RT期刊文章SR电子T1开发和外部验证的学习算法来识别和定位蛛网膜下腔出血CT扫描摩根富林明神经病学神经学乔FD Lippincott Williams &威尔金斯SP e1257 OP e1266 10.1212 / WNL。半岛投注体育官网100签证官0000000000201710是12 A1安东尼奥Thanellas A1海基Peura A1 Mikko Lavinto A1托米- Ruokola A1莫伊拉Vieli A1 Victor e . Staartjes A1塞巴斯蒂安Winklhofer A1卡罗塞拉A1卢卡Regli A1 Miikka Korja年2023 UL //www.ebmtp.com/content/100/12/e1257.abstract半岛投注体育官网 AB医学成像的背景和目标,数量有限的培训深入学习算法外部验证,公开发布。我们假设深学习算法可以被训练识别和定位蛛网膜下腔出血(SAH)头部电脑断层扫描(CT)扫描和训练模式时执行令人满意地使用外部和真实数据进行了测试。方法我们用头部CT图像的病人承认赫尔辛基大学医院在2012年和2017年之间。我们手动分割(即。,划定)长官在90头CT扫描和使用分割CT扫描在一起22负(SAH)控制CT扫描在训练一个开源卷积神经网络(U-Net)来识别和定位长官。然后我们测试训练算法的性能通过使用外部数据集(137 SAH和1242控制情况下)中收集2国外也通过创建一个数据集的连续紧急头部CT扫描(8 SAH和511例)控制在5种不同表现在待命时间国内医院2021年9月。我们评估了算法的能力来识别SAH病人通过计算限制电平性能指标,如敏感性和特异性。结果在外部验证组1379例,137例SAH的算法确定了136正确(敏感性99.3%,特异性63.2%)。49064轴头CT片,该算法识别和本地化SAH的2110片1845年长官(敏感性87.4%,特异性95.3%)。519年连续紧急头部CT扫描成像在2021年9月,该算法正确地识别所有8例SAH(敏感性100.0%,特异性75.3%)。 The slice-level (27,167 axial slices in total) sensitivity and specificity were 87.3% and 98.8%, respectively, as the algorithm identified and localized SAH in 58 of 77 slices with SAH. The performance of the algorithm can be tested on through a web service.Discussion We show that the shared algorithm identifies SAH cases with a high sensitivity and that the slice-level specificity is high. In addition to openly sharing a high-performing deep learning algorithm, our work presents infrequently used approaches in designing, training, testing, and reporting deep learning algorithms developed for medical imaging diagnostics.Classification of Evidence This study provides Class III evidence that a deep learning algorithm correctly identifies the presence of subarachnoid hemorrhage on CT scan.CE=Conformité Européenne; DICOM=Digital Imaging and Communications in Medicine; HUH=Helsinki University Hospital; ICD=International Classification of Diseases; MPR=multiplanar reformatted; NIfTI=Neuroimaging Informatics Technology Initiative; PACS=Picture Archiving and Communication Systems; SAH=subarachnoid hemorrhage; U-Net=neural network

map