跳到主要內容

臺灣博碩士論文加值系統

(44.223.39.67) 您好!臺灣時間:2024/05/26 13:45
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

: 
twitterline
研究生:林嘉葦
研究生(外文):LIN,CHIA-WEI
論文名稱:應用Wasserstein GAN於深度學習之少量資料擴增 – 以多通道酸鹼阻抗訊號診斷咽喉胃酸逆流為例
論文名稱(外文):Augmenting Small Data in Deep Learning with Wasserstein GAN - A Case Study on Diagnosing Pharyngeal Acid Reflux Episodes using Multi-Channel pH Impedance Signals
指導教授:傅家啟傅家啟引用關係
指導教授(外文):FU,JA-CHIH
口試委員:呂明山白炳豐
口試委員(外文):LU,MING-SHANPAI,PING-FENG
口試日期:2023-06-27
學位類別:碩士
校院名稱:國立雲林科技大學
系所名稱:工業工程與管理系
學門:工程學門
學類:工業工程學類
論文種類:學術論文
論文出版年:2023
畢業學年度:111
語文別:中文
論文頁數:101
中文關鍵詞:咽喉胃酸逆流深度學習Wasserstein GANDropout
外文關鍵詞:Laryngopharyngeal refluxDeep learningWasserstein GANDropout
相關次數:
  • 被引用被引用:0
  • 點閱點閱:72
  • 評分評分:
  • 下載下載:17
  • 收藏至我的研究室書目清單書目收藏:0
咽喉胃酸逆流是胃內容物倒流所引起的咽喉或黏膜的發炎反應,症狀主要為聲音嘶啞、慢性咳嗽和呼吸困難等,主要以臨床診斷為主,搭配24小時咽喉酸鹼阻抗監測儀進行測量,而訊號回傳需經由2位以上醫師進行診斷是否罹患咽喉胃酸逆流,但由於監測儀回傳之訊號龐大,人工判別耗時,因此使用卷積神經網路與長短期記憶神經網路(CNN-LSTM)以及多變量長短期記憶神經網路與全卷積神經網路(MLSTM-FCN)來對於咽喉胃酸逆流訊號進行偵測,共截出84筆咽喉胃酸逆流訊號以及53筆非逆流訊號,將訊號分為6秒、11秒及21秒長度資料,並建立個別模型以及級聯級成模型,本研究將結合Dropout以及WGAN以提升咽喉胃酸逆流準確率,利用不同Dropout機率值加入個別模型以及級連集成模型內進行比較,並選出最佳值,再比較WGAN擴增資料數量,將擴增資料以及原始測試資料放入含有最佳Dropout機率值之模型,進行咽喉胃酸逆流的診斷,以解決取樣受測者數目限制所導致的小數據以及資料不平衡。透過使用WGAN資料擴增以及Dropout後提升模型之準確率,於資料擴增後MLSTM-FCN優於CNN-LSTM,且MLSTM-FCN在測試資料準確率達96.55%,於Cohort study準確率達86.54%,結果可看出在MLSTM-FCN在擴增至500筆以及不設置Dropout值能夠得到較佳的準確率,透過資料擴增能夠使模型準確率提高,以輔助醫師於咽喉胃酸逆流診斷,亦節省診斷時間。
Laryngopharyngeal reflux disease (LPRD) is an inflammatory reaction of the throat or mucous membrane caused by the backflow of stomach contents. The symptoms are mainly hoarseness, chronic cough, and breathing difficulties. The diagnosis of pharyngeal acid reflux is mainly based on clinical symptoms with a 24-hour pharyngeal pH-impedance monitor. The signals are judged by more than two physicians for diagnosis if there are symptoms of LPRD. However, since the amount of signals information returned by the monitor is too huge and time-consuming for human interpretation, and to identify them using Convolutional Neural Network and Long Short-Term Memory Network and Multivariate Long Short-Term Memory Network for detection of regurgitation signals the results showed that a total of 84 signals of LPRD and 53 signals without LPRD. The signal has divided into 6 seconds, 11 seconds and 21 seconds. And signal models and cascaded ensemble models are constructed. In this study, Dropout and WGAN will be combined to improve the accuracy of throat acid reflux, and different Dropout probability values will be added to individual models and cascading integration models for comparison. The best value was selected, the number of WGAN data augmentation was compared, and the data augmentation and the original data were into a model containing the best Dropout probability value to the diagnosis of Laryngopharyngeal reflux disease to solve the small data and data unbalance caused by the limitation of the number of sample subjects. By using WGAN data augmentation and Dropout, the model's accuracy has been improved. After data augmentation, MLSTM-FCN outperforms CNN-LSTM, achieving an accuracy of 96.55% on the test data set and 86.54% in the cohort study. The results indicate that MLSTM-FCN achieves better accuracy when augmented to 500 samples and without setting a Dropout value. Through data augmentation, the model's accuracy improves, assisting physicians in diagnosing gastroesophageal reflux and saving diagnosis time.
