臺灣博碩士論文加值系統

細胞圖像自動分割是在多種生物醫學應用中的一項重要任務。目前影像切割有六大類：強度閾值(intensity thresholding)、特徵檢測(feature detection)、形態學濾波器(morphological ?ltering)、區域累積(region accumulation)、可變形模型擬合(deformable model ?tting)、及其他方法。

本論文提出利用細胞胞器的空間分布集群來切割3D影像中的個別細胞範圍。我們利用中國倉鼠卵巢細胞(CHO)的影像作測試。我們先將細胞核螢光影像用雙Otsu法切割出細胞核範圍，再將粒線體螢光影像用自適性區域臨界值法切割出粒線體範圍。接著計算粒線體和細胞核的空間質心及螢光亮度加權重心進行胞器分群，用分成同一群的粒線體的空間範圍作為單一細胞的範圍。在本論文中我們將測試並比較用不同的分群方法得到的細胞分割的正確率。

我們測試的分群方法包括高斯混和模型分群(Gaussian Mixture Model Clustering)、K-means clustering、階層式分群(Hierarchical Clustering)、正規化分割法(Normalized Cuts)。比較的基礎是使用手標3D影像中的所有細胞範圍，與個別分群法的細胞分割範圍計算重疊比例作為準確率。結果顯示在有加入細胞核的質心做為資料點的情況下，K-means以細胞核質心作為起始群中心的準確率(81.43%)及高斯混和模型的準確率(81.75%)比用平均連結聚合的階層式分群法(77.18%)高。在以細胞核質心作為起始群中心的K-means，有加入細胞核質心做為資料點(81.43%)或沒有的準確率(81.22%)相比，結果是相似的。在有加入細胞核的質心做為資料點的情況下，階層式分群法(Hierarchical Clustering)用平均連結聚合(Average linkage)(77.18%)和完整連結聚合(Complete linkage)(77.02%)的準確率相似。總體而言，K-means及高斯混和模型分群對於球形及短的細胞有較好的分群效果。用高斯混和模型加入細胞核質心分群的正確率最高(81.75%)。高斯混和模型無法整張影像細胞分割，因為邊界會有許多其他細胞的粒線體，造成粒線體群比細胞核數多。所以我們將正確率第二好的K-means不加入細胞核質心、並將細胞核質心當作起始中心的分群方法製作成使用者界面(GUI)。另外我們手標並將整張3D共軛交螢光顯微鏡影像輸入GUI測試，達到正確率66.71%。GUI方便使用者可以批次匯入大量影像檔案做細胞分割。本篇論文可以運用在分析染其他細胞胞器的影像，自動分割細胞。

Automatic segmentation of cell images is an essential task in a variety of biomedical applications. There are six main classes of approaches: intensity thresholding, feature detection, morphological ?ltering, region accumulation, deformable model ?tting, and other approaches.

In this thesis, we investigate whether spatial clustering of subcellular organelles is useful for 3D cell segmentation. We used CHO cell 3D images as our dataset. The nuclear channel is segmented by double Otsu methods and mitochondrial channel is segmented by adaptive local thresholding. We calculated the spatial centroid and weighted centroid of the mitochondria and nuclei, and then used unsupervised clustering to group the mitochondria. We used the spatial extent of mitochondria in the same group as individual cell regions. Because there are several unsupervised clustering methods, we hope to know which method yields higher accuracy for cell segmentation.

We compared the performance of GMM clustering, K-means, hierarchical clustering and normalized cuts methods. Regions of interest (ROI) for each cell in the 3D images are manually labeled slice-by-slice, and used as the gold standard for accuracy calculation. The following are results using methods that include nucleus centroids as data point. K-means clustering (81.43%) and GMM clustering (81.75%) with nucleus centroids initialization have higher accuracy than hierarchical clustering with average linkage (77.18%). We compared K-means with (81.22%) or without (81.43%) using nuclei centroids as initial cluster centers, and their accuracies are similar. Hierarchical clustering with nucleus centroids as data points with average (77.18%) or complete (77.02%) linkage has the same performance. Overall, K-means and GMM clustering in round and short cells have better accuracy than flat cells. GMM clustering with nucleus centroids as data points has the highest accuracy of 81.75%. GMM clustering is not suitable for whole field images, because there are many mitochondria from cells truncated by the image boundary, resulting in more mitochondrial clusters than nuclei. We designed a graphical user interface (GUI) system for K-means clustering without using nuclei centroids as initial cluster centers. The GUI was tested on another whole field 3D confocal image with manual cell ROI and achieved accuracy of 66.71%. Users can import a large number of image files for cell segmentation in our GUI. The proposed method can be applied to cell images with different subcellular organelle labels for automatic cell segmentation.

