臺灣博碩士論文加值系統

目的：為了比較相同併發症下正常組織併發症機率模型的預測效力。除了常規的因子挑選，本研究想以實證醫學的方式透過統合分析(Meta analysis)與自然語言處理的結合來進行頭頸癌預後因子間的探討，並對統合分析流程中的檢索文獻流程進行自然語言處理以達自動化。
材料與方法： 本研究透過PICOS原則進行統合分析問題的研擬，隨後在WOS及Pubmed這些文庫進行黏膜炎 ( Mucositis ) 、口乾 ( Xerostomia ) 、吞嚥困難 ( Dysphagia ) 三種併發症的放射治療後正常組織併發症概率 ( Normal tissue complication probability, NTCP ) 相關文獻檢索，接著對檢索到的文獻做統合分析之自然語言處理CNN模型 ( Convolutional Neural Networks for Natural Language Processing ) 流程結合對統合分析收錄的摘要做初步審查。經過全文條件的排除與PROBAST ( Prediction model Risk Of Bias ASsessment Tool ) 評分工具的評分，再將其提取數據進行效應量合併與偏倚、異質性分析用以輸出其森林圖與漏斗圖。完成上述流程後，進行綜合成效評估，分為兩方面，其一是文獻品質評估方面為PROBAST評估結果與偏倚發表，其二是模型評估方面內容則包含使用WPM ( Words per minute ) 比較人工與自然語言處理CNN模型兩方流程時間效率、分析文獻覆蓋率驗證其檢索的取代工作量的程度，以及透過混淆矩陣、準確性 (Accuracy, ACC) 、召回率 (Recall ) 、F1 score、與之分類模型的Loss值做為評估指標對模型訓練集、測試集與模型內部成效進行評估。
結果 :本研究除了呈現統合分析之研究選擇與特徵表、文獻PROBAST評分表、偏倚分析漏斗圖、森林圖結果圖確立較佳模型與特徵因子組合外。在自然語言處理成效則展示CNN進行摘要與題目檢索的文獻覆蓋率和時間效率比較，以及CNN對文本分類辨識的最直觀指標初步的準確率 ( Accuracy, ACC ) 66.7%，再經由優化器與超參數設定後提升至94.87%。
結論： 本研究使用統合分析與自然語言處理的結合進行頭頸癌特徵因子間的探討，並在研究結果部分表明至少具有一組以上的模型與特徵組合在預測同一種併發症的情況下，除了具有預測能力外也具有較高成效的預測能力並與虛無假設H_1相符。此外再結果的部分展示使用捲機神經網路的自然語言處理模型能一定程度地簡化統合分析流程，並將此工具納入總流程能減少多變量正常組織併發症預測與頭頸癌特徵因子間的探討文獻之識別與審查的工作量與時間。

Purpose : In order to compare the predictive performance of normal tissue complication probability ( NTCP ) models for the same complications. Besides conventional factor selection, this study aims to empirically explore the prognostic factors of head and neck cancer through the integration of meta-analysis and natural language processing ( NLP ), and to automate the literature retrieval process within the meta-analysis procedure.
Materials and methods : This study formulated research questions based on the PICOS principle for meta-analysis. Subsequently, relevant literature on three complications ( mucositis, xerostomia, and dysphagia ) of normal tissue after head and neck cancer radiotherapy was retrieved from databases such as WOS and PubMed. The retrieved literature was then subjected to the NLP CNN ( Convolutional Neural Networks for Natural Language Processing ) model for initial review within the meta-analysis process. After exclusion based on full-text conditions and assessment using the PROBAST ( Prediction model Risk Of Bias ASsessment Tool ) scoring tool, data extraction was performed for effect size combination and bias and heterogeneity analysis, which was used to generate forest plots and funnel plots. Following these steps, a comprehensive assessment was conducted, including PROBAST assessment results and bias publication for literature quality evaluation, as well as a model evaluation that compared the time efficiency between manual and NLP CNN processes using Words per Minute ( WPM ), verified the extent of literature coverage replacing workload, and evaluated the model's performance using confusion matrix, accuracy ( ACC ), recall, F1 score, and Loss value for training set, test set, and internal model effectiveness.
Results : In addition to presenting the selection and feature table of meta-analysis studies, PROBAST scoring table, bias analysis funnel plot, and forest plot results to establish optimal model and feature factor combinations, the effectiveness of natural language processing is demonstrated. The coverage and time efficiency comparison of literature retrieval for abstracts and titles using CNN, as well as the preliminary accuracy (ACC) of CNN for text classification recognition, were 66.7%. After optimization with optimizers and hyperparameter settings, the accuracy was improved to 94.87%.
