臺灣博碩士論文加值系統

根據統計資料，我國近幾年早產兒佔所有新生兒比例約10%，其中懐孕週數小於29 週的極端早產出生的新生兒大約佔1%，是一群需要特別照護的群體。這些提早3-4個多月出生的早產兒，由於極度早產導致各種器官不成熟，需在新生兒加護病房的保溫箱住上數週至數個月之久。住院期間須接受各器官系統的重度急救及加護醫療，並暴露在許多的危險因子之中，影響其出院之預後。其中長時間的呼吸器使用及氧氣治療，也是其中一個影響預後的重要因子。
本研究主要在探討氧氣暴露軌跡的集群問題，可視為一個函數型資料之集群問題。根據Abraham (2003)提出的無監督曲線集群(unsupervised curve clustering)方法，我們採用具有正交性質的多解析薄板樣條基底函數來進行配適模型，並加入影響曲線變化的重要解釋變數後，進行氧氣暴露軌跡的集群。由初步的結果可看出，提出之方法可適當的分出具有意義的兩個軌跡群，並檢視不同分群與預後之間的關係。

Based on statistical data, preterm births in Taiwan account for approximately 10% of all newborns in recent years, with extremely preterm infants (born before 29 weeks of gestation) constituting about 1%. These infants, born 3 to 4 months early, suffer from severely underdeveloped organs and require several weeks to months of care in the neonatal intensive care unit (NICU) incubators. During their hospital stay, they undergo intensive emergency and critical care for various organ systems and are exposed to numerous risk factors that affect their prognosis upon discharge. Among these, prolonged use of mechanical ventilation and oxygen therapy are significant factors influencing their outcomes.
This study focuses on the clustering problem of oxygen exposure trajectories, which can be considered a functional data clustering problem. Utilizing the unsupervised curve clustering method proposed by Abraham (2003), we employ orthogonal multiresolution thin-plate spline basis function to fit the model. By incorporating important explanatory variables affecting the curve changes, we cluster the oxygen exposure trajectories. Preliminary results indicate that two meaningful trajectory groups can be appropriately identified, and the relationship between different clusters and prognosis is examined.

中文摘要 i
Abstract ii
1 前言 1
2 研究分析資料 3
2.1 資料介紹 3
2.1.1 分析變數說明 3
2.2 資料前處理 6
2.2.1 缺失值插補 6
2.2.2 資料合併 8
2.3 用氧濃度及呼吸器使用模式分布 10
2.3.1 所有早產兒分布情形 10
2.3.2 各週數分布情形 13
2.3.3 呼吸器使用模式與用氧濃度的關係 16
2.4 資料中的離群值 20
3 研究方法介紹 22
3.1 去除離群值 22
3.1.1 用氧濃度的距離 24
3.1.2 離群值的判斷方法 25
3.2 標準化 30
3.3 建構用氧濃度的時間相關模型 30
3.3.2 B樣條基底(B-spline) 32
3.3.3 多解析薄板樣條基底（MRTS） 34
3.4 k-平均集群（k-means clustering） 35
4 研究結果 39
4.1 集群結果 39
5 討論與未來工作 44
5.1 臨床預後的影響因子 44
5.2 變數篩選 46
5.3 具關聯性多軌跡集群 51
參考文獻 52

