Atashi Alireza, Tohidinezhad Fariba, Dorri Sara, Nazeri Najmeh, Ghousi Rouzbeh, Marashi Sina, Hajialiasgari Fatemeh
E-Health Department, Virtual School, Tehran University of Medical Sciences, Tehran, Iran.
Department of Medical Informatics, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran.
Stud Health Technol Inform. 2019 Jul 4;262:142-145. doi: 10.3233/SHTI190037.
The aim is to recognize the unknown atterns in a real breast cancer dataset using data mining algorithms as a new method in medicine. Due to excessive missing data in the collection only data on 665 of 809 patients were available. The other missing values were estimated using the EM algorithm in SPSS21 software. Fields have been converted into discrete fields and finally the APRIORI algorithm has been used to analyze and explore the unknown patterns. After the rule extraction, experts in the field of breast cancer eliminated redundant and meaningless relations. 100 association rules with a confidence value of more than 0.9 explored by the APRIORI algorithm and after the clinical expert feedback, 10 clinically meaningful relations have been detected and reported. Due to the high number of risk factors, the use of data mining is effective for cancer data. These patterns provide the future study hypotheses of specific clinical studies.
目的是使用数据挖掘算法作为医学中的一种新方法，识别真实乳腺癌数据集中的未知模式。由于数据收集过程中存在大量缺失数据，809名患者中仅有665名患者的数据可用。其他缺失值使用SPSS21软件中的期望最大化（EM）算法进行估计。字段已转换为离散字段，最后使用APRIORI算法分析和探索未知模式。在规则提取之后，乳腺癌领域的专家消除了冗余和无意义的关系。APRIORI算法探索出100条置信度值大于0.9的关联规则，经过临床专家反馈，检测并报告了10条具有临床意义的关系。由于风险因素数量众多，数据挖掘在癌症数据方面的应用是有效的。这些模式为特定临床研究提供了未来的研究假设。
