《计算机应用研究》|Application Research of Computers

肿瘤类疾病的过度与错误医疗检查控制机制与模型的研究

Study on evaluation mechanism of excessive treatment and misdiagnosis of tumor diseases

免费全文下载 (已被下载 次)  
获取PDF全文
作者 朱诗生,汪昕蓉,毛礼厅,柳学国
机构 1.汕头大学 计算机系,广东 汕头 515063;2.中山大学附属第五医院,广东 珠海 519000
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2019)05-031-1428-05
DOI 10.19734/j.issn.1001-3695.2018.04.0267
摘要 针对当前肿瘤类疾病诊治过程中存在的错误与过度医疗问题,本研究基于医疗大数据提取出相似病案专家处方中的影像信息,利用机器学习分类模型提出了发现错误与过度诊治的检查控制机制与解决方案。该方案依托医院长期积累的各类肿瘤疾病病历中的CT、MRI图像,以每次诊疗过程中的实际肿瘤类型为依据,从医疗数据库中选择对应类型的影像数据进行特征提取、特征选择、模型构建,得到该类型肿瘤的预测分类器,预测当前病例的良恶性;并通过跟医生诊断结果的对比判断诊疗过程中是否存在过度与错误医疗问题。其核心是提高不依赖人工判别方法的判别正确率来降低肿瘤类疾病的错诊可能性,通过实验证明结合了Spearman去冗余方法的SVM_RFE降维,与传统的SVM_RFE方法相比,在肺结节良恶性分类问题的SVM模型中表现更佳,同时也优于传统的radiomics方法。该方案能及时发现错误与过度医疗问题并提出预警,发挥监督提醒的作用,在实现预防和避免诊治错误的同时减少对人工鉴别的依赖,为错误医疗问题及减轻患者负担提供一种新的解决途径。
关键词 错误医疗; 机器学习; Spearman; SVM_RFE; SVM分类模型
基金项目 广东省科技计划资助项目(20140401)
本文URL http://www.arocmag.com/article/01-2019-05-031.html
英文标题 Study on evaluation mechanism of excessive treatment and misdiagnosis of tumor diseases
作者英文名 Zhu Shisheng, Wang Xinrong, Mao Liting, Liu Xueguo
机构英文名 1.Dept. of Computer,Shantou University,Shantou Guangdong 515063,China;2.The 5th Affiliated Hospital of Sun Yat-Sen University,Zhuhai Guangdong 519000,China
英文摘要 Aiming at solving the problem of erroneous and excessive medical treatment of tumor diseases, this paper extracted image information from similar medical record experts′ prescriptions based on medical big data, and used a machine learning classification model to quantitatively analyze the level of medical treatment. The program relies on the CT, MRI images of tumor diseases accumulated in the hospital over a long period of time. Based on the tumor type in each treatment case, it selected the corresponding type of image data from the medical database for feature extraction and feature selection, to construct models to obtain a predictor for this tumor disease, to predict the benign and malignant of the current case, and to determine whether there were excessive and erroneous medical problems in the diagnosis process by comparing with the results of the doctor′s diagnosis. The core of the method is to improve the accuracy of discriminating tumors without relying on human beings. Compared with the traditional SVM_RFE, the investigated method which combined SVM_RFE with the Spearman correlation is experimentally proved to perform better in the SVM model of benign and malignant classification of pulmonary nodules. In ge-neral, it offered better performance compared with traditional radiomics method. The solution can detect erroneous and excessive medical treatment issues in real-time and provide warning, which can play a role in supervision and reminding. It potentially reduces the reliance on manual identification and minimizes the burden on patients while preventing and avoiding errors in diagnosis and treatment.
英文关键词 error treatment; machine learning; Spearman; SVM_RFE; SVM classifier model
参考文献 查看稿件参考文献
 
收稿日期 2018/4/7
修回日期 2018/5/17
页码 1428-1432
中图分类号 TP391
文献标志码 A