[1]Abraham, C., Cornillon, P. A., Matzner‐Løber, E. R. I. C., and Molinari, N. (2003). Unsupervised curve clustering using B‐splines. Scandinavian journal of statistics. 30(3): 581-595.
[2]Collaco, J. M., and McGrath-Morrow, S. A. (2018). Respiratory phenotypes for preterm infants, children, and adults: bronchopulmonary dysplasia and more. Annals of the American Thoracic Society. 15(5): 530-538.
[3]Faramarzi, R., Darabi, A., Emadzadeh, M., Maamouri, G., and Rezvani, R. (2023). Predicting neurodevelopmental outcomes in preterm infants: A comprehensive evaluation of neonatal and maternal risk factors. Early Human Development. 184: 105834.
[4]Figueras-Aloy, J., Palet-Trujols, C., Matas-Barceló, I., Botet-Mussons, F., and Carbonell-Estrany, X. (2020). Extrauterine growth restriction in very preterm infant: etiology, diagnosis, and 2-year follow-up. European Journal of Pediatrics. 179: 1469-1479.
[5]Giacofci, M., Lambert-Lacroix, S., Marot, G., and Picard, F. (2013). Wavelet-based clustering for mixed-effects functional models in high dimension. Biometrics. 69(1): 31-40.
[6]Grelli, K. N., Keller, R. L., Rogers, E. E., Partridge, J. C., Xu, D., Barkovich, A. J., and Gano, D. (2021). Bronchopulmonary dysplasia precursors influence risk of white matter injury and adverse neurodevelopmental outcome in preterm infants. Pediatric research. 90(2): 359-365.
[7]Hien D. N., Geoffrey J. M., and Ian A. W. (2016). Mixtures of spatial spline regressions for clustering and classification. Computational Statistics and Data Analysis. 93: 76-85.
[8]Higgins, R. D., Jobe, A. H., Koso-Thomas, M., Bancalari, E., Viscardi, R. M., Hartert, T. V., ... and Raju, T. N. (2018). Bronchopulmonary dysplasia: executive summary of a workshop. The Journal of pediatrics. 197: 300-308.
[9]James, G. M., and Sugar, C. A. (2003). Clustering for Sparsely Sampled Functional Data. Journal of the American Statistical Association. 98(462): 397-408.
[10]Kakatsaki, I., Papanikolaou, S., Roumeliotaki, T., Anagnostatou, N. H., Lygerou, I., and Hatzidaki, E. (2023). The prevalence of small for gestational age and extrauterine growth restriction among extremely and very preterm neonates, using different growth curves, and its association with clinical and nutritional factors. Nutrients. 15(15): 3290.
[11]Kang, I., Choi, H., Yoon, Y. J., Park, J., Kwon, S. S., and Park, C. (2023). Fréchet distance-based cluster analysis for multi-dimensional functional data. Statistics and Computing, 33(4), 75.
[12]Lasky, R. E., and Williams, A. L. (2009). Noise and light exposures for extremely low birth weight newborns during their stay in the neonatal intensive care unit. Pediatrics. 123(2): 540-546.
[13]Ma, P., and Zhong, W. (2008). Penalized clustering of large-scale functional data with multiple covariates. Journal of the American Statistical Association. 103(482): 625-636.
[14]Martínez-Jiménez, M. D., Gómez-García, F. J., Gil-Campos, M., and Pérez-Navero, J. L. (2020). Comorbidities in childhood associated with extrauterine growth restriction in preterm infants: a scoping review. European Journal of Pediatrics. 179: 1255-1265.
[15]Nguyen, H. D., Ullmann, J. F., McLachlan, G. J., Voleti, V., Li, W., Hillman, E. M., ... and Janke, A. L. (2018). Whole‐volume clustering of time series data from zebrafish brain calcium images via mixture modeling. Statistical Analysis and Data Mining: The ASA Data Science Journal. 11(1): 5-16.
[16]Norman, M., Jonsson, B., Söderling, J., Björklund, L. J., and Håkansson, S. (2023). Patterns of respiratory support by gestational age in very preterm infants. Neonatology. 120(1): 142-152.
[17]Northway Jr, W. H., Rosan, R. C., and Porter, D. Y. (1967). Pulmonary disease following respirator therapy of hyaline-membrane disease: bronchopulmonary dysplasia. New England Journal of Medicine. 276(7): 357-368.
[18]Ray, S., and Mallick, B. (2006). Functional clustering by Bayesian wavelet methods. Journal of the Royal Statistical Society Series B: Statistical Methodology. 68(2): 305-332.
[19]Shennan, A. T., Dunn, M. S., Ohlsson, A., Lennox, K., and Hoskins, E. M. (1988). Abnormal pulmonary outcomes in premature infants: prediction from oxygen requirement in the neonatal period. Pediatrics. 82(4): 527-532.
[20]Suarez, A. J., and Ghosal, S. (2016). Bayesian clustering of functional data using local features. Bayesian Analysis. 11(1): 71-98.
[21]Tarpey, T., and Kinateder, K. K. (2003). Clustering functional data. Journal of classification. 20(1): 93-114.
[22]Tarpey, T. (2007). Linear transformations and the k-means clustering algorithm: applications to clustering curves. The American statistician. 61(1): 34-40.
[23]Yu, W. H., Chu, C. H., Lin, Y. C., Chen, R. B., Iwata, O., and Huang, C. C. (2022). Early‐life respiratory trajectories and neurodevelopmental outcomes in infants born very and extremely preterm: A retrospective study. Developmental Medicine & Child Neurology. 64(10): 1246-1253.