臺灣博碩士論文加值系統

中文摘要
論文名稱：慢波睡眠與心率變異性的AI模型之可行性研究
臺北醫學大學醫學院人工智慧醫療碩士在職專班
研究生姓名： 陳家祺
指導教授：蘇家玉 臺北醫學大學 教授

研究背景：傳統上評估慢波睡眠的情形需依賴睡眠多項生理檢查，這種檢查過程繁瑣對患者而言，頗為不便，這限制了一些患者接受睡眠檢查的意願。為解決此一問題，本研究目的在探討一種基於心率變異性的人工智慧模型，用以預測慢波睡眠。透過應用機器學習方法或深度學習技術，提供一種更為簡便且可靠的方法，來獲取和預測睡眠週期中慢波睡眠的實際分佈。

研究方法：本研究爲回溯性研究。資料來源為衛生福利部的雙和醫院
睡眠中心，505位受試者的睡眠多項生理檢查記錄，以睡眠多項生理檢查記錄進行了連續的心率變異性、血氧飽和度和電腦自動偵測慢波，頻率為0.4Hz ~ 4Hz。然後根據睡眠技師判定的睡眠分期記錄，以是否出現慢波數據進行標記。模型的特徵包括平均心率、九個心率變異性指標、和二個血氧飽和度指標；然後將數據集分為80%的訓練集和20%的測試集，使用四種機器學習方法與三種基於深度學習的時間序列模型進行模型訓練。在模型評估過程中，採用5折交叉驗證來計算模型的平均準確率（Accuracy）、接收者操作特徵曲綫下面積（AUROC）和精確率-召回率曲綫下面積（AUPRC），用以評估在這三個指標上，表現最佳的模型。另外，通過田口方法有效地優化多組合模型的最佳超參數組合。

研究結果：第三組實驗為最佳。數據集共有271,125筆，包含非慢波睡
眠事件共計188,931筆及慢波睡眠事件共計82,194筆。經由5折交叉驗證後，取模型評估指標Accuracy，AUROC及 AUPRC分別計算平均值，共有8個組合模型達到預期目標。機器學習方法以隨機森林（Random Forest）模型，應用隨機過採樣（Random Oversampling）效果最佳，達到的評估指標，Accuracy = 86.78%、AUROC = 0.934、AUPRC = 0.953。深度學習方法以長短期記憶（LSTM）模型，應用SMOTE Oversampling為最佳，Accuracy = 84.07%、AUROC = 0.915、AUPRC = 0.933。

研究數據顯示，以心率變異性作為人工智慧模型的特徵，適合用於預測慢波睡眠。這不僅提升了居家檢查的可行性和患者的接受度，還增强了醫生進行長期監測的能力。

ABSTRACT
Title of Thesis：Feasibility Study of an AI Model for Slow-Wave
Sleep and Heart Rate Variability
Author：Chia-Chi Chen
Thesis advised by：Emily Chia-Yu Su
Taipei Medical University,
Professional Master Program in Artificial Intelligence in Medicine

Traditionally, assessing Slow-Wave Sleep relies on Polysomnography, a cumbersome and inconvenient procedure for patients, limiting some patients' willingness to undergo sleep studies. To address this issue, this research aims to explore an Artificial Intelligence model based on Heart Rate Variability indicators to predict Slow-Wave Sleep. By applying Machine Learning and Deep Learning techniques, a more convenient and reliable method is provided to obtain and predict the actual distribution of Slow-Wave Sleep during sleep cycles.

This retrospective study sourced data from the Sleep Center at Shuang Ho Hospital, Ministry of Health and Welfare, involving 505 participants with Polysomnography records. Continuous Heart Rate Variability, Oxygen Saturation , and automatic detection of Slow Waves, specifically Slow Waves (0.4Hz - 4Hz), were recorded. Sleep stages were marked based on whether Slow Waves data appeared, as determined by sleep technicians. The model's features included mean Heart Rate , nine Heart Rate Variability indicators, and two Oxygen Saturation indicators. The dataset was divided into an 80% training set and a 20% testing set, using four machine learning methods and three deep learning-based time series models for training. Model evaluation involved 5-fold cross-validation to calculate average Accuracy, Area Under the Receiver Operating Characteristic Curve (AUROC), and Area Under the Precision-Recall Curve (AUPRC), identifying the best-performing models. Furthermore, the Taguchi method was employed to optimize the best hyperparameter combination for multi-model ensembles.

Research Results: The third experiment was the best overall. The dataset comprises a total of 271,125 records, including 188,931 records of non-slow-wave sleep events and 82,194 records of slow-wave sleep events. Using 5-fold cross-validation, the average values of the model evaluation metrics Accuracy, AUROC, and AUPRC were calculated. There are eight ensemble models that have achieved the expected goal. Among machine learning methods, the Random Forest model with Random Oversampling performed the best, achieving the following evaluation metrics: Accuracy：86.78%, AUROC：0.934, and AUPRC：0.953. For deep learning methods, the LSTM model with SMOTE Oversampling was the best performer, achieving the following evaluation metrics: Accuracy：84.07%, AUROC：0.915, and AUPRC：0.933.

The research data indicates that using heart rate variability as a feature for artificial intelligence models is suitable for predicting slow-wave sleep. This not only enhances the feasibility of home examinations and patient acceptance but also improves the ability of doctors to conduct long-term monitoring.

目錄
審定書 i
誌謝 ii
目錄 iii
表目錄 v
中文摘要 viii
ABSTRACT x
第一章 緒論 1
1-1研究背景 1
1-2研究動機 2
1-3研究目的 3
第二章 文獻回顧與探討 4
2-1睡眠分期（Sleep Stage） 4
2-2慢波睡眠（SWS） 4
2-3心率變異性（Heart Rate Variability, HRV） 5
2-4時間序列數據 7
2-5心電圖（ECG） 8
2-6基於深度學習的時間序列模型 9
2-7田口方法（Taguchi Methods） 10
2-8文獻探討 13
第三章 研究資料與方法 15
3-1資料來源 15
3-2睡眠多項生理檢查（PSG） 15
3-3研究方法 16
3-4特徵數據分析 19
3-5人工智慧模型 31
3-6模型訓練統計方法 31
3-7數據預處理 32
3-8數據不平衡解決方法 33
3-9超參數最佳化 34
第四章 研究結果 36
4-1模型超參數 36
4-2原始數據集訓練結果 46
4-3使用SMOTE Oversampling訓練結果 47
4-4 Random Oversampling訓練結果 49
4-5 Random Downsampling 50
4-6模型訓練時間比較 51
第五章 研究討論與結論 52
5-1研究討論 52
5-2研究限制 54
5-3結論 54
參考文獻 56
IRB證明函 59

表目錄

表1 L9（34）正交陣列 12
表2 SW的時間序列窗口、總和時間與註記 20
表3 新增特徵時間序列窗口欄位及SW Event標注欄位 21
表4 PSG檢查分類 26
表5 事件的起始時間及結束時間 27
表6 SW Events記錄 27
表7 SW Events匹配EKG起始時間 28
表8 EKG時間序列 28
表9 以時間軸調整睡眠分期位置 28
表10 合併慢波發生位置註記、特徵時間窗口及標註 29
表11 完成標注的二維矩陣數據 30
表12 機器學習法各個模型的超參數 36
表13 深度學習法各個模型的超參數 36
表14 機器學習法各個模型的最佳化超參數 37
表15 LSTM模型與GRU模型的L9（34）正交陣列 39
表16 InceptionTime模型的L9（34）正交陣列 39
表17 LSTM模型應用SMOTE Oversampling實驗數據 40
表18 Accuracy指標噪聲比SN（1） 41
表19 Accuracy指標的因子反應分析表 42
表20 AUROC指標的因子反應分析表 43
表21 GRU模型應用SMOTE Oversampling在田口實驗組合最佳組合及實驗數據 44
表22 GRU模型應用SMOTE Oversampling的Accuracy指標的最佳超參數組合 44
表23 Accuracy指標最佳超參數組合5-fold cross-validation 45
表24 深度學習法各個模型應用SMOTE Oversampling的最佳化超參數 45
表25 原始數據集訓練結果統計表 47
表26 執行SMOTE Oversampling所獲得之機器學習模型最佳化超參數 48
表27 執行SMOTE Oversampling所獲得之深度學習模型最佳化超參數 48
表28 執行SMOTE Oversampling所獲得之各模型訓練結果統計表 48
表29 執行Random Oversampling所獲得之機器學習模型最佳化超參數 49
表30 執行Random Oversampling所獲得之深度學習模型最佳化超參數 50
表31 執行Random Oversampling所獲得之各模型訓練結果統計表 50
表32 執行Random Downsampling所獲得之各模型訓練結果統計表 51
表33 組合模型的訓練時間比較表 51

