請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/31523
完整後設資料紀錄
DC 欄位 | 值 | 語言 |
---|---|---|
dc.contributor.advisor | 陳中明(Chung-Ming Chen) | |
dc.contributor.author | Chih-Pin Yen | en |
dc.contributor.author | 嚴志彬 | zh_TW |
dc.date.accessioned | 2021-06-13T03:14:13Z | - |
dc.date.available | 2006-08-09 | |
dc.date.copyright | 2006-08-09 | |
dc.date.issued | 2006 | |
dc.date.submitted | 2006-08-03 | |
dc.identifier.citation | 1. http://www.doh.gov.tw/statistic/index.htm
2. Yeh WC, Huang SW and Li PC. Liver fibrosis grade classification with B-mode ultrasound. Ultrasound in Med. & Biol. 2003; 29: 1229-1235. 3. Ogawa K, Fukushima M, Kubota K, Hisa N. Computer-aided diagnostic system for diffuse liver diseases with ultrasonography by neural networks. IEEE Trans. Nuclear Science 1998; 45: 3069-3074 4. Hong MH, Sun YN, Lin XZ. Texture feature coding method for classification of liver sonography. Computer Med. Imaging and Graphics 2002; 26: 33-42. 5. Bleck JS, Ranft U, Gebel M, et al. Random field models in the textural analysis of ultrasonic images of the liver,” IEEE Trans. Med. Imaging 1996; 15: 796-801. 6. Sun YN, Horng MH, Lin XZ, Wang JY. Ultrasonic image analysis for liver diagnosis: A noninvasive alternative to determine liver disease. IEEE Engineering in Medicine and Biology 1996; 15: 93-101. 7. Mojsilovic A, Popovic M, Markovic S, Krstic M. Characterization of visually similar diffuse diseases from B-scan liver images using nonseparable wavelet transform. IEEE Trans. Med. Imaging 1998; 17: 541-549. 8. Kadah YM, Farag AA, Zurada JM, Badawi AM, Youssef AM. Classification algorithms for quantitative tissue characterization of diffuse liver disease from ultrasound images. IEEE Trans. Med. Imaging 1996; 15: 466-478. 9. Wu CM, Chen YC, Hsieh KS. Texture features for classification of ultrasonic liver images. IEEE Trans. Med. Imaging 1992; 11: 141-152. 10. Lang M, Ermert H, Heuser L. In-vivo study of on-line liver tissue classification based on envelope power spectrum analysis. Ultrasonic Symposium Proceedings 