| DC 欄位 |
值 |
語言 |
| DC.contributor | 跨領域轉譯醫學研究所 | zh_TW |
| DC.creator | 蕭成儀 | zh_TW |
| DC.creator | Cheng-Yi Hsiao | en_US |
| dc.date.accessioned | 2024-1-30T07:39:07Z | |
| dc.date.available | 2024-1-30T07:39:07Z | |
| dc.date.issued | 2024 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:444/thesis/view_etd.asp?URN=106888003 | |
| dc.contributor.department | 跨領域轉譯醫學研究所 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 心房顫動（Atrial fibrillation，簡稱AF）是最常見的心律不整 (arrhythmia)，許多患者因缺乏明顯症狀而未及時就診接受檢查與治療。先前的研究顯示利用大型資料集建構的深度神經網路（deep neural network，簡稱DNN），能透過12導程心電圖（electrocardiogram，簡稱ECG）篩檢出在竇性心律（sinus rhythm，簡稱SR）下的心房顫動。然而，這些訓練好的模型不僅取得不易，使用對外公開的小型資料集也很難重現相同的結果。本研究所提出的方法，能針對數量有限的資料提取有意義的特徵，藉以偵測在竇性心律下的心房顫動。我們從臺北醫學大學附設醫院的64,196位病人中，蒐集到94,224筆12導程心電圖資料，從中挑選213位病人在心房顫動診斷前的竇性心律心電圖，並隨機選擇247位年齡配對的非心房顫動受試者作為對照。我們所開發的訊號處理技術，簡稱MA-UPEMD，利用主成分分析（principal component analysis，簡稱PCA）與導程間關聯性，可從12導程心電圖中分離出P波，並且定量其時序與空間特徵。合併這些特徵後，機械學習模型產生之AUC為0.64。我們的研究顯示即使只有如此少量的資料，使用我們開發的方法，也能描繪出表現心房電活性的P波，所提取特徵的效能亦優於帶通濾波器 (bandpass filter) 或深度神經網路擷取之P波。我們所提供的方法具生理可解釋性及再現性，能辨識在竇性心律下的心房顫動患者。 | zh_TW |
| dc.description.abstract | Many AF patients are not properly diagnosed and treated due to the absence of evident signs or symptoms. Researchers used deep neural networks (DNN) with large datasets of 12-lead electrocardiograms (ECG) to detect AF during sinus rhythm (SR). However, these models are not only publicly inaccessible, but similar results are also irreproducible with small datasets. In our study, we established a method to identify AF during SR with explainable features extracted from limited data. We collected 94,224 12-lead ECG recordings from 64,196 patients who visited the Taipei Medical University Hospital. Two hundred and thirteen patients with SR ECG before AF diagnosis were selected, and 247 age-matched normal subjects were also selected to serve as controls. The technique we developed, MU-UPEMMD, extracts P-wave from 12-lead ECG and quantifies its spatiotemporal features by principal component analysis (PCA) and inter-lead relationships. With the combination of these features, the machine learning (ML) model yielded an AUC of 0.64. Our study demonstrates that MU-UPEMMD can delineate P-waves that represent the electrical activity of the atrium even with limited data. The performance of features extracted from P-wave is also superior to those from bandpass filter or DNN. The proposed approach can classify AF patients during SR with results that are physiologically explainable and reproducible. | en_US |
| DC.subject | 心房顫動 | zh_TW |
| DC.subject | 12導程心電圖 | zh_TW |
| DC.subject | 主成分分析 | zh_TW |
| DC.subject | P 波迴路 | zh_TW |
| DC.subject | atrial fibrillation | en_US |
| DC.subject | 12 lead ECG | en_US |
| DC.subject | principal component analysis | en_US |
| DC.subject | P loop | en_US |
| DC.title | 自12導程心電圖擷取P波特徵辨識竇性心律下之 心房顫動高風險病患 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Identification of patients with potential atrial fibrillation during sinus rhythm using isolated P wave characteristics from 12-lead ECGs | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |