摘要
目的:探讨应用最大似然法鉴别周围性肺癌、肺错构瘤及结核球3种孤立性肺结节(SPN)的诊断价值。方法:对150例经临床和手术病理证实的SPN(周围性肺癌、肺错构瘤、结核球各50例)的HRCT影像学征象进行分类统计分析,运用最大似然判别法,将其各种征象转化为记分值,以数值的大小来判定肺结节所属的类型。结果:最大似然法对周围型肺癌、肺错构瘤及结核球的诊断正确率分别为86%、92%及90%,平均诊断正确率为89.3%,高于常规阅片法的82%,但两组间差异无统计学意义(χ2=2.434,P>0.05)。最能提示为肺癌的征象依次为空泡征、分叶征、支气管充气征和血管集束征;最能提示为肺错构瘤的征象依次为脂肪、边缘清楚及钙化;最能提示为结核球的征象依次为空洞、卫星灶、钙化及胸膜凹陷征。结论:最大似然法对常见SPN的鉴别诊断正确率有所提高,是一种辅助影像学诊断的有价值的方法。
Objective:To evaluate the diagnostic values of discrimination method of large log-likelihood study in differentiating the 3 kinds of SPN among peripheral lung carcinoma, hamartoma and tuberculoma. Methods: 150 patients proved by pathology and clinic with peripheral lung carcinomas( n = 50), hamaltomas(n = 50) and tuberculomas(n = 50) were studied statistically with HRCT manifestations. With the discrimination method of large log-likelihood, CT signs of the 3 kinds of SPN were analyzed and then translated them into probability scores to identify the nodular category. Results: The diagnostic accuracy with discrimination method of large log-likelihood of peripheral lung carcinoma, hamartoma and tuberculoma were 86 %, 92 % and 90%, the whole accuracy (89.3 % ) of discrimination method of large log-likelihood was higher than that(82% ) of routine diagnostic method, but there was no statistical difference (χ^2 = 2. 434, P 〉 0.05 ). The most probable CT signs for lung cancer were vacuole sign, lobulation, bronchogram and vascular convergence; The most probable CT signs for hamartoma were fat, well-defined margin and calcification; The most probable CT signs for tuberculoma were cavitation, satellite, calcification and pleural indentation. Conclusion: The discrimination method of large log-likelihood is a kind of valuable method to assist imaging diagnosis, and it can make the differential diagnostic accuracy of SPN higher than that of routine diagnostic method.
出处
《医学影像学杂志》
2007年第4期352-356,共5页
Journal of Medical Imaging
关键词
肺癌
肺孤立结节
结核球
错构瘤
最大似然法
体层摄影术
X线计算机
Lung carcinoma
Solitary pulmonary nodule
Tuberculoma
Hamartoma
Maximal log-likelihood
Tomography, X-ray computed