運用統計方法及影像處理計算顯微影像中結腸癌細胞凋亡之比例

Abstract

本研究的目的是運用影像處理以及統計分析方法，從結腸癌細胞的電子顯微影像中，自動化估計出自然凋亡細胞的比例。這個比例在治療癌症的體外試驗中可以提供診斷的資訊。 分水嶺轉換是對可見光的顯微影像進行初始化的分割，而無母數的檢定，例如Kolmogorov-Smirnov檢定是將鄰近相似的區域合併成一個區域的方法，這些被分割出來的區域是被用來當成訓練資料，從中獲得細胞的平均半徑以及標準差。用大小不同的濾波器做摺積的目的是偵測細胞在可見光和螢光顯微影像上的位置以及個數，再用型態學中的方法來記數細胞的個數以及比例。這些比例的信賴區間是用來評估有關臨床影像自動化檢查方法的表現成效。

This study is aimed to estimate the ratio of apoptosis cells for colon cancer automatically from microscope images with image processing and statistical analysis methods. This ratio will provide diagnosis information for cancer treatments with in vitro cells. Watershed transform is used to obtain the initial segmentation in light microscope images. Nonparametric tests, such as the Kolmogorov-Smirnov test, will be used to merge similar regions. These segmented regions will be used as the training set to obtain the average radius of cells and their standard deviations. Based on these sizes, convolution filters will be proposed to detect the number and location of cells in both images of light and fluorescentmicroscopes. Morphology methods are used to count the number of cells and the ratio. The confidence intervals of the ratios are computed to evaluate the performance of this automatic approach with respect to expert inspection for clinical images.