圖目錄

圖1 連續的時間序列數據 7
圖2 數據收集和處理的流程圖 18
圖3 60秒的時間窗口和1秒的步長，計算連續HRV特徵 19
圖4 EKG片段訊號 22
圖5 不正確的R波峰波峰 22
圖6 噪聲訊號波峰 23
圖7 使用帶通濾波單一條件 24
圖8 使用帶通濾波及R-R間隔距離條件 24
圖9 正確R波波峰 25
圖10 Accuracy指標的因子反應分析圖 42
圖11 AUROC指標的因子反應分析圖 43

參考文獻

Ako, M., Kawara, T., Uchida, S., Miyazaki, S., Nishihara, K., Mukai, J., . . . Okubo, Y. (2003). Correlation between electroencephalography and heart rate variability during sleep. Psychiatry and clinical neurosciences, 57(1), 59-65.
Aldredge, J. L., & Welch, A. J. (1973). Variations of heart rate during sleep as a function of the sleep cycle. Electroencephalography and Clinical Neurophysiology, 35(2), 193-198.
Bakker, J. P., Ross, M., Vasko, R., Cerny, A., Fonseca, P., Jasko, J., . . . Anderer, P. (2021). Estimating sleep stages using cardiorespiratory signals: validation of a novel algorithm across a wide range of sleep-disordered breathing severity. Journal of Clinical Sleep Medicine, 17(7), 1343-1354.
Bobra, N. P., Wang, Z., Zhang, W., & Luo, A. (2013). A high-quality, low-energy, small-size system-on-chip (soc) solution enabling ECG mobile applications. Paper presented at the IECON 2013-39th Annual Conference of the IEEE Industrial Electronics Society.
Born, J., & Wilhelm, I. (2012). System consolidation of memory during sleep. Psychological research, 76, 192-203.
Cénat, J. M., Blais-Rochette, C., Kokou-Kpolou, C. K., Noorishad, P.-G., Mukunzi, J. N., McIntee, S.-E., . . . Labelle, P. R. (2021). Prevalence of symptoms of depression, anxiety, insomnia, posttraumatic stress disorder, and psychological distress among populations affected by the COVID-19 pandemic: A systematic review and meta-analysis. Psychiatry research, 295, 113599.
Chaudhuri, S., Pawar, T. D., Duttagupta, S., Chaudhuri, S., Pawar, T. D., & Duttagupta, S. (2009). Review of ECG Analysis. Ambulation Analysis in Wearable ECG, 15-26.
Chi, D.-J., & Chu, C.-C. (2021). Artificial intelligence in corporate sustainability: Using LSTM and GRU for going concern prediction. Sustainability, 13(21), 11631.
Herzig, D., Eser, P., Omlin, X., Riener, R., Wilhelm, M., & Achermann, P. (2018). Reproducibility of heart rate variability is parameter and sleep stage dependent. Frontiers in physiology, 8, 1100.
Holgado-Cuadrado, R., Plaza-Seco, C., Lovisolo, L., & Blanco-Velasco, M. (2023). Characterization of noise in long-term ECG monitoring with machine learning based on clinical criteria. Medical & Biological Engineering & Computing, 61(9), 2227-2240.
Huyett, P., Siegel, N., & Bhattacharyya, N. (2021). Prevalence of sleep disorders and association with mortality: results from the NHANES 2009–2010. The Laryngoscope, 131(3), 686-689.
Iber, C. (2007). The AASM manual for the scoring of sleep and associated events: rules, terminology, and technical specification. (No Title).
Jurysta, F., Van De Borne, P., Migeotte, P.-F., Dumont, M., Lanquart, J.-P., Degaute, J.-P., & Linkowski, P. (2003). A study of the dynamic interactions between sleep EEG and heart rate variability in healthy young men. Clinical neurophysiology, 114(11), 2146-2155.
Kuo, C.-F., Tsai, C.-Y., Cheng, W.-H., Hs, W.-H., Majumdar, A., Stettler, M., . . . Tseng, C.-H. (2023). Machine learning approaches for predicting sleep arousal response based on heart rate variability, oxygen saturation, and body profiles. Digital Health, 9, 20552076231205744.
Léger, D., Debellemaniere, E., Rabat, A., Bayon, V., Benchenane, K., & Chennaoui, M. (2018). Slow-wave sleep: from the cell to the clinic. Sleep medicine reviews, 41, 113-132.
