簡易檢索 / 詳目顯示

回結果列表
研究生： 王韋凱
Wang, Wei-Kai
論文名稱： 多重相關性資料補值及多重感測器Lasso回歸於躁鬱症疾患評估
Multiple Correlation Data Imputation and MLP-based Lasso Regression for Bipolar Disorder Assessment
指導教授： 吳宗憲
Wu, Chung-Hsien
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 資訊工程學系
Department of Computer Science and Information Engineering
論文出版年： 2021
畢業學年度： 109
語文別： 英文
論文頁數： 59
中文關鍵詞： 躁鬱症評估 、漢氏憂鬱量表因素 、楊氏躁症量表因素 、數位足跡 、智慧型手機 、資料補值
外文關鍵詞： Bipolar Disorder Assessment, The factors of HAMD, The factors of YMRS, Digital Phenotyping, smart phones, data imputation
相關次數： 點閱：85  下載：8 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 躁鬱症為最常見的心理疾病之一，躁鬱症可以透過即時的治療來改善病患病狀，但評估過程需要的醫療人力及時間需求相當龐大。為了減少社會醫療資源的耗損，本論文透過智慧型手機App向23位健康對照組以及142位躁鬱症患者，平均年齡分別為34.37和24.75，收集資料不同型態(位置資訊、自評量表、每日心情、睡眠時間及多媒體紀錄)等數位足跡資料，並且從中提取特徵合併後做為評估病症的依據，訓練評估鬱期症狀的HAMD和躁期YMRS量表分數預測系統，以協助醫生評估病人的資訊。
    在預測量表方面，由於量表因素可以給予醫生更多病人的資訊，因此本論文預測目標除了原本的量表總分外，也額外預測量表因素。而在蒐集資料方面由於缺失值嚴重，導致完整資料只有187筆。本論文透過knn結合多重相關係數的補值方法增加更可靠的資料量，將完整資料增加至641筆。預測模型方面，使用MLP-based Lasso減少模型複雜度的方式，減少過擬合的問題，也可以利用其模型係數看出輸入特徵與預測目標的關係，使預測模型具有可解釋性。
    在模型表現上，在增加資料量後，MLP-based Lasso表現較Lasso Regression更佳。MLP-based Lasso在HAMD的平均因素及總分的平均絕對誤差為0.55和2.00，較Lasso Regression的0.61和2.07更減少了0.06及0.07，而在YMRS的量表方面，MLP-based Lasso在HAMD的平均因素及總分的誤差為0.34和0.95，較Lasso Regression的0.46和1.05更減少了0.12以及0.1。

    Bipolar disorder is one of the most common mental illnesses, but the patient's symptoms can be controlled through immediate treatment. However, the evaluation process requires a huge amount of medical manpower and time. In order to reduce the consumption of social medical resources, this thesis designed a convenient system to monitor patients and provide early warning of illness. We collected data from 23 healthy control groups and 142 bipolar disorder patients with an average age of 34.37 and 24.75 through a smartphone App. The different types of collected data contains location information (GPS), self-rating scale, daily mood, sleep time, and multimedia records (text, voice and video). Then we extracted the features from them and combined the features as the basis for evaluating symptoms. Finally, we trained a predictive system to evaluate HAMD and manic YMRS scale scores for depressive symptoms to assist doctors in assessing patient condition.
    In terms of predictive scales, because scale factors can provide more patients’ information, the prediction target of this thesis can predict the scale in addition to the original total score, and the possible data is due to the serious value. There are only 187 complete data. This thesis uses knn combined with the correlation coefficient to increase the amount of more reliable data, increasing the number of complete data to 641. In terms of predictive models, the use of MLP-based lasso can reduce the complexity of the model and solve the problem of over-fitting, and the model can also be used to generate the relationship between the input features and the prediction target, so that the predictive model is interpretable.
    In terms of model performance, after increasing the amount of data, MLP-based Lasso performs better than Lasso Regression. In prediction of HAMD scale, mean absolute error on the average factor and total score of Lasso based on MLP were 0.55 and 2.00, which was a decrease of 0.06 and 0.07 compared with 0.61 and 2.07 of Lasso Regression. In terms of YMRS scale, mean absolute error on the average factor and total score of Lasso based on MLP were 0.34 and 0.95, which was a decrease of 0.12 and 0.1 compared with 0.46 and 1.05 of Lasso Regression.

    摘要 I Abstract III 誌謝 V Contents VI List of Tables VIII List of Figures XI Chapter 1 Introduction 1 1.1 Background 1 1.2 Motivation 3 1.3 Literature Review 4 1.3.1 The characteristics related to bipolar disorder 4 1.3.2 data imputation for incomplete data 4 1.3.3 Target of Prediction – the factors of scale 5 1.3.4 Prediction Model 5 1.4 Problems 6 1.5 Brief description of the proposed method 6 Chapter 2 Database design and collection 8 2.1 Criteria of Participants 8 2.2 Collection Process 8 Chapter 3 Feature Extraction 11 3.1 Feature Extraction of various data types 11 3.1.1 Feature parameter extraction process 11 3.1.2 Feature extraction-location information (GPS) 12 3.1.3 Feature extraction-self-rating scale 13 3.1.4 Feature extraction-daily valence 15 3.1.5 Feature extraction-sleep time 15 3.1.6 Feature extraction- daily seven emotions 18 3.1.7 Feature extraction-emotion profile of multimedia 18 3.2 The statistics of various feature types parameters 22 Chapter 4 Proposed Method 25 4.1 knn imputation with multiple correlation 25 4.2 Factors of HAMD and YMRS 28 4.3 Prediction Models 31 4.3.1 Lasso Regression 31 4.3.2 Multilayer Perceptron 31 4.3.3 MLP-based Lasso Regression 32 Chapter 5 Experiment and Discussion 34 5.1 Multimedia emotion recognition model 34 5.1.1 Text emotion recognition model -BERT 34 5.1.2 Audio emotion recognition model-Gradient Boosting Classifier 34 5.1.3 Video emotion recognition model-Emotion-FAN 35 5.2 Evaluation of data imputation methods 36 5.3 Evaluation of the different factor prediction models under different training data 40 5.4 The influence of various feature on scale factors and total score 52 Chapter 6 Conclusion and Future Work 56 Reference 58