Conclusions : This study integrates meta-analysis and natural language processing to explore the prognostic factors of head and neck cancer. The results suggest that at least one set of models and feature combinations have predictive ability and high predictive performance for the same complication, and are consistent with the null hypothesis H_1. Furthermore, the results demonstrate that the natural language processing model using convolutional neural networks can simplify the meta-analysis process to some extent. Incorporating this tool into the overall process can reduce the workload and time required for the identification and review of literature discussing the prediction of normal tissue complications and prognostic factors of head and neck cancer.

摘要 i
致謝 vi
目錄 vii
表目錄 ix
圖目錄 ix
符號縮寫 xi
符號說明 xii
第一章 緒論 1
1.1 動機 1
1.3相關文獻探討 6
第二章 材料與方法 8
2.1前言 8
2.2主題及檢索資料 11
2.3系統性文獻回顧與統合分析 15
2.3.1隨機效應模型 ( RANDOM-EFFECTS MODEL ) 15
2.4文獻管理及評估 16
2.4.1收錄文獻納入與排除條件 17
2.5 NLP程序設計 17
2.5.1 CNNNLP ( CONVOLUTIONAL NEURAL NETWORKS FOR NATURAL LANGUAGE PROCESSING ) 18
2.5.2 資料預處理 19
2.5.3詞向量嵌入 ( WORD EMBEDDING ) 20
2.5.4卷積層 ( CONVOLUTIONAL LAYER ) 22
2.5.5池化層 ( POOLING LAYER ) : 23
2.5.6啟動函數 ( ACTIVATION FUNCTION ) 24
2.5.7全連接層 ( FULLY CONNECTED LAYER ) : 26
2.5.8多層感知器 ( MULTILAYER PERCEPTRON，MLP ) : 26
2.5.9優化器優化 28
2.5.10超參數設定 28
2.6文獻品質分析 29
2.7 統合分析效應量合併 32
2.8綜合評估指標 34
2.8.1時間效率 34
2.8.2單詞每分鐘 35
2.8.3文獻覆蓋率 36
2.8.4模型評估指標 37
第三章 結果 40
3.1前言 40
3.2納入統合分析文獻流程圖 41
3.2.1 統合分析檢索收錄文獻 42
3.2.2收錄研究特徵表 47
3.3森林圖 ( FOREST PLOT ) 52
3.4綜合分析及效能評估 54
3.4.1文獻PROBAST評估表結果 54
3.4.2研究出版偏倚 55
3.4.3 CNNNLP模型成效 56
3.4.4文獻覆蓋率時間效率 66
第四章 討論 70
4.1統合分析研究結果探討 70
4.2 CNNNLP模型成效、優化器優化與覆蓋率探討 72
4.3統合分析研究改善與侷限性 76
4.4NLP研究改善與侷限性 77
第五章 結論 80
5.1結論 80
第六章 未來研究 81
6.1未來研究 81
參考文獻: 83
附錄 88
附錄1.資料庫明細檢索明細表 88
附錄2. PROBAST評分工具問題評估內容 90
附錄3. WPM計算程式 92
作者資訊: 93

1.Chen, A.M., et al., Quality of life among long-term survivors of head and neck cancer treated by intensity-modulated radiotherapy. JAMA Otolaryngology–Head & Neck Surgery, 2014. 140(2): p. 129-133.
2.Nguyen, N., et al., Dysphagia following chemoradiation for locally advanced head and neck cancer. Annals of Oncology, 2004. 15(3): p. 383-388.
3.Jensen, S.B., et al., Salivary gland hypofunction and xerostomia in head and neck radiation patients. JNCI Monographs, 2019. 2019(53): p. lgz016.
4.Gueiros, L.A., M.S.M. Soares, and J.C. Leao, Impact of ageing and drug consumption on oral health. Gerodontology, 2009. 26(4): p. 297-301.
5.Sackett, D.L., et al., Evidence based medicine: What it is and what it isn't - It's about integrating individual clinical expertise and the best external evidence. British Medical Journal, 1996. 312(7023): p. 71-72.
6.Borenstein, M., et al., Introduction to meta-analysis. 2021: John Wiley & Sons.
7.Straus, S.E., et al., Evidence-Based Medicine E-Book: Evidence-Based Medicine E-Book. 2018: Elsevier Health Sciences.
8.Murad, M.H., et al., New evidence pyramid. BMJ Evidence-Based Medicine, 2016. 21(4): p. 125-127.
9.Deng, Z., et al., Validation of a semiautomated natural language processing–based procedure for meta-analysis of cancer susceptibility gene penetrance. JCO clinical cancer informatics, 2019. 3: p. 1-9.
