利用定量相位顯微術判斷視網膜色素上皮細胞之細胞凋亡

Abstract

藥物初步開發的流程分為標靶篩選及表徵篩選，而後者為將候選藥物化合物加入培養之細胞或者組織，並使用檢驗的方法量化藥物化合物的療效。細胞死亡檢驗是最常見的表徵篩選檢驗，細胞死亡大致分為細胞凋亡及細胞壞死，區分何種細胞死亡可以提供藥物篩選額外的資訊。使用螢光染劑以觀測不同種類的細胞死亡現象是普遍被接受的方法，其具備高靈敏度及特異性，並且已經出現商業系統，包含顯微鏡、影像處裡軟體及自動資料庫建立流程。但是使用螢光染劑需要樣本前處理、等待染劑反應時間、觀測時可能發生光漂白現象、需要昂貴的螢光染劑等，這些缺點降低了細胞死亡檢測的通量。 定量相位顯微術是一種免標記、快速、可動態觀測的技術，其拍攝之相位影像包含樣本厚度及樣本內部折射率的資訊，在細胞凋亡的檢測上具有優勢，有機會發展為高通量、低成本的檢測工具。最近有研究使用定量相位顯微術判斷細胞壞死及細胞凋亡，在本次研究希望可以使用更多能夠描述細胞凋亡現象的特徵以提升準確率。此研究以螢光標定細胞並使用螢光顯微鏡及定量相位顯微鏡觀測同一視野，從各顆細胞之相位影像萃取出可能代表細胞凋亡之形態變化的特徵，最後將螢光作為判定細胞凋亡的標準以訓練分類器，使用分類器區分細胞是否為細胞凋亡。結果顯示，使用細胞圓形性、離心率、光學體積、細胞核邊界梯度、細胞邊緣相位平均值、細胞邊緣及中央相位比值對於分類結果有最佳的表現，初步分類的結果顯示預測的正確率約87%。其分析流程中，影像分割、特徵的計算方式等等，仍有最佳化的空間，預期未來可以提升正確率並且加速判斷流程，發展為高通量的檢測工具。

In the development of drug, drug discovery can be divided into target-based screening and phenotypic screening. The latter is adding drug in cultural cell or tissue and quantifying the performance of those candidates. Cell death assay is one of the most common assay in phenotypic screening, and it includes apoptosis and necrosis assay. Different sorts of cell death represent different information for phenotypic screening. It is prevalent to use fluorescence-based method to sensitive and specific observe cell death. It has developed a commercial system integrated with microscope, image processing software and database. However, there are some drawbacks such as sample preprocessing, time-consuming incubation, photobleaching, expensive fluorescent reagent, which decrease the throughput of assay. Quantitative phase imaging(QPI) is a label-free, fast, time-lapse technique which generates quantitative phase images related to both the intracellular refractive index and the cell thickness. QPI has the potential to be developed as a high throughput and cost-efficient tool to identify apoptosis. In this research, we expect to find more features to increase the accuracy of prediction. We apply fluorescence-based method and QPI to observe cells in the same FOV, and extract some features from each cell in phase images. A classifier is trained by using these features and the fluorescent label which is regarded as the ground truth of apoptosis. In the result, the accuracy of identification is approximate 87% by using features, including circularity, eccentricity, optical volume, nuclear edge gradient index, peripheral phase, and the ratio of peripheral and central phase. The procedures such as image segmentation, calculation of features need to be improved. To sum up, we expect QPI can develop into a high throughput tool after enhancing the accuracy and accelerating the procedures.