| 摘要: | 近年來，根據世界衛生組織(WHO)的統計，心血管疾病已經成為導致死亡和殘疾的重要原因之一，最常見的心血管疾病則為心律不整。通常，臨床醫生通過觀察長時間的心電圖（ECG）進行診斷是相當耗時且困難的。現在，良好的醫護環境逐漸受到重視，但醫療資源依然有限。幸運的是，今天深度學習的發展在影像識別和生物醫學訊號分析領域取得了巨大成功。 卷積神經網絡（CNN）是其中一種著名的深度學習方法，它具有先進的特徵自動提取和強健性。 在這個研究中，我們開發了一種新穎且高精度的CNN系統，用於心房顫動（Atrial Fibrillation, AFIB），心室期外收縮（Premature ventricular contraction, PVC），心房期外收縮（Premature atrial contraction, PAC），左束支傳導阻滯（Left bundle branch block, LBBB），右束支傳導阻滯（Right bundle branch block, RBBB），心室顫動（Ventricular Fibrillation,VFIB） 和心室性心動過速（Ventricular Tachycardia, VT）7類心電圖疾病和正常竇性心律（Normal sinus rhythm, NSR）的辨識，在這項研究中，所有用於訓練和測試的心電圖數據皆取自於MIT-BIH數據庫。 本系統對上述七種類型的心臟疾病和正常的心電圖數據進行分類，準確度達到95％。 這項研究證明了其在臨床應用中的可行性，加以改進後未來可作為臨床醫師診斷的輔助工具。;In recent years, cardiovascular disease has become the leading cause of death and disability according to the statistics by the World Health Organization. The most common form of cardiovascular disease being arrhythmia. Sometimes, it becomes time-consuming and difficult for clinicians to observe electrocardiogram (ECG) and analyze the arrhythmia. Fortunately, deep learning has brought great success in the fields of image recognition and biomedical signal analysis. The Convolutional Neural Network (CNN) is a such famous method of deep learning with advanced automatic feature extraction and robustness. In this work, we developed a novel system of CNN for automatic detection of arrhythmia based on ECG signals. The ECG signals were obtained from a publicly available arrhythmia database. We have obtained Normal sinus rhythm (NSR), Atrial Fibrillation(AFIB), Premature ventricular contraction (PVC), Premature atrial contraction (PAC), Left bundle branch block(LBBB), Right bundle branch block(RBBB), Ventricular Fibrillation (VFIB) and Ventricular Tachycardia (VT) ECG data from MIT-BIH arrhythmia database which includes recordings of many common and life-threatening arrhythmias along with clinical annotation. Our system has achieved an detection accuracy of 95% for the aforementioned seven types of arrhythmia. Hence, it is evident that our work has potential to be implemented in clinical settings to serve as an adjunct tool for the doctor. Furthermore, this work will be improved to achieve better performance in the future. |