摘要 i
ABSTRACT ii
目錄 iii
表目錄 vi
圖目錄 ix
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 3
1.3 研究流程 4
第二章 文獻探討 5
2.1 咽喉胃食道逆流緒論及深度學習用於多通道訊號之判讀 5
2.1.1 咽喉胃酸逆流 5
2.1.2 咽喉胃酸逆流檢測方式 5
2.1.3 CNN-LSTM, MLSTM-FCN應用於時間序列分類 8
2.1.4 級聯集成模型(Cascade ensemble) 11
2.2 深度學習應用於小數據集之議題 12
2.2.1 拋棄層(Dropout) 13
2.2.2 Wasserstein GAN (Wasserstein Generative Adversarial Network, WGAN) 15
第三章 研究方法 23
3.1 研究資料與架構 23
3.2 模型比較組合 26
3.3 資料截取判定 28
3.3.1 咽喉胃食道逆流診斷之黃金標準 29
3.3.2 咽喉胃酸逆流訊號偵測與資料截取區分 31
3.4模型訓練 38
3.4.1 Dropout(拋棄層)與個別模型(CNN-LSTM , MLSTM-FCN) 38
3.4.2 Wasserstein GAN資料擴增 42
3.4.3 個別模型訓練 44
3.4.4 級聯集成模型訓練(Cascade ensemble) 45
3.5 績效衡量 47
第四章 實驗結果與分析 49
4.1 實驗樣本 49
4.2實驗設備 50
4.3實驗績效 50
4.3.1 訊號偵測結果 50
4.3.2 各模型參數比較 51
4.3.3 原數據之績效 53
4.3.4 Dropout值之績效比較 55
4.3.5 WGAN資料擴增之績效比較 58
4.4小結 72
第五章 結論與後續研究 73
5.1 結論 73
5.2後續研究 74
參考文獻 75
附錄 79
附錄一、卷積神經網路與長短期記憶神經網路(CNN-LSTM)公式 79
附錄二、多變量長短期記憶神經網路與全卷積神經網路(MLSTM-FCN)公式 81
附錄三、原數據個別模型績效 82
附錄四、個別模型-Dropout值績效比較 83
附錄五、個別模型-擴增值比較 88


1.台北榮民總醫院. 無線膠囊測食道酸鹼 胃食道逆流治療更精準. 2019; Available from: https://www.vghtpe.gov.tw/News!one.action?nid=5774&gcode=A05.
2.Wang, L., et al., Relationship between laryngopharyngeal reflux disease and gastroesophageal reflux disease based on synchronous esophageal and oropharyngeal Dx-pH monitoring. American Journal of Otolaryngology, 2020. 41(3): p. 102441.
