酵母菌細胞之分割與分類

Segmentation and Classification of Yeast Cell

張浩翔(Hao-Hsiang Chang)

指導教授 ： 林維暘 ； 蔡佳玲

國立中正大學/工學院/資訊工程學系/碩士(2016年)

摘要

本篇論文提出一個自動化酵母菌細胞切割及分類的方法來幫助生物學家處理分析大量的酵母菌細胞影像。我們的方法分成兩個部分，細胞切割及細胞分類。在細胞切割這部分我們首先結合 HOG (Histogram of Oriented Gradients) 和頂帽轉換兩種方法來對細胞做種子點偵測，再以此種子點為基礎做分水嶺切割。在分類的部分，我們使用通用傅立葉描述元作為形狀特徵來區分不同酵母菌細胞形狀之間的差異，並透過支持向量機器 (Support vector machine) 將酵母菌分成五類。我們的實驗資料是由 Iona 大學生物學系 Eric Muller 教授所提供的 34 張變異酵母菌細胞影像。本論文的細胞切割結果的 F1-score 達到 0.8976。雖然在整體系統的實驗中 (以本論文的切割結果所做的細胞分類)，五類細胞的分類結果受到切割效果及異常細胞的影響效果較差。但若改以 ground truth 來做四類細胞 (不包含異常細胞) 的分類就會有較好的分類準確度。

關鍵字

酵母菌 ； 細胞切割 ； 分水嶺 ； 細胞分類

並列摘要

In this paper, we present a fully automatic segmentation and classification method to help biologist process large-scale analysis on yeast cell images. Our proposed method consists of two major parts, namely cell segmentation and cell classification respectively. In the segmentation part, we detect the seeds of yeast cells and background using Histogram of Oriented Gradients and top-hat transformation, then we perform watershed segmentation based on these seed points. In the classification part, we indicate the yeast cell with different category by generic fourier descriptor shape features, and pass through Support Vector Machine to classify yeast cell into five categories. In our experimental, we have collected 34 images from Professor Eric Muller, Department of Biology, Iona College. The F1-score of our cell segmentation result is 0.8976. The whole system in our experiment (cell classified by our segmentation result) has a poor classification result in five categories affected by segmentation result and abnormal cell. However, we have a better classify efficient when we classify four categories without abnormal cell and use ground truth to do our classify experiment.

並列關鍵字

Yeast Cell ； Cell Segmentation ； Watershed ； Cell Classification

參考文獻

Cell nuclei and cytoplasm joint segmentation using the sliding band lter. IEEE

[2] Bahram Parvin, Qing Yang, Ju Han, Hang Chang, Bjorn Rydberg, and Mary Helen

Barcellos-Ho . Iterative voting for inference of structural saliency and characterization of subcellular events. IEEE Transactions on Image Processing, 16 (3):615-623, 2007.

saliency through iterative voting. In Pattern Recognition, 2004. ICPR 2004.

Proceedings of the 17th International Conference on, volume 1, pages 63-66.