基于磁心电图参数的胸痛患者冠心病检测：机器学习模型。
Detection of coronary artery disease in patients with chest pain: A machine learning model based on magnetocardiography parameters.
机构信息
Department of Cardiology, The 8th Medical Center, Chinese PLA General Hospital, Bejing, China.
出版信息
Clin Hemorheol Microcirc. 2021;78(3):227-236. doi: 10.3233/CH-200905.
BACKGROUD
Patients with chest pain and suspected of coronary artery disease(CAD) need further test to confirm the diagnosis. Magnetocardiography (MCG) is a non-invasive and emission-free technology which can detect and measure the weak magnetic fields created by the electrical activity of the heart.
OBJECTIVE
This study aimed to investigate the usefulness of the 10 MCG parameters to detect CAD in patients with chest pain by means of a machine learning method of multilayer perceptron(MLP) neural network.
METHODS
209 patients who were suffering from chest pain and suspected of CAD were enrolled in this cross-sectional study. In all patients, 12-lead electrocardiography(ECG) and MCG test were performed before percutaneous coronary angiography(PCA). 10 MCG parameters were analyzed by MLP neural networks.
RESULTS
11 diagnostic models(M1 to M11) were established after MLP analysis. The accuracies ranged from 71.2% to 90.5%. Two models(M10 and M11) were further analyzed. The accuracy, sensitivity, specificity, PPV, NPV, PLR and NLR were 89.5%, 89.8%, 88.9%, 92.7%, 84.7%, 11.10 and 0.11, of M10, and were 90.0%, 91.4%, 87.7%, 92.1%, 86.6%, 7.43 and 0.10, of M11.
CONCLUSIONS
By a method of MLP neural network, MCG is applicable in identifying CAD in patients with chest pain, which seems beneficial for detection of CAD.
背景
胸痛且疑似患有冠心病（CAD）的患者需要进一步的检查来确诊。磁心电图（MCG）是一种非侵入性、无辐射的技术，可以检测和测量心脏电活动产生的微弱磁场。
目的
本研究旨在通过多层感知器（MLP）神经网络的机器学习方法，探讨 10 个 MCG 参数在胸痛患者中检测 CAD 的有用性。
方法
本横断面研究纳入了 209 例胸痛且疑似 CAD 的患者。所有患者均在经皮冠状动脉造影术（PCA）前行 12 导联心电图（ECG）和 MCG 检查。通过 MLP 神经网络分析 10 个 MCG 参数。
结果
MLP 分析后建立了 11 个诊断模型（M1 至 M11）。准确率范围为 71.2%至 90.5%。进一步分析了两个模型（M10 和 M11）。M10 的准确率、敏感度、特异度、PPV、NPV、PLR 和 NLR 分别为 89.5%、89.8%、88.9%、92.7%、84.7%、11.10 和 0.11，M11 的准确率、敏感度、特异度、PPV、NPV、PLR 和 NLR 分别为 90.0%、91.4%、87.7%、92.1%、86.6%、7.43 和 0.10。
结论
通过 MLP 神经网络方法，MCG 可用于识别胸痛患者中的 CAD，这似乎有利于 CAD 的检测。
Zotero 插件