10.Takeshita, M., R. Rzepka, and K. Araki, Speciesist language and nonhuman animal bias in English Masked Language Models. Information Processing & Management, 2022. 59(5): p. 103050.
11.Walls, G., et al., Radiomics for predicting lung cancer outcomes following radiotherapy: a systematic review. Clinical oncology, 2022. 34(3): p. e107-e122.
12.El Naqa, I., et al., Multivariable modeling of radiotherapy outcomes, including dose–volume and clinical factors. International Journal of Radiation Oncology* Biology* Physics, 2006. 64(4): p. 1275-1286.
13.Buckland, S.T., K.P. Burnham, and N.H. Augustin, Model selection: an integral part of inference. Biometrics, 1997: p. 603-618.
14.Bentzen, S.M., et al., Quantitative Analyses of Normal Tissue Effects in the Clinic (QUANTEC): an introduction to the scientific issues. International Journal of Radiation Oncology* Biology* Physics, 2010. 76(3): p. S3-S9.
15.Marks, L.B., et al., Use of normal tissue complication probability models in the clinic. International Journal of Radiation Oncology* Biology* Physics, 2010. 76(3): p. S10-S19.
16.Goguen, L.A., et al., Examining the need for neck dissection in the era of chemoradiation therapy for advanced head and neck cancer. Archives of Otolaryngology–Head & Neck Surgery, 2006. 132(5): p. 526-531.
17.Wang, R., et al., Statistics in medicine—reporting of subgroup analyses in clinical trials. New England Journal of Medicine, 2007. 357(21): p. 2189-2194.
18.Schochet, P.Z. and H.S. Chiang, Estimation and identification of the complier average causal effect parameter in education RCTs. Journal of Educational and Behavioral Statistics, 2011. 36(3): p. 307-345.
19.Zhao, Q.Y., D.S. Small, and W.J. Su, Multiple Testing When Many p-Values are Uniformly Conservative, with Application to Testing Qualitative Interaction in Educational Interventions. Journal of the American Statistical Association, 2019. 114(527): p. 1291-1304.
20.Bang, C., et al., Artificial intelligence to predict outcomes of head and neck radiotherapy. Clinical and Translational Radiation Oncology, 2023. 39: p. 100590.
21.Moher, D., et al., Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Annals of internal medicine, 2009. 151(4): p. 264-269.
22.Concato, J., N. Shah, and R.I. Horwitz, Randomized, controlled trials, observational studies, and the hierarchy of research designs. New England journal of medicine, 2000. 342(25): p. 1887-1892.
23.Friedman, L.M., et al., Fundamentals of clinical trials. 2015: Springer.
24.Booth, A., “Brimful of STARLITE”: toward standards for reporting literature searches. Journal of the Medical Library Association, 2006. 94(4): p. 421.
25.Hoffmann, T.C., et al., Better reporting of interventions: template for intervention description and replication (TIDieR) checklist and guide. Bmj, 2014. 348.
26.Spiteri, L., A simplified model for facet analysis: Ranganathan 101. Canadian journal of information and library science, 1998. 23(1-2): p. 1-30.
27.Sandbank, M., et al., Project AIM: Autism intervention meta-analysis for studies of young children. Psychological bulletin, 2020. 146(1): p. 1.
28.Mellenbergh, G.J., A conceptual introduction to psychometrics: Development, analysis and application of psychological and educational tests. 2011: The HagueEleven international publishing.
29.Leandro, G., Meta-analysis in medical research: The handbook for the understanding and practice of meta-analysis. 2005: John Wiley & Sons.
30.Guolo, A. and C. Varin, Random-effects meta-analysis: the number of studies matters. Statistical methods in medical research, 2017. 26(3): p. 1500-1518.
31.Huedo-Medina, T.B., et al., Assessing heterogeneity in meta-analysis: Q statistic or I² index? Psychological methods, 2006. 11(2): p. 193.
32.Schmidt, F.L., I.S. Oh, and T.L. Hayes, Fixed‐versus random‐effects models in meta‐analysis: Model properties and an empirical comparison of differences in results. British Journal of Mathematical and Statistical Psychology, 2009. 62(1): p. 97-128.
33.Johnson, R. and T. Zhang, Convolutional neural networks for text categorization: Shallow word-level vs. deep character-level. arXiv preprint arXiv:1609.00718, 2016.
34.He, X., J. Gao, and L. Deng. Deep learning for natural language processing: Theory and practice (tutorial). in Cikm. 2014.
35.Mikolov, T., et al., Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013.
36.Pennington, J., R. Socher, and C.D. Manning. Glove: Global vectors for word representation. in Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP). 2014.
