| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊工程學系 | zh_TW |
| DC.creator | 林姿伶 | zh_TW |
| DC.creator | Zi-Ling Lin | en_US |
| dc.date.accessioned | 2021-7-23T07:39:07Z | |
| dc.date.available | 2021-7-23T07:39:07Z | |
| dc.date.issued | 2021 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=108522088 | |
| dc.contributor.department | 資訊工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 一群細菌中存在不同程度的抗生素抗藥性，對抗生素的反應程度不同而導致難以精確施用抗生素。若高強度的抗藥性細菌在抗生素環境下生存，將造成嚴重威脅。因此需要及時提供細菌抗藥性訊息，以降低抗藥性的發生。近年微生物檢驗開始採用基於光學誘導介電泳（Optically Induced Dielectrophoresis, ODEP）的微流體系統，先前研究亦透過該系統對不同微生物或細胞得到不同資訊，該系統能根據細菌具備的極性不同，分離細菌並移動到不同位置。為了找出細菌在流體裡面的所在位置進而快速辨識細菌影片的抗藥性。本研究藉由長庚醫院發明的光誘導介泳系統錄製的大腸桿菌（Escherichia coli）與金黃色葡萄球菌（Staphylococcus aureus）分離影片，使用兩種特徵選取法並基於機器學習從細菌的特徵與分布位置的影片辨認細菌是否抗藥。本研究首先在特徵選取中結合了面積規則與卷積神經網路（Convolutional Neural Network）辨認大腸桿菌與金黃色葡萄球菌的細菌特徵之準確度分別達0.90與1.0，而後將細菌辨認結果與該座標結合轉換成影片資料的特徵，並以羅吉斯回歸（Logistic Regression）分類器所建構之模型，透過留一法交叉（Leave-one-out Cross Validation）驗證辨認細菌影片的細菌抗藥性訓練資料準確率為0.9521。此實驗結果可提供影片中的細菌位置或是提供影片是否抗藥，幫助臨床實驗不同程度的細菌抗藥性相關性，未來可提供醫師更準確的用藥。 | zh_TW |
| dc.description.abstract | There are different levels of antibiotic resistance in a group of bacteria, and the level of response to antibiotics is different, which makes it difficult to accurately administer antibiotics. In recent years, microbiological testing has begun to use optically induced dielectrophoresis-based microfluidic system. The system can separate bacteria and move to different locations according to their polarities. In order to quickly detect the antibiotics resistance of the bacteria in the video through the location of the bacteria in the microfluid. In this study, we used the Escherichia coli and Staphylococcus aureus isolation videos recorded by the optically induced dielectrophoresis system invented by Chang Gung Memorial Hospital to combine two kinds of feature extractions and machine learning to quickly determine whether the video is antibiotics resistant from the distribution and characteristics of bacteria. In feature extraction, this research combines the area rule and convolutional neural network to recognize Escherichia coli and Staphylococcus aureus with accuracy of 0.90 and 1.0 respectively. Then, the bacteria identification results combined with the coordinates are converted as the features of the video data, and the model constructed by logistic regression classifier. The accuracy of identifying bacterial resistance of videos in ODEP data through leave-one-out cross-validation was 0.9521. This study can help clinical trials with varying levels of antibiotics resistance result and provide more appropriate medications, or uses the location of bacteria. | en_US |
| DC.subject | 光介電泳 | zh_TW |
| DC.subject | 影像辨識 | zh_TW |
| DC.subject | 抗生素抗藥性 | zh_TW |
| DC.subject | 機器學習 | zh_TW |
| DC.subject | optically induced dielectrophoresis | en_US |
| DC.subject | image recognition | en_US |
| DC.subject | antibiotic resistance | en_US |
| DC.subject | machine learning | en_US |
| DC.title | 基於光誘導介電泳影像處理檢測流體抗藥性 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Detection of Microfluidic Bacterial Resistance Based on Optically Induced Dielectrophoresis (ODEP) Image Processing | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |