩特市科技计划项目(2021-社-10)。

摘　　要：目的探究CT增强扫描灰度直方图纹理分析技术对肺癌患者术前恶性程度评估及对预后的预测价值。方法选择2018年5月至2022年9月在我院就诊的且经术后病理证实的肺癌患者580例作为研究对象。依据术后病理分期,将研究对象分为高分化组(183例)、中分化组(195例)、低分化组(202例)。根据预后情况,将其分为预后良好组(356例)和预后不良组(224例)。由2名医师提取患者CT增强扫描灰度直方图纹理参数。受试者工作特征(receiver operating characteristic,ROC)曲线分析特征参数对肺癌患者术前恶性程度的诊断效能及预后的预测价值。多因素Logistic回归分析预后不良的重要影响因素并构建人工神经网络模型。Pearson分析检验参数间的相关性。结果均值、10%分位、50%分位等参数在高、中、低分化三组中有显著差异。三个参数对评估肺癌患者恶性程度均具有一定的诊断效能,且三者联合诊断效能最佳。均值、10%分位、50%分位是预后不良的保护因素,三者联合预测肺癌患者的曲线AUC值大于各指标单独预测的AUC值。ROC曲线和累积增益图表明该模型预测能力良好。纹理特征参数中,69.44%的参数具有相关性。结论CT增强扫描灰度直方图纹理参数在一定程度上反应肺癌患者术前恶性程度信息,给术前预测以及预后提供了重要方法。Objective To explore the predictive value of texture analysis of ADC map for histological grading of pancreatic cancer.Methods From May 2018 to September 2022,580 patients with lung cancer confirmed by postoperative pathology were selected as the research object.According to postoperative pathological stage,the subjects were divided into highly differentiated group(183 cases),moderately differentiated group(195 cases)and poorly differentiated group(202 cases).According to the prognosis,they were divided into a good prognosis group(356 cases)and a bad prognosis group(224 cases).Texture parameters of gray histogram of enhanced CT scan were extracted by 2 physicians.The predictive value of characteristic parameters of receiver operating characteristic(ROC)curve on preoperative malignant degree diagnosis and prognosis of patients with gastric cancer.Multivariate Logistic regression was used to analyze the important influencing factors of poor prognosis and construct the artificial neural network model.The correlations between the parameters were examined by Pearson analysis.Results The parameters of mean value,10%and 50%were significantly different among the three groups with high,middle and low differentiation.All three parameters have certain diagnostic efficacy in evaluating the malignancy degree of lung cancer patients,and the combination of the three parameters has the best diagnostic efficacy.The mean,10%and 50%fractions were protective factors for poor prognosis.The AUC value of the combined prediction curve of the three fractions was greater than that of the AUC value predicted by each index alone.ROC curve and cumulative gain graph show that the model has good predictive ability.Among the texture feature parameters,69.44%of the parameters have correlation.Conclusion The texture parameters of the gray histogram of enhanced CT scan can reflect the preoperative malignant degree information of patients with gastric cancer to a certain extent,and provide an important method for preoperative prediction and prognos

关 键 词：肺癌 CT增强扫描 灰度直方图 纹理分析 

分 类 号：R734.2[医药卫生—肿瘤]