Long, X., Fonseca, P., Aarts, R. M., Haakma, R., Rolink, J., & Leonhardt, S. (2015). Detection of nocturnal slow wave sleep based on cardiorespiratory activity in healthy adults. IEEE Journal of Biomedical and Health Informatics, 21(1), 123-133.
Mensen, A., Zhang, Z., Qi, M., & Khatami, R. (2016). The occurrence of individual slow waves in sleep is predicted by heart rate. Scientific Reports, 6(1), 29671.
Merrigan, J. M., Buysse, D. J., Bird, J. C., & Livingston, E. H. (2013). JAMA patient page. Insomnia. Jama, 309(7), 733.
Nagarkoti, S. K., Singh, B., & Kumar, M. (2012). An algorithm for fetal heart rate detection using wavelet transform. Paper presented at the 2012 1st International Conference on Recent Advances in Information Technology (RAIT).
Nam, B., Bark, B., Lee, J., & Kim, I. Y. (2024). InsightSleepNet: the interpretable and uncertainty-aware deep learning network for sleep staging using continuous Photoplethysmography. BMC Medical Informatics and Decision Making, 24(1), 50.
Samann, F., & Schanze, T. (2023). Multiple ECG segments denoising autoencoder model. Biomedical Engineering/Biomedizinische Technik, 68(3), 275-284.
Schuurmans, A. A., de Looff, P., Nijhof, K. S., Rosada, C., Scholte, R. H., Popma, A., & Otten, R. (2020). Validity of the Empatica E4 wristband to measure heart rate variability (HRV) parameters: A comparison to electrocardiography (ECG). Journal of medical systems, 44, 1-11.
Shaffer, F., & Ginsberg, J. P. (2017). An overview of heart rate variability metrics and norms. Frontiers in public health, 5, 290215.
Shahrbabaki, S. S., Ahmed, B., Penzel, T., & Cvetkovic, D. (2016). Pulse transit time and heart rate variability in sleep staging. Paper presented at the 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC).
Shinar, Z., Baharav, A., Dagan, Y., & Akselrod, S. (2001). Automatic detection of slow-wave-sleep using heart rate variability. Paper presented at the Computers in Cardiology 2001. Vol. 28 (Cat. No. 01CH37287).
Smagulova, K., & James, A. P. (2019). A survey on LSTM memristive neural network architectures and applications. The European Physical Journal Special Topics, 228(10), 2313-2324.
Stein, P. K., & Pu, Y. (2012). Heart rate variability, sleep and sleep disorders. Sleep medicine reviews, 16(1), 47-66.
Su, C.-T. (2013). Quality engineering: off-line methods and applications: CRC press.
Trinder, J., Kleiman, J., Carrington, M., Smith, S., Breen, S., Tan, N., & Kim, Y. (2001). Autonomic activity during human sleep as a function of time and sleep stage. Journal of sleep research, 10(4), 253-264.
Usmankhujaev, S., Ibrokhimov, B., Baydadaev, S., & Kwon, J. (2021). Time series classification with inceptionfcn. Sensors, 22(1), 157.
Verma, A. K., Nandakumar, B., Acedillo, K., Yu, Y., Marshall, E., Schneck, D., . . . Howell, M. J. (2024). Slow-wave sleep dysfunction in mild parkinsonism is associated with excessive beta and reduced delta oscillations in motor cortex. Frontiers in Neuroscience, 18, 1338624.
Van Gent, P., Farah, H., Van Nes, N., & Van Arem, B. (2019). HeartPy: A novel heart rate algorithm for the analysis of noisy signals. Transportation research part F: traffic psychology and behaviour, 66, 368-37
Wang, C., & Holtzman, D. M. (2020). Bidirectional relationship between sleep and Alzheimer’s disease: role of amyloid, tau, and other factors. Neuropsychopharmacology, 45(1), 104-120.
Wang, Y., Yi, X., Luo, M., Wang, Z., Qin, L., Hu, X., & Wang, K. (2023). Prediction of outpatients with conjunctivitis in Xinjiang based on LSTM and GRU models. Plos one, 18(9), e0290541.