Use of Machine Learning to Identify Follow-Up Recommendations in Radiology Reports.
Use of Machine Learning to Identify Follow-Up Recommendations in Radiology Reports.
复制标题
DOI：
10.1016/j.jacr.2018.10.020
复制
发表时间：
2019-03
期刊：
影响因子：
--
通讯作者：
Khorasani R
中科院分区：
文献类型：
--
作者：
Carrodeguas E;Lacson R;Swanson W;Khorasani R
评估放射学报告中的随访建议，开发和评估传统机器学习（TML）和深度学习（DL）模型以确定随访，并将其与自然语言处理（NLP）系统进行基准测试。这项符合HIPAA标准、经IRB批准的研究在一家学术医疗中心进行，每年产生> 500，000份放射学报告。对2016年生成的1,000份随机选择的超声、X射线、计算机断层扫描和磁共振成像报告进行了手动审查和注释，以提供后续建议。传统的机器学习（支持向量机、随机森林、逻辑回归）和深度学习（递归神经网络）算法在850份报告（训练数据）上构建和训练，随后优化模型架构和参数。精确度，召回率和F1分数计算其余150份报告（测试数据）。之前开发和验证的NLP系统（iSCOUT）也应用于测试数据，并计算了等效度量。12.7%的报告有后续行动建议。TML算法在测试数据上获得了0.75（随机森林）、0.83（逻辑回归）和0.85（支持向量机）的F1分数。DL Recurrent Neural Nets的F1评分为0.71; iSCOUT的F1评分也为0.71。通过F1分数的TML和DL方法的性能在训练时500-700个样本后出现平台。TML和DL是确定后续建议的可行方法。这些方法对于放射学报告中的随访建议的近实时监测具有巨大潜力。
Assess follow-up recommendations in radiology reports, develop and assess traditional machine learning (TML) and deep learning (DL) models in identifying follow-up, and benchmark them against a natural language processing (NLP) system. This HIPAA-compliant, IRB approved study, was performed at an academic medical center generating >500,000 radiology reports annually. 1,000 randomly-selected ultrasound, x-ray, computed tomography and magnetic resonance imaging reports generated in 2016 were manually reviewed and annotated for follow-up recommendations. Traditional machine learning (Support Vector Machines, Random Forest, Logistic Regression) and deep learning (Recurrent Neural Nets) algorithms were constructed and trained on 850 reports (training data), with subsequent optimization of model architectures and parameters. Precision, recall and F1-score were calculated on the remaining 150 reports (test data). A previously-developed and validated NLP system (iSCOUT) was also applied to the test data, with equivalent metrics calculated. 12.7% of reports had follow-up recommendations. The TML algorithms achieved F1 scores of 0.75 (Random Forest), 0.83 (Logistic Regression), and 0.85 (Support Vector Machine) on the test data. DL Recurrent Neural Nets had an F1 score of 0.71; iSCOUT also had an F1 score of 0.71. Performance of both TML and DL methods by F1-scores appeared to plateau after 500–700 samples while training. TML and DL are feasible methods to identify follow-up recommendations. These methods have great potential for near real-time monitoring of follow up recommendations in radiology reports.
登录
查看更多内容
影响因子：
4.4
作者：
Zhou, Yihua;Amundson, Per K.;Wippold, Franz J.
通讯作者：
Wippold, Franz J.
影响因子：
19.7
作者：
Sistrom, Christopher L.;Dreyer, Keith J.;Thrall, James H.
通讯作者：
Thrall, James H.
影响因子：
6.2
作者：
Dutta, Sayon;Long, William J.;Reisner, Andrew T.
通讯作者：
Reisner, Andrew T.
DOI：
10.1377/hlthaff.27.6.1491
发表时间：
2008-11
期刊：
Health affairs (Project Hope)
影响因子：
--
作者：
Smith-Bindman R;Miglioretti DL;Larson EB
通讯作者：
Larson EB
影响因子：
4.5
作者：
Lacson, Ronilda;Prevedello, Luciano M.;Khorasani, Ramin
通讯作者：
Khorasani, Ramin