利用深度学习增强专家从 12 导联心电图对早期冠状动脉闭塞的检测
应用统计
2019-11-19 v4 人工智能
定量方法
摘要
基于心电图（ECG）表现对急性冠状动脉闭塞进行早期诊断，对于及时实施直接经皮冠状动脉介入治疗至关重要。当前的 ST 段抬高（STE）标准具有特异性但敏感性不足。因此，很可能许多患者错失了潜在挽救生命的治疗。专家将非特异性 ECG 改变与 STE 相结合，以更高的敏感性检测缺血，但代价是特异性降低。我们表明，深度学习模型检测由急性冠状动脉闭塞引起的缺血，比 STE 标准、现有计算机分析器或心内科专家具有更好的敏感性与特异性平衡。
引用
@article{arxiv.1903.04421,
title = {Augmenting expert detection of early coronary artery occlusion from 12 lead electrocardiograms using deep learning},
author = {Rob Brisk and Raymond R Bond. Dewar D Finlay and James McLaughlin and Alicja Piadlo and Stephen J Leslie and David E Gossman and Ian B A Menown and David J McEneaney},
journal= {arXiv preprint arXiv:1903.04421},
year = {2019}
}
备注
Our attempts to produce what we considered to be an acceptable level of explainability from our algorithm have not yielded a satisfactory account of its internal logic and we do not feel this is acceptable from a clinical application. We will publish a fuller account of our work on this issue and its implications on the validation of clinical deep learning algorithms in the near future