3.Koufman, J.A., Laryngopharyngeal reflux is different from classic gastroesophageal reflux disease. Ear, Nose & Throat Journal, 2002. 81(9): p. 7-9.
4.gastroenterology, A.c.o. Acid Reflux. 2022; Available from: https://gi.org/topics/acid-reflux/#tabs2.
5.曾美晴, 火燒心到喉嚨 竟是咽喉逆流. 2017.
6.Lechien, J.-R., et al., Review of management of laryngopharyngeal reflux disease. European Annals of Otorhinolaryngology, Head and Neck Diseases, 2021. 138(4): p. 257-267.
7.Wu, D., et al., Narrow band imaging might contribute to the diagnosis of laryngopharyngeal reflux. American Journal of Otolaryngology, 2020. 41(3): p. 102403.
8.Williams, R.B., et al., Esophagopharyngeal acid regurgitation: dual pH monitoring criteria for its detection and insights into mechanisms. Gastroenterology, 1999. 117(5): p. 1051-1061.
9.Domingos, P., A few useful things to know about machine learning. Communications of the ACM, 2012. 55(10): p. 78-87.
10.Shin, H.-C., et al. Medical image synthesis for data augmentation and anonymization using generative adversarial networks. in International workshop on simulation and synthesis in medical imaging. 2018. Springer.
11.Srivastava, N., et al., Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research, 2014. 15(1): p. 1929-1958.
12.湯書亞 and 台灣家庭醫學醫學會, 喉咽反流與胃食道逆流. 2016.
13.莊宗憲 and 台灣家庭醫學醫學會, 氫離子幫浦阻斷劑(PPI)藥物之比較. 2012.
14.李宜展. 喉嚨卡卡真不舒服 -- 咽喉逆流簡介. 2016; Available from: http://albertleemd.blogspot.com/2016/12/blog-post_28.html.
15.曾屏輝 and 台灣胃腸神經與蠕動學會. 食道酸鹼值檢查. 2021; Available from: http://tnms.org.tw/?action=health_one&news=756231094.
16.吳文傑, 胃食道逆流新檢測儀器 96小時無線膠囊 (Bravo capsule) 酸鹼度監測. 2021.
17.Lai, C.-J., et al., Detecting intraoperative gastric regurgitation by using preattached esophageal multichannel intraluminal impedance and pH monitoring on a solid-state manometry: a case series study. Journal of clinical monitoring and computing, 2020. 34(4): p. 853-859.
18.連敏秀, 應用可解釋人工智慧(XAI)深度學習模型於多通道酸鹼阻抗訊號之咽喉胃酸逆流偵測及分類. 2022.
19.Rai, H.M. and K. Chatterjee, Hybrid CNN-LSTM deep learning model and ensemble technique for automatic detection of myocardial infarction using big ECG data. Applied Intelligence, 2022. 52(5): p. 5366-5384.
20.Shoeibi, A., et al., Automatic diagnosis of schizophrenia in EEG signals using CNN-LSTM models. Frontiers in Neuroinformatics, 2021. 15.
21.Nishikawa, K., et al. Detecting system Alzheimer’s dementia by 1d CNN-LSTM in Japanese speech. in 2021 IEEE International Conference on Consumer Electronics (ICCE). 2021. IEEE.
22.Karim, F., et al., LSTM fully convolutional networks for time series classification. IEEE access, 2017. 6: p. 1662-1669.
23.Karim, F., et al., Multivariate LSTM-FCNs for time series classification. Neural Networks, 2019. 116: p. 237-245.
24.Abu-gellban, H., L. Nguyen, and F. Jin, Gfdlecg: Pac classification for ecg signals using gradient features and deep learning, in Advances in Data Science and Information Engineering. 2021, Springer. p. 369-382.
25.Kladis, E., et al. An Empirical Evaluation of Early Time-Series Classification Algorithms. in EDBT/ICDT Workshops. 2021.
