论文部分内容阅读
目的探讨数据挖掘技术对周围型肺癌影像诊断规则提取的价值。方法收集58例经过临床病理证实的周围型肺癌病例,对其临床及CT表现属性进行标准化认定,输入数据库,分别采用自主开发的基于关联规则知识发现程序与通用数据分析工具ROSETTA中的粗糙集约简算法和遗传分类算法对58例周围型肺癌临床及影像学数据进行挖掘对比研究。结果由Johnson’s Algorithm粗糙集约简算法产生诊断规则51条,由ROSETTA遗传分类算法所产生的诊断规则有5千多条,基于关联规则的挖掘算法所产生的诊断规则有123条。这3种不同的数据挖掘方法产生的最重要的诊断规则基本上都将性别、年龄、位置、毛刺、形状、毛玻璃样密度等属性作为诊断周围型肺癌的主要依据。结论数据挖掘技术在医学影像诊断和鉴别诊断中具有潜在的应用价值。
Objective To explore the value of data mining in the extraction of diagnostic rules of peripheral lung cancer. Methods Fifty-eight cases of peripheral lung cancer confirmed by clinical pathology were collected and their clinical and CT performance attributes were standardized and entered into the database. Rough set reduction in ROSETTA, a self-developed knowledge discovery program based on association rules, and a common data analysis tool Algorithm and genetic classification algorithm for 58 cases of peripheral lung cancer clinical and imaging data mining comparative study. The results showed that there were 51 diagnostic rules generated by Johnson’s Algorithm and Rough Set Reduction Algorithm. The number of diagnostic rules generated by ROSETTA genetic algorithm was more than 5000, and 123 based on association rules. The three most important diagnostic rules generated by these three different data mining methods basically include the attributes of gender, age, location, burr, shape and frosted glass density as the main basis for diagnosing peripheral lung cancer. Conclusion Data mining has potential applications in medical imaging diagnosis and differential diagnosis.