- 期刊
Application of an Image Processing Technique to Estimate the Concentration of Beads and Trapped Escherichia coli (E. coli) on a Dielectrophoretic Chip
應用影像處理技術估計介電泳晶片乳膠顆粒及捕抓大腸桿菌濃度
摘要
介電泳晶片（Dielectrophoretic chips, DEP chips）因具有分離捕抓特定病原菌、使用的樣本量較少、與傳統生化分析方式比較可有較快速的作業等優點，近年來它被廣泛的應用在醫藥、食品及微生物分析等相關的研究上。為有效的使用此種晶片，本研究利用5mm及2 mm直徑的乳膠顆粒（Beads）及影像處理技術，建立評估晶片分離及捕抓效率的方法，應用的技術包括遮罩處理及雙閥值法等。另外，本研究應用由灰階空間矩陣（Gray-Tone Spatial-Dependent Matrix, GTSDM）計算13種紋理特徵值，藉以分辨晶片背景與所捕抓大腸桿菌（E. coli）的區域影像。試驗結果顯示，晶片在乳膠顆粒的分離與捕抓效率平均分別為91.1%及87.6%，而透過影像處理估計乳膠顆粒濃度，平均判定係數R^2達0.95，迴歸的濃度每微升（ml）界於5 × 10^4 到5 × 10^6顆粒。利用13種紋理特徵，區別大腸桿菌區域及晶片背景區域影像結果顯示，第11種特徵─熵差（difference entropy）具有最低的辨識誤差0.039，此時估計大腸桿菌所占的面積比例是9.0%。後續的研究，建議建立捕抓細菌面積與實際細菌濃度的迴歸，以直接推估大腸桿菌或其他目標菌的濃度。
並列摘要
Dielectrophoretic chips (DEP) have been widely studied and have recently been used in medical, food, and microbiological assay applications. The advantages of these chips include the ability to sort and trap specific pathogens, the limited amount of sample required, and the convenient processing procedure compared to conventional bio-chemical processes. To facilitate the efficient use of DEP chips, this study proposes an evaluation process for the sorting and trapping performance of DEP chips that uses an image processing technique utilizing 5 mm and 2 mm beads. Our technique involves the use of one masking operation and two thresholding operations. An analysis was also performed to determine 13 features from a microscopic image of an E. coli pathogen trapped by a DEP chip. These texture features were calculated from a Gray-Tone Spatial-Dependent Matrix (GTSDM) derived from the E. coli image. The results showed that the average sorting efficiency and trapping efficiency were 91.1% and 87.6%, respectively. The correlation between the number of beads and the area of the bead region in the digital image was observed to have an average R^2 value about 0.95 in beads concentration between 5 × 10^4 to 5 × 10^6 particles/ml. Among the 13 examined texture features, difference entropy (feature 11) was seen to have the lowest error (0.039) for distinguishing E. coli and the background image. Under the circumstance, proportion of E. coli region was about 9.0%. Further study is suggested, in which correlation of texture features are calculated for various concentrations of E. coli for the purpose of formulating a predictive equation.