基於深度學習的心臟疾病的分類 = Classification of heart diseases based on deep learning / 李孟珅.
- 作者： 李孟珅
- 其他題名：
- Classification of heart diseases based on deep learning
- 主題： 心電圖 心律不整 深度學習 卷積神經網路. , electrocardiogram arrhythmia deep learning convolutional neural network.
- URL：
電子資源
- 一般註：指導教授: 陳瓊安, 林君玲. 學年度: 113.
- 書目註：參考書目： 葉.
-
讀者標籤：
- 系統號： 005185101 | 機讀編目格式
館藏資訊
摘要註
心電圖（Electrocardiogram, ECG）是診斷心律不整的關鍵工具，能夠根據心臟在收縮與舒張過程中產生的電訊號變化，有效偵測異常節律。本研究設計了一種基於深度學習的卷積神經網路（Convolutional Neural Network, CNN）模型，用於分類五種類型的心跳。研究以 MIT- BIH 心律不整資料庫為基礎，專注於 MLII 通道，採用短時傅立葉變換 （Short-Time Fourier Transform, STFT）將一維信號轉換為二維時頻譜圖作為模型輸入。實驗結果顯示，本方法達成了準確率 86.16%、靈敏度 86.16%、特異性 96.54%；相較於傳統心律不整檢測方法，其準確率提升 7%至 38%，靈敏度提高 38%，展現出更優的分類性能與臨床應用價值。. Electrocardiogram (ECG) is a key tool for diagnosing cardiac arrhythmia. It can effectively detect abnormal rhythms based on the changes in electrical signals generated by the heart during contraction and relaxation. In this study, a deep learning-based convolutional neural network (CNN) model was designed to classify five types of heartbeats.The study is based on the MIT-BIH arrhythmia database, focusing on the MLII channel, and using short-time Fourier transform (STFT) to convert one-dimensional signals into two-dimensional spectrograms as model input. Experimental results show that this method achieved an accuracy of 86.16%, a sensitivity of 86.16%, and a specificity of 96.54%; compared with traditional arrhythmia detection methods, its accuracy was increased by 7% to 38%, and its sensitivity was increased by 38%, showing better classification performance and clinical application value..