26.Sharma, A.K., et al., Dermatologist-Level Classification of Skin Cancer Using Cascaded Ensembling of Convolutional Neural Network and Handcrafted Features Based Deep Neural Network. IEEE Access, 2022. 10: p. 17920-17932.
27.Li, X., et al., Classification of COVID-19 chest CT images based on ensemble deep learning. Journal of Healthcare Engineering, 2021. 2021.
28.Singh, I., et al. Dropout-VGG based convolutional neural network for traffic sign categorization. in Congress on Intelligent Systems. 2022. Springer.
29.Brigato, L. and L. Iocchi. A close look at deep learning with small data. in 2020 25th International Conference on Pattern Recognition (ICPR). 2021. IEEE.
30.Ha, C., et al., Eliminating overfitting of probabilistic topic models on short and noisy text: The role of dropout. International Journal of Approximate Reasoning, 2019. 112: p. 85-104.
31.Yadav, M., et al. Estimating public speaking anxiety from speechsignals using unsupervised transfer learning. in 2019 Affective Computing and Intelligent Interaction Conference (ACII 2019). 2019.
32.Zhang, Y.-D., et al., Multiple sclerosis identification by convolutional neural network with dropout and parametric ReLU. Journal of computational science, 2018. 28: p. 1-10.
33.Shi, H., et al. Data augmentation with improved generative adversarial networks. in 2018 24th International Conference on Pattern Recognition (ICPR). 2018. IEEE.
34.Wang, L. and Z. Huang. Research on Stock Price Volatility Prediction Based on Generative Adversarial Network. in 2021 International Conference on Networking, Communications and Information Technology (NetCIT). 2021. IEEE.
35.Nielsen, F., On a generalization of the Jensen–Shannon divergence and the Jensen–Shannon centroid. Entropy, 2020. 22(2): p. 221.
36.Li, L.-L., et al. EEG Generation of Virtual Channels Using an Improved Wasserstein Generative Adversarial Networks. in International Conference on Intelligent Robotics and Applications. 2022. Springer.
37.Song, M., Q. Su, and M. Zhang. Handwriting Digit Generation Based on GAN Model. in 2021 3rd International Conference on Machine Learning, Big Data and Business Intelligence (MLBDBI). 2021. IEEE.
38.Kong, J. and F. Zhang. SAR Target Recognition with Generative Adversarial Network (GAN)-based Data Augmentation. in 2021 13th International Conference on Advanced Infocomm Technology (ICAIT). 2021. IEEE.
39.Lan, T., et al. Arrhythmias Classification Using Short-Time Fourier Transform and GAN Based Data Augmentation. in 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). 2020. IEEE.
40.Frid-Adar, M., et al. Synthetic data augmentation using GAN for improved liver lesion classification. in 2018 IEEE 15th international symposium on biomedical imaging (ISBI 2018). 2018. IEEE.
41.dhyaaalayed, wgan-gaussian. 2019.
42.Han, L., et al. An intrusion detection method based on WGAN and deep learning. in 2021 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC). 2021. IEEE.
43.Bouallegue, G. and R. Djemal. EEG data augmentation using Wasserstein GAN. in 2020 20th International Conference on Sciences and Techniques of Automatic Control and Computer Engineering (STA). 2020. IEEE.
44.Petruţiu, V.M., et al. Enhancing the classification of EEG signals using Wasserstein generative adversarial networks. in 2020 IEEE 16th International Conference on Intelligent Computer Communication and Processing (ICCP). 2020. IEEE.
45.Munia, M.S., M. Nourani, and S. Houari. Biosignal oversampling using wasserstein generative adversarial network. in 2020 IEEE International Conference on Healthcare Informatics (ICHI). 2020. IEEE.
46.Wang, Q., et al., WGAN-based synthetic minority over-sampling technique: Improving semantic fine-grained classification for lung nodules in CT images. IEEE Access, 2019. 7: p. 18450-18463.


QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