自動計算癌細胞群落數之影像分析系統

Abstract

隨著生物醫學上應用影像處理等技術來協助判讀與研究的需求增加 ，從前需要大量人力觀察辨識的工作也日漸需要一套標準化的系統來協助處理，以降低在時間上與人力上所需的大量資源。以肝癌治療效果之檢測為例，本實驗選用國人獨有的HA22T肝癌細胞，當前的標準方法為克隆檢驗(Clonogenic assay)，其步驟為：種植(seeding)過程、治療過程、等待過程與染色定位及計數過程，其中可以節省人力資源與時間的步驟就在於讓計數過程之自動化；且在此類辨識與計數過程中會與專業訓練成熟度存在密切關係，研究人員的主觀看法皆可能影響到最後的結果，因此極需要客觀的資訊工具來輔助。所以本研究主要運用影像處理的技術配合上模糊推論系統來對培養皿內部的群落數(colony)來做辨識與計數，過程中會先用掃描器來擷取培養皿影像，將影像儲存到電腦中，以影像處理的方法來進行分析，先利用霍氏轉換(Hough transform)找到培養皿內部相對位置，影像分割部分是用影像相減的方法來處理，特徵擷取方面是憑藉著經驗法則來擷取，配合上模糊推論系統來決策，計算出影像中的群落數以完成目標。本論文的主要貢獻有三，第一，解決此類研究中用面積來判斷群落內細胞數目時所發生的分佈密集面積小數量多以及分佈稀疏面積足夠數量卻不足的問題。第二，提出一個以模推論系統(fuzzy inference system, FIS)為核心的模糊癌細胞群落辨識系統，有成功地分辨出肉眼無法辨識的癌細胞群落。第三，本論文系統計數結果可取代人工計數結果。

Recently, requirements of image processing techniques used for helping biomedical diagnosis have become popular. In traditional systems, they’re time consuming because a lot of observations and reorganizations of human resources are needed. For solving this problem, a standard system that decreases lots of time and extra human resources is needed. About this, in this thesis, an image analysis system for automatic counting cancer cell colonies is proposed. In a detection of a curative effect of liver cancer, the Clonogenic assay is a golden standard. It’s used to assay the cells that are named HA22T of liver cancer. The steps of Clonogenic assay consist of seeding process, treating process, waiting process, making a location by dyeing, and counting process. This thesis proposed an automatic counting system to take place of counting by human for solving the problem that mentioned above. The automatic counting system consists of an image process and a fuzzy inference system (FIS). In an image process, the scanner is used to scan the image of a dish and store the scanned images into the computer. After that, an image process method that called Hough transform is used to find the relative position of the dish. After finding the relative position of the dish, an image subtraction is used to separate targets image and backgrounds image and perform a feature extraction according to the experience of a doctor. In the FIS, the total number of the cancer cell colonies is distinguished and calculated. The advantages of this thesis are summarized as follows: 1) the proposed system can distinguish whether the cells form a dense or sparse region in the colony; 2) the proposed system adopts a fuzzy inference system (FIS) to obtain a better performance of distinguishing the cancer cell colonies; 3) the proposed system can take place of counting by human.