電機資訊學院: 生醫電子與資訊學研究所指導教授: 林啟萬; 邱銘章胡翔崴Hu, Hsiang-WeiHsiang-WeiHu2017-03-022018-07-052017-03-022018-07-052016http://ntur.lib.ntu.edu.tw//handle/246246/272629憂鬱症盛行率相當高，為全球人類失能最重要疾病之一，且造成社會經濟嚴重的負擔，關於重度憂鬱症的療法，抗憂鬱藥物，物理 性刺激電療(ECT : electro-convulsive treatment)與磁刺激(rTMS : repetitive transcranial magnetic stimulation)皆是治療的選擇，在各類治療上的高副作用及各治療療效限制之下，期望能找出預測憂鬱症療效的客觀判定方式，增強重度憂鬱症精準治療依據，過去的研究曾藉由腦波偵測發現腦前額葉的theta波段能夠預測重度憂鬱症的療效，並且在藥物治療及磁刺激治療的療效預測上有不錯的成果，但關於電療療效的預測方面尚未有明確的成果，僅驗證theta波段與電療療效有相關性 。 因此，本研究將驗證運用向量支持機(SVM : Support vector machine)對於腦波theta波段的分析來預測電療與藥物的治療效果，並且取出Theta cordance、頻譜量值、近似熵、變異數等特徵值，來進行多維度分類分群，分析電療以及藥物治療的腦波訊號對於療效的關聯性，來預測最佳的抗憂鬱療法，研究預測模型的成果達到藥物短期療效上，精準度為83.1%,靈敏度為81.9%及特異度為78.8%，而藥物長期療效上，精準度為80.3%,且電療療效短療效上，特異度為81.7%及靈敏度具79.2%，電療療效精準度 79.5%,且特異度為78.2%及靈敏度為76.1%，而ROC曲線積分面積(AUC)，可得藥物療效短期組高達0.852，藥物療效長期達0.837，電療療效短期預期達0.814，具有初步性的突破，未來隨著收案數的增加將會有更精準的成果。 未來建立臨床資料庫之後，將用以提供病患在治療上的精準選擇判定，免於受到在ECT副作用之苦卻結果無效的風險，或是在治療上可避免用錯治療方式而延誤造成惡化。The prevalence of depression is very high, and it will be one of the most important and incapacitating human disease worldwide. Major depressive disorder also cause severe social and economic burden. For the therapy of severe depression, antidepressant medication, ECT (electro-convulsive treatment) and rTMS (repetitive transcranial magnetic stimulation) treatment is one of the options. Because of treatment in high side effects and treatment efficacy limit, find out the effect of depression predict by objective determination, and increasing the precision of severe depression treatment based on existing studies. In the past, It have found that the brain detected by electroencephalogram (EEG) can predict the effect of efficacy prefrontal by theta band. There have been good results for prediction of drug efficacy and magnetic stimulation efficacy, but there is no related outcomes for ECT. It just verify that theta band is correlated with the efficacy of ECT. Therefore, this study will validate that using SVM: Support vector machine) by EEG theta band will predict the therapeutic effect of ECT and drugs. This study uses a series of feature such us energy, variance, approximate entropy and theta cordance to analyze multi-dimensional clustering classification to find correlation efficacy of electrotherapy and drug treatment by EEG and to predict the best antidepressant therapy. The results of the forecast model is that the prediction for the short-term efficacy of the drug is accuracy of 83.1%, sensitivity of 81.9% and specificity of 78.8%, and the long-term efficacy of the drug is the accuracy is 80.3%, specificity of 81.7% and sensitivity of 79.2%, and ECT efficacy is accuracy of 79.5% , specificity of 78.2% and a sensitivity of 76.1%. The ROC curve integral area (AUC) for the short-term efficacy of the drug group is 0.852, and for the long-term efficacy of the drug group is 0.837, short-term efficacy of ECT is 0.814 .The research is kind of breakthrough . In the future, with the increasing number of cases will receive more precise results. After establishing clinical data in future, it will be used to provide accurate selection in the treatment decision. There will be free from the side effects of ECT having the risk of invalidating the results. The wrong treatment can be prevented to avoid the deterioration of delayed treatment.1888470 bytesapplication/pdf論文公開時間: 2020/8/30論文使用權限: 同意有償授權(權利金給回饋學校)重度憂鬱症電器痙攣治療近似熵變異數向量支持機major depressive disorderelectro-convulsive treatmentcordanceapproximate entropyvariancesupport vector machine[SDGs]SDG3thesis10.6342/NTU201603552http://ntur.lib.ntu.edu.tw/bitstream/246246/272629/1/ntu-105-R03945009-1.pdf