论文部分内容阅读
目的分析筛选出与孤立性肺结节(solitary pulmonary nodules,SPN)恶性概率相关的一组临床资料,建立并验证了SPN良恶性判别的数学模型,并将该模型与国内李运模型和国外Mayo模型、VA模型进行比较。方法分别收集2011年1月至2014年11月在第二军医大学长海医院手术切除并明确病理的资料252例,总结性别、年龄、症状、吸烟史、肺部基础疾病史、既往肿瘤家族史、结节部位、最大直径、边界清楚、边缘光滑、毛刺、分叶、棘突、胸膜凹陷征、血管集束征、透亮影等资料。从252例资料中选出83例作为验证组(B组),剩余169例作为建模组(A组);同时从B组数据中剔出6例使得其剩余的77例数据均符合其他3个模型的入选和排除条件并组成C组。通过Logistic分析A组资料筛选出与SPN良恶性相关的5个独立因子,构建良恶性概率判别模型。并用B组验证本文模型、C组分别对四个模型进行统一验证和比较。结果年龄、既往肿瘤史、最大直径、钙化、透亮影这5项因素的差异在良性和恶性SPN之间有统计学意义(P<0.05)。建立的SPN良恶性概率数学判别方程,将B组数据代入公式,得出的model ROC(receiver operating characteristic)曲线下面积(AUC)为0.905±0.036,灵敏性为79.3%、特异性为84.0%、阳性似然比为4.957、阴性似然比为0.246、阳性预测值为0.920、阴性预测值0.636。将C组数据验证长海模型AUC为0.893±0.040,李运模型AUC为0.817±0.056,Mayo模型AUC为0.804±0.050,VA模型AUC为0.780±0.057。结论患者年龄、肿瘤史、结节最大直径、钙化、透亮影是SPN良、恶性判别的独立预测因子,通过Logistic回归建立的数学模型有一定的临床应用价值。对于本研究的患者病例,长海模型比李运模型、Mayo模型、VA模型预测效果都更有效。
Objective To analyze and screen a set of clinical data related to the malignant probability of solitary pulmonary nodules (SPN), establish and verify the mathematical model of benign and malignant SPN, and compare the model with the domestic Li Yun model and foreign Mayo Model, VA model for comparison. Methods 252 cases of surgical resection and definite pathology in Changhai Hospital, Second Military Medical University from January 2011 to November 2014 were collected respectively. The data of gender, age, symptom, smoking history, pulmonary underlying disease history, family history of tumor, Nodules, the largest diameter, clear boundaries, smooth edges, burr, lobulation, spinous process, pleural indentation, vascular bundles sign, translucent shadow and other information. 83 cases were selected from 252 cases as the verification group (group B), and the remaining 169 cases were used as the modeling group (group A). Meanwhile, 6 cases were excluded from the data of group B, and the remaining 77 cases were in line with the other 3 A model of the inclusion and exclusion conditions and the formation of C group. Logistic analysis of A group of data screening and benign and malignant SPN associated with five independent factors, to establish a benign and malignant probability discriminant model. Group B is used to verify the model, and group C is used to verify and compare the four models. Results The differences of age, previous history of tumor, maximal diameter, calcification, and transillumination were statistically significant between benign and malignant SPNs (P <0.05). (AUC) was 0.905 ± 0.036, the sensitivity was 79.3%, the specificity was 84.0%, the specificity of the receiver operating characteristic curve The positive likelihood ratio was 4.957, the negative likelihood ratio was 0.246, the positive predictive value was 0.920, the negative predictive value was 0.636. The AUC of C model was 0.893 ± 0.040, the AUC of Liyun model was 0.817 ± 0.056, the AUC of Mayo model was 0.804 ± 0.050, and the AUC of VA model was 0.780 ± 0.057. Conclusion The age, tumor history, maximum nodule diameter, calcification, and transillumination are independent predictors of benign and malignant SPN. The mathematical model established by Logistic regression has some clinical value. For the patient cases in this study, the Changhai model is more effective than the Li Yun model, the Mayo model and the VA model prediction effect.