藉由機器學習判别伴隨心房顫動之中風病患的嚴重程度

Abstract

心律不整是一個當今老化的社會中盛行的心血管疾病其中一個嚴峻的課題，嚴重的情況可能導致死亡。其中最常見型態的心律不整為心房顫動。 房顫是由於心臟組織失調的電位活動，導致心臟異常的收縮甚至產生顫動的情形。然而，有關於中風相關的疾病背後的病理學牽涉複雜，因此我們排除其他非缺血性中風的其他病理學所導致的中風。 為了描述這個問題，我們採用由臨床內科醫師確診為心房顫動暨中風症狀的病患的來自加護病房生理訊號，探索使用機器學習的方法來判別病患的中風嚴重程度。 我們的目標是試圖探索那些特徵是有效足以讓我們能夠有效鑑別不同中風嚴重程度的病患。 我們研究的發現即時血壓分析之於彌補心率變異性分析（HRV）的重要性，特別是心電圖(ECG)的量測。如果沒有涉及到與收縮壓在時間和資訊熵域條件特徵，我們很難泛化各種背後複雜的病理學。此外，光體積描記法（PPG）對於研究血流速率和心血管健康也是至關重要。

Arrhythmia is emerging as a significant cardiovascular disease in our modern aging society and can lead to fatal outcomes; One of the most prevalent types of arrhythmia is Atrial Fibrillation (AF). AF originates from abnormal discharges in cardiac tissue, resulting in incomplete atrial contraction and atrial fibrillation. To address this issue, we analysed the recordings of acute ischaemic stroke patients associated with AF symptoms diagnosed by clinicians for further analysis and explored the methodology to classify them into different severity of stroke by Machine Learning Technique. Our goal is to explore what features are significant for distinguishing the different outcomes of this patient group. Our study findings underscore the critical role of real-time blood pressure analysis in complementing Heart Rate Variability (HRV), especially in ECG measurements. Without incorporating haemodynamic condition features related to Systolic blood pressure in both Time and Entropy domains, it's difficult to generalise the complex metrics underlying pathology. Additionally, Photoplethysmography (PPG) is indispensable for investigating flow rate and cardiovascular health.