臺灣博碩士論文加值系統

阿米巴原蟲是一種會導致人類產生許多疾病的生物，例如有名的阿米巴性痢疾，全球每年估計有五億人感染過阿米巴性痢疾，十萬人死於阿米巴性痢疾。阿米巴原蟲導致的許多疾病其診斷方式必須仰賴病理學家以鏡檢方式判斷是否感染阿米巴原蟲。但此方式常會受到人為條件所限制與影響，例如醫護人員的生理狀況、心理狀況、專業知識等，而可能導致診斷的錯誤、降低診斷品質與效率。若能透過電腦來協助判斷，提升診斷的效率、判斷的準確度，降低人為因素對診斷的影響，使病理學家快速獲得疾病診斷之相關資訊，在病情的治療上能夠更加專注。因此，在本篇研究中，提出一個以影像為基礎的阿米巴原蟲偵測方法。
我們所提出的方法主要分為兩個部分，第一部分是針對整張影像做處理，而第二部分是針對處理後的影像將其做區域標籤(Region Labeling)，對切割不足的區域做處理。首先，將彩色影像轉為灰階影像，再對影像做對比強化及輪廓強化。將強化後的影像做二值化，再對此黑白影像做切割，對切割後的影像做區域標籤(Region Labeling)，透過形狀判斷是否切割不足，若切割不足，便將此區域對應回原始的灰階影像，再將之做對比、輪廓的強化以及二值化。對此黑白影像做切割，以及雜訊的去除。最後剩下的部分即是阿米巴原蟲。
本研究總共使用了115張影像，其中包含了146個阿米巴原蟲。透過精確率(Precision)、召回率(Recall)與F-Measure來評估本研究之結果，可看出本研究之方法能夠輔助偵測到阿米巴原蟲，但卻未達理想值，所以我們也提出了一些本研究所遇到之問題與困難及未來發展。

The amoeba can cause people to infect many diseases. For example, the famous amebic dysentery, each year estimated half a billion people worldwide infected with amoebic dysentery, 100,000 people died of amoebic dysentery. The diagnostic modality of amoebic diseases relies on pathologists use microscopy. However, this method is heavily impacted by human’s behaviors such as physical factors, mental factors, and professional knowledge, etc. That can lead to diagnostic errors, reduce diagnostic quality and efficiency. If the computer technique can assist pathologist to diagnose, that can improve diagnostic efficiency, accuracy and lessen the impact of human factors. Let pathologists obtain information about the diagnosis of the disease quickly, and in the treatment of the illness can be more focused. Therefore, in this paper, we propose an amoeba detection methods.
Our proposed method is divided into two parts. In the first part, we process the entire image. In the second part, we do Region Labeling against the processed images, then have further processing on those incomplete regions. First, we convert color images into grayscale images, enhance the images contrast and contour. Then we convert the images into binarization and segment the images. We process images after segmentation by Region Labeling and then confirm whether every region segment sufficiently through shapes. If not, this region corresponds to the original grayscale image, enhance the region contrast and contour. Then we convert the region into binarization, segment the region and remove noise. The remaining regions are amoebae.
In this paper, we use a total of 115 images, which contains 146 amoebae. We use Precision, Recall and F-Measure to evaluate the results of this paper. It can be seen our proposed method can detect amoebae, but the result is not as expected. So, we offer some challenges and difficulties in this paper.

摘要 i
Abstract ii
List of Contents iv
List of Tables vi
List of Figures vii
Chapter 1. Introduction 1
Chapter 2. Related Works 5
2.1. Amoeba Image Characteristics 5
2.2. Histogram Equalization 6
2.3. Run-Length 6
2.4. Otsu’s Method 8
2.5. Mathematical Morphology 9
2.5.1. Dilation and Erosion 9
2.5.2. Opening 10
2.5.3. Region Filling 11
2.6. Region Labeling 12
2.7. F-Measure 16
Chapter 3. Amoeba Detection 18
3.1. Color Images Convert to Grayscale Images 19
3.2. Image Enhancement 20
3.2.1. Histogram Equalization 20
3.2.2. Gamma Correction 20
3.2.3. Run-Length 21
3.3. Image Binarization 22
3.4. Image Segmentation 23
3.4.1. Region Filling 23
3.4.2. Opening 24
3.5. Confirm Region Segmentation 24
3.5.1. The Ratio of Region Area and Region Perimeter Convert to Circle Area 25
3.5.2. The Ratio of Region Area and Region Maximum Distance Convert to Circle Area 26
3.6. Region Processing 26
3.7. Remove Small Regions 28
Chapter 4. Results and Discussions 30
4.1. Samples of Experiment 30
4.2. Result of Experiment 30
4.3. Challenges of Experiment 32
4.3.1. The Different Color Depth of Images 32
4.3.2. The Inconsistent Shapes of Amoebae 33
4.3.3. The Different Size of Amoebae Area in Images 34
Chapter 5. Conclusions and Future Work 35
Reference 37

[1] 衛生福利部疾病管制署“阿米巴性痢疾” http://www.cdc.gov.tw/professional/index.aspx
[2] 衛生福利部疾病管制署“棘狀阿米巴”
http://www.cdc.gov.tw/professional/index.aspx
[3] A. M. Baig, “Pathogenesis of amoebic encephalitis: Are the amoebas being credited to an ''inside job'' done by the host immune response?”, Acta Trop, vol.148,pp. 72-78, 2015.
[4] A. M. Baig, “Granulomatous amoebic encephalitis: ghost response of an immunocompromised host?”, Journal of Medical Microbiology, vol.63, pp.1763-6, 2014.
[5] C. D. Gregorio, F. Rivasi, N. Mongiardo, B. D. Rienzo, S. Wallace, G.S. Visvesvara, “Acanthamoeba meningoencephalitis in a patient with acquired immunodeficiency syndrome”, Archives of Pathology & Laboratory Medicine, vol.116 (12), pp. 363–5, 1992.
[6] C. J. van Rijsbergen, “Information Retrieval (2nd ed.)”, Butterworths, 1979.
[7] D. Zhang, W. J. Park, S. J. Lee, K. A. Choi, S. J. Ko, “Histogram partition based gamma correction for image contrast enhancement”, Consumer Electronics (ISCE), IEEE 16th International Symposium, pp. 1-4, 2012.

[8] G. Xu, J. Su, H. Pan, Z. Zhang, H. Gong, “An image enhancement method based on gamma correction”, 2009 Second International Symposium on Computational Intelligence and Design, vol. 1, pp. 60-63, 2009.
[9] K. Chan and C. C. Chang, “Image Matching Using Run-length Feature”, Pattern Recognition Letters, vol. 22, no.5, pp. 447-455, 2001.
[10] N. Otsu, “A Threshold Selection Method from Gray-level Histogram”, IEEE Transactions on System Man Cybernetics, vol. 9, no. 1, pp. 62-66, 1979.
[11] R. W. Hall, T. Y. Kong, A. Rosenfeld, “Shrinking binary images”, Machine Intelligence and Pattern Recognition”, vol. 19, pp. 31-98., 1996.
[12] S. Pierre, “Morphological Image Analysis: Principles and Applications”, Springer-Verlag, pp. 173-174, 1999.
[13] Wikipedia The Free Encyclopedia, “Mathematical Morphology”, https://en.wikipedia.org/wiki/Mathematical_morphology
[14] WHO | World Health Organization, http://www.who.int/