應用電腦視覺於腎臟組織切片之腎小球偵測研究

Abstract

在腎臟疾病的診斷中，篩檢方式主要有下列幾種：尿液常規檢查、抽血檢查、腎臟影像學檢查以及腎臟組織切片檢查。而其中腎臟組織切片檢查是許多慢性腎臟病的重要確診方法。目前腎臟組織切片檢驗方法中，主要依靠專業檢驗人士來完成工作，而其必須用肉眼從光學顯微鏡下觀察與辨識腎小球，此種方式在人力及時間上的耗費都相當大。 本論文主要研究對象為腎臟組織切片影像，目標是利用電腦視覺，能夠在廣泛的案例中將腎小球從組織影像的複雜背景中自動偵測出來，進一步能夠方便專業檢驗人員進行分析，進而增加判讀品質、提高診斷正確性。 本論文主要貢獻如下：第一，針對背景複雜的腎臟組織切片影像，本論文提出一些背景特徵，可有效地簡化背景。第二，基於第一項貢獻所得到的結果影像，本論文提供了一個有效判斷腎小球的方法，經由實驗結果證實本方法辨識結果相當良好。

There are some ways to diagnose kidney disease, including blood tests, urine tests, image tests, and removing a sample of kidney tissue for testing. Examining all kidney tissue is an important method for diagnosing kidney disease. Recently, the clinical observation of kidney tissue image is done by a medical technologist who spends plenty of time examining the entire tissue image in a light microscope. In order to help the physician analyze tissue image easily and correctly, the aim of this thesis is to automatically detect glomerulus from kidney tissue image by using computer vision technique. The contributions of the thesis are summarized as follows: 1) Some background features are proposed based on other tissue definition. By using these features, some background of the image can be removed quickly and effectively, and the task of glomerulus detection is become more simplified. 2) Based on the above result, we proposed an effective method to recognize glomerulus. Experiment results indicate that the proposed method has good performance for detecting glomerulus.