基於混合深度學習模型之血液透析患者血壓預測;Blood Pressure Prediction for Hemodialysis Patients Based on a Hybrid Deep Learning Model

NCUIR > College of Electrical Engineering & Computer Science > Executive Master of Communication Engineering > Electronic Thesis & Dissertation > Item 987654321/96271

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/96271

Title:	基於混合深度學習模型之血液透析患者血壓預測;Blood Pressure Prediction for Hemodialysis Patients Based on a Hybrid Deep Learning Model
Authors:	許寬廷;Hsu, Kuan-Ting
Contributors:	通訊工程學系在職專班
Keywords:	血壓預測;血液透析;卷積神經網路;循環神經網路;注意力機制;Blood Pressure Prediction;Hemodialysis;Convolutional Neural Networks;Recurrent Neural Networks;Attention Mechanism
Date:	2025-03-21
Issue Date:	2025-04-09 17:31:18 (UTC+8)
Publisher:	國立中央大學
Abstract:	血液透析患者的血壓波動受多種生理與治療因素影響，若無法即時掌握變化趨勢，可能導致患者不適並增加臨床風險。因此，準確預測透析過程中的下一次收縮壓值，對於提升病患安全性與優化臨床決策至關重要。當多名患者同時出現血壓異常時，準確的即時預測可幫助醫護人員根據病情嚴重程度進行風險分層與資源調度，確保高風險患者獲得優先處置。本研究提出混合深度學習模型 (Hybrid Deep Learning Model)，結合卷積神經網路 (CNN)、長短期記憶網路 (LSTM) / 門控循環單元 (GRU) 及注意力機制 (Attention Mechanism) 以提升血壓預測準確性。CNN 用於擷取局部特徵，RNN 建模長期趨勢，而 Attention 機制則動態調整時間步權重，使模型聚焦於關鍵資訊，進而提高預測精度。實驗結果顯示，該混合模型能有效提升血壓預測準確性，協助醫護人員快速識別高風險患者並優化臨床資源分配，提高患者安全性與透析照護效率。未來可進一步發展個人化病患模型、應用 Transformer 技術、整合穿戴式裝置數據，並推動模型於臨床應用，以提升血液透析患者的健康管理與治療品質。 ;Blood pressure fluctuations in hemodialysis patients are influenced by various physiological and treatment-related factors. Failure to promptly monitor these changes may lead to patient discomfort and increased clinical risks. Therefore, accurately predicting the next systolic blood pressure (SBP) value during dialysis is crucial for enhancing patient safety and optimizing clinical decision-making. When multiple patients experience abnormal blood pressure fluctuations simultaneously, accurate real-time predictions enable healthcare professionals to prioritize high-risk patients and allocate resources efficiently. This study proposes a Hybrid Deep Learning Model that integrates Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) / Gated Recurrent Units (GRUs), and the Attention Mechanism to improve blood pressure prediction accuracy. CNNs extract local features, RNNs capture long-term trends, and the Attention Mechanism dynamically adjusts weight distribution across time steps, allowing the model to focus on critical information and enhance prediction precision. Experimental results demonstrate that the proposed hybrid model significantly improves blood pressure prediction accuracy, assisting healthcare professionals in identifying high-risk patients and optimizing clinical resource allocation. Future research may focus on developing personalized patient models, incorporating Transformer architectures, integrating wearable device data, and advancing clinical applications to enhance health management and treatment quality for hemodialysis patients.
Appears in Collections:	[Executive Master of Communication Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	23	View/Open

社群 sharing

Loading...