CONTENTS
CHINESE ABSTRACT i
ABSTRACT iii
CONTENTS v
LIST OF FIGURES vii
LIST OF TABLES ix
Chapter 1 Introduction 1
1.1 Cell Image Analysis 2
1.2 Cell Image Segmentation 3
1.2.1 Intensity Thresholding 4
1.2.2 Miscellaneous approaches 4
Chapter 2 Specific Aims 6
Chapter 3 Materials and Methods 7
3.1 Raw Data 7
3.2 Pre-processed Images 9
3.2.1 3D Median Filter 10
3.2.2 3D Gaussian Filter 12
3.3 Double Otsu’s Method 13
3.4 3D Adaptive Local Thresholding 15
3.5 Centroid and Weighted Centroid 18
3.5.1 Centroid 18
3.5.2 Weighted Centroid 20
3.6 Clustering Methods 21
3.6.1 K-means 22
3.6.2 Hierarchical Clustering 23
3.6.3 Gaussian Mixture Model Clustering 24
3.6.4 Normalized Cuts 26
3.7 Cell Segmentation Accuracy Calculation 29
Chapter 4 Result 31
4.1 Clustering Methods Comparison 32
4.2 Clustering Methods Comparison Using Weighted Centroids 37
4.3 Whole Field Clustering Result 42
4.4 Cell Segmentation GUI 45
Chapter 5 Discussion and Conclusion 47
References 49

LIST OF FIGURES
Fig. 1 1 Blob image. 4
Fig. 3 1 Z maximum intensity projection of a sample image. 7
Fig. 3 2 Channel 1: Mitochondria 8
Fig. 3 3 Channel 2: Differential interference contrast image 8
Fig. 3 4 Channel 3: Nuclei 8
Fig. 3 5 Window or mask of size 3×3×3 in 3D (Kumar et al., 2013) 10
Fig. 3 6 Example of 3×3×3 median filter calculation 11
Fig. 3 7 3D median filtering results (z-slices: 4, 7, and 14). 11
Fig. 3 8 3D Gaussian filtering results (z-slices: 4, 7, and 14) 12
Fig. 3 9 Double Otsu’s method 14
Fig. 3 10 3D reconstruction of double Otsu’s segmentation result 14
Fig. 3 11 Comparison of segmentation methods for mitochondrial image 16
Fig. 3 12 3D reconstruction of Otsu’s method mitochondrial binary image 16
Fig. 3 13 3D reconstruction of ALT method mitochondrial binary image 17
Fig. 3 14 3D reconstruction of ALT method mitochondrial and double Otsu’s method Nuclei binary image 17
Fig. 3 15 Centroid scatter plot of the image in Fig. 3-13 19
Fig. 3 16 Weighted centroid scatter plot of the image in Fig. 3-13 20
Fig. 3 17 K-means clustering with two clusters 22
Fig. 3 18 Hierarchical clustering 23
Fig. 3 19 Data points 25
Fig. 3 20 The estimated probability density contours for the mixture distribution. 25
Fig. 3 21 The Gaussian mixture model clustering fit using two clusters. 26
Fig. 3 22 A case where minimum cut gives a bad partition. 28
Fig. 3 23 The K-means clustering result 29
Fig. 3 24 The gold stander result 30
Fig. 3 25 The clustering confusion matrix 30
Fig. 4 1 Comparison of clustering methods 35
Fig. 4 2 Comparison of clustering methods 40
Fig. 4 3 Cell segmentation GUI 46

LIST OF TABLES
Table 4.1 Comparison of clustering methods 33
Table 4.2 Comparison of clustering methods 38
Table 4.3 Whole field comparison of clustering methods 43
Table 4.4 Whole field comparison of clustering methods 44
Table 5.1 Comparison of cell types and clustering methods. 47
Table 5.2 Comparison of cell types and clustering methods. 48