    1. 精神疾病介紹-2躁鬱症(Bipolar Disorder). 2018.
    2. Hirschfeld, R. and L.A. Vornik, Bipolar disorder—costs and comorbidity. Am J Manag Care, 2005. 11(3 Suppl): p. S85-S90.
    3. 劉嘉逸, 憂鬱好文章 - 漢氏憂鬱量表.
    4. 財務委員會審核二零一七至一八年度開支預算管制人員的答覆. 2017.
    5. Muaremi, A., et al. Assessing bipolar episodes using speech cues derived from phone calls. in International symposium on pervasive computing paradigms for mental health. 2014. Springer.
    6. Valenza, G., et al., Mood states modulate complexity in heartbeat dynamics: A multiscale entropy analysis. EPL (Europhysics Letters), 2014. 107(1): p. 18003.
    7. Bopp, J.M., et al., The longitudinal course of bipolar disorder as revealed through weekly text messaging: a feasibility study. Bipolar disorders, 2010. 12(3): p. 327-334.
    8. Malik, A., G.M. Goodwin, and E.A. Holmes, Contemporary approaches to frequent mood monitoring in bipolar disorder. Journal of experimental psychopathology, 2012. 3(4): p. 572-581.
    9. Gold, A.K. and L.G. Sylvia, The role of sleep in bipolar disorder. Nature and science of sleep, 2016. 8: p. 207.
    10. Harvey, A.G., Sleep and circadian rhythms in bipolar disorder: seeking synchrony, harmony, and regulation. American journal of psychiatry, 2008. 165(7): p. 820-829.
    11. Murray, G. and A. Harvey, Circadian rhythms and sleep in bipolar disorder. Bipolar disorders, 2010. 12(5): p. 459-472.
    12. Plante, D.T. and J.W. Winkelman, Sleep disturbance in bipolar disorder: therapeutic implications. American Journal of Psychiatry, 2008. 165(7): p. 830-843.
    13. Karam, Z.N., et al. Ecologically valid long-term mood monitoring of individuals with bipolar disorder using speech. in 2014 IEEE international conference on acoustics, speech and signal processing (ICASSP). 2014. IEEE.
    14. Palmius, N., et al., Detecting bipolar depression from geographic location data. IEEE Transactions on Biomedical Engineering, 2016. 64(8): p. 1761-1771.
    15. Su, H.-Y., et al. Assessment of Bipolar Disorder Using Heterogeneous Data of Smartphone-Based Digital Phenotyping. in ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2021. IEEE.
    16. Vazquez-Montes, M.D., et al., Control charts for monitoring mood stability as a predictor of severe episodes in patients with bipolar disorder. International journal of bipolar disorders, 2018. 6(1): p. 1-15.
    17. Jadhav, A., D. Pramod, and K. Ramanathan, Comparison of performance of data imputation methods for numeric dataset. Applied Artificial Intelligence, 2019. 33(10): p. 913-933.
    18. Liu, X., X. Lai, and L. Zhang. A hierarchical missing value imputation method by correlation-based K-nearest neighbors. in Proceedings of SAI Intelligent Systems Conference. 2019. Springer.
    19. Beiwinkel, T., et al., Using smartphones to monitor bipolar disorder symptoms: a pilot study. JMIR mental health, 2016. 3(1): p. e4560.
    20. Lattuada, E., et al., Symptomatologic analysis of psychotic and non-psychotic depression. Journal of Affective Disorders, 1999. 54(1-2): p. 183-187.
    21. Altman, E.G., et al., The Altman self-rating mania scale. Biological psychiatry, 1997. 42(10): p. 948-955.
    22. Meng, D., et al. Frame attention networks for facial expression recognition in videos. in 2019 IEEE International Conference on Image Processing (ICIP). 2019. IEEE.
    23. Muthukrishnan, R. and R. Rohini. LASSO: A feature selection technique in predictive modeling for machine learning. in 2016 IEEE international conference on advances in computer applications (ICACA). 2016. IEEE.
    24. Chou, H.-C., et al. NNIME: The NTHU-NTUA Chinese interactive multimodal emotion corpus. in 2017 Seventh International Conference on Affective Computing and Intelligent Interaction (ACII). 2017. IEEE.
    25. Dupuis, K. and M.K. Pichora-Fuller, Recognition of emotional speech for younger and older talkers: Behavioural findings from the Toronto emotional speech set. Canadian Acoustics, 2011. 39(3): p. 182-183.
    26. Burkhardt, F., et al. A database of German emotional speech. in Ninth european conference on speech communication and technology. 2005.
    27. Dhall, A., et al. Emotiw 2018: Audio-video, student engagement and group-level affect prediction. in Proceedings of the 20th ACM International Conference on Multimodal Interaction. 2018.

    下載圖示 校內：2022-08-31公開
    校外：2022-08-31公開
    QR CODE