37.Jo, H. and S.J. Choi, Extrofitting: Enriching word representation and its vector space with semantic lexicons. arXiv preprint arXiv:1804.07946, 2018.
38.Devlin, J., et al., Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018.
39.Kim, Y., Convolutional neural networks for sentence classification. arXiv preprint arXiv:1408.5882, 2014.
40.Glorot, X., A. Bordes, and Y. Bengio. Deep sparse rectifier neural networks. in Proceedings of the fourteenth international conference on artificial intelligence and statistics. 2011. JMLR Workshop and Conference Proceedings.
41.Goodfellow, I., Y. Bengio, and A. Courville, Deep learning. 2016: MIT press.
42.He, K., et al. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. in Proceedings of the IEEE international conference on computer vision. 2015.
43.Conneau, A., et al., Very deep convolutional networks for text classification. arXiv preprint arXiv:1606.01781, 2016.
44.Scale, N.-O., Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. 2014.
45.Higgins, J.P., et al., A tool to assess the quality of a meta‐analysis. Research Synthesis Methods, 2013. 4(4): p. 351-366.
46.Debray, T.P., et al., A guide to systematic review and meta-analysis of prediction model performance. bmj, 2017. 356.
47.Moons, K.G., et al., PROBAST: a tool to assess risk of bias and applicability of prediction model studies: explanation and elaboration. Annals of internal medicine, 2019. 170(1): p. W1-W33.
48.Fernandez-Felix, B.M., et al., CHARMS and PROBAST at your fingertips: a template for data extraction and risk of bias assessment in systematic reviews of predictive models. BMC Medical Research Methodology, 2023. 23(1): p. 1-8.
49.Cochran, W.G., The comparison of percentages in matched samples. Biometrika, 1950. 37(3/4): p. 256-266.
50.Higgins, J.P., et al., Measuring inconsistency in meta-analyses. Bmj, 2003. 327(7414): p. 557-560.
51.Ntonti, P., et al., A systematic review of reading tests. International Journal of Ophthalmology, 2023. 16(1): p. 121.
52.Lee, T.F., et al., Using Multivariate Regression Model with Least Absolute Shrinkage and Selection Operator (LASSO) to Predict the Incidence of Xerostomia after Intensity-Modulated Radiotherapy for Head and Neck Cancer. Plos One, 2014. 9(2).
53.Lee, T.F., et al., LASSO NTCP predictors for the incidence of xerostomia in patients with head and neck squamous cell carcinoma and nasopharyngeal carcinoma. Scientific Reports, 2014. 4.
54.Gabrys, H.S., et al., Design and Selection of Machine Learning Methods Using Radiomics and Dosiomics for Normal Tissue Complication Probability Modeling of Xerostomia. Frontiers in Oncology, 2018. 8.
55.van Dijk, L.V., et al., CT image biomarkers to improve patient-specific prediction of radiation-induced xerostomia and sticky saliva. Radiotherapy and Oncology, 2017. 122(2): p. 185-191.
56.Ursino, S., et al., Incorporating dose–volume histogram parameters of swallowing organs at risk in a videofluoroscopy-based predictive model of radiation-induced dysphagia after head and neck cancer intensity-modulated radiation therapy. Strahlentherapie und Onkologie, 2021. 197: p. 209-218.
57.Dean, J., et al., Incorporating spatial dose metrics in machine learning-based normal tissue complication probability (NTCP) models of severe acute dysphagia resulting from head and neck radiotherapy. Clinical and translational radiation oncology, 2018. 8: p. 27-39.
58.Dean, J.A., et al., Normal tissue complication probability (NTCP) modelling using spatial dose metrics and machine learning methods for severe acute oral mucositis resulting from head and neck radiotherapy. Radiotherapy and Oncology, 2016. 120(1): p. 21-27.
59.Cumpston, M., et al., Updated guidance for trusted systematic reviews: a new edition of the Cochrane Handbook for Systematic Reviews of Interventions. The Cochrane database of systematic reviews, 2019. 2019(10).
60.Montavon, G., G. Orr, and K.-R. Müller, Neural networks: tricks of the trade. Vol. 7700. 2012: springer.
61.Downs, S.H. and N. Black, The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and non-randomised studies of health care interventions. Journal of epidemiology & community health, 1998. 52(6): p. 377-384.
62.Brown, T., et al., Language models are few-shot learners. Advances in neural information processing systems, 2020. 33: p. 1877-1901.
63.Esteva, A., et al., Dermatologist-level classification of skin cancer with deep neural networks. nature, 2017. 542(7639): p. 115-118.
64.Zheng, T., et al., Detection of medical text semantic similarity based on convolutional neural network. BMC medical informatics and decision making, 2019. 19: p. 1-11.