【摘要】 为了实现利用XGBoost方法对隆突下淋巴结的良恶性进行分类,提出从CT图像中无创识别隆突下淋巴结的良恶性,利用影像组学方法提取CT图像中隆突下淋巴结的特征。对40例良性和40例恶性淋巴结的CT图像进行分类实验,结果表明,隆突下淋巴结的良恶性准确度为0.813,敏感度为0.825,特异度为0.800。通过递归式特征消除方法进行特征分析,选择熵、中值、联合熵3个特征时,识别性能最好。更多还原

【Abstract】 It is very important for the diagnosis and prognosis of lung cancer and esophageal cancer to accurately evaluate the subcarinal lymph node metastasis.A noninvasive method was presented to identify the benign and malignant of subcarinal lymph node from computed tomography(CT)images.The features of the subcarinal lymph nodes in CT images were extracted by the method of Pyradiomics.The recursive feature elimination method is used for feature analysis.The recognition performance is the best when the features of entropy,median and joint entropy are selected.XGBoost was used for benign and malignant classification.The classification performance was test by 40 benign and 40 malignant subcarinal lymph nodes.The experimental results show that the accuracy is 0.813,the sensitivity is 0.825 and the specificity is 0.800.更多还原