[1] Bengtsson, Ewert, C. Wahlby, and Joakim Lindblad. "Robust cell image segmentation methods." PATTERN RECOGNITION AND IMAGE ANALYSIS C/C OF RASPOZNAVANIYE OBRAZOV I ANALIZ IZOBRAZHENII. 14.2 (2004): 157-167.
[2] Bovermann, Till, Julian Rohrhuber, and Helge Ritter. Durcheinander–Understanding Clustering via Interactive Sonification." Proceedings of the 14th International Conference on Auditory Display. 2008.
[3] Coelho, Luís Pedro, Aabid Shariff, and Robert F. Murphy. "Nuclear segmentation in microscope cell images: a hand-segmented dataset and comparison of algorithms." Biomedical Imaging: From Nano to Macro, 2009. ISBI'09. IEEE International Symposium on. IEEE, 2009.
[4] Dawkins, Richard. In Bateson, Paul Patrick Gordon; Hinde, Robert A. Growing points in ethology: based on a conference sponsored by St. John's College and King's College, Cambridge. Cambridge, England: Cambridge University Press. , 1976.
[5] Gonzalez, Rafael C., and Richard E. Woods. "Digital image processing." Prentice Hall (2002): 462-463.
[6] Jyh-Ying Peng, Chung-Chih Lin and Chun-Nan Hsu. Adaptive Image Enhancement for Fluo-rescence Microscopy. Proceedings of the Conference on Technologies and Applications of Articial Intelligence (TAAI2010), Hsinchu, Taiwan, Nov 18-20, 2010.
[7] Kumar, K. Ganesh, and K. Kiran Kumar. "3D Median Filter Design for Iris Recognition." International Journal of Modern Engineering Research (IJMER), Vol. 3, Issue. 5, Sep - Oct. 2013 pp-3008-3011
[8] Luncien, M., Le Cam, and Jerzy Neyman, eds. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. University of California, 1967.
[9] McLachlan, G., and D. Peel. "Finite Mixture Models." Hoboken, NJ: John Wiley & Sons, Inc., 2000.
[10] Meijering, Erik, Oleh Dzyubachyk, and Ihor Smal. "Methods for cell and particle tracking." Methods Enzymol 504.9 (2012): 183-200.
[11] Meijering, Erik. "Cell segmentation: 50 years down the road [life sciences]."Signal Processing Magazine, IEEE 29.5 (2012): 140-145
[12] Michalet, Xavier, et al. "The power and prospects of fluorescence microscopies and spectroscopies." Annual review of biophysics and biomolecular structure32.1 (2003): 161-182.
[13] Monteiro, Fernando C., and Aurélio Campilho. "Watershed framework to region-based image segmentation." Pattern Recognition, 2008. ICPR 2008. 19th International Conference on. IEEE, 2008.
[14] Otsu, Nobuyuki. "A threshold selection method from gray-level histograms." System, Man and Cybernetics, IEEE Transactions on 9, (1975): 62-66.
[15] Prewitt, Judith, and Mortimer L. Mendelsohn. "The analysis of cell images*."Annals of the New York Academy of Sciences 128.3 (1966): 1035-1053.
[16] Rittscher, Jens. "Characterization of biological processes through automated image analysis." Annual review of biomedical engineering 12 (2010): 315-344.
[17] Segaran, Toby. Programming collective intelligence: building smart web 2.0 applications. "O'Reilly Media, Inc.", 2007.
[18] Shi, Jianbo, and Jitendra Malik. "Normalized cuts and image segmentation." Pattern Analysis and Machine Intelligence, IEEE Transactions on 22.8 (2000): 888-905.
[19] Srinivasa, Gowri, et al. "Active mask segmentation of fluorescence microscope images." Image Processing, IEEE Transactions on 18.8 (2009): 1817-1829.
[20] Tao, Wenbing, Hai Jin, and Yimin Zhang. "Color image segmentation based on mean shift and normalized cuts." Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on 37.5 (2007): 1382-1389.
[21] Tolles, W. E. "SECTION OF BIOLOGY: THE CYTOANALYZER—AN EXAMPLE OF PHYSICS IN MEDICAL RESEARCH*." Transactions of the New York Academy of Sciences 17.3 Series II (1955): 250-256.
[22] Zhang, Jin, et al. "Creating new fluorescent probes for cell biology." Nature Reviews Molecular Cell Biology 3.12 (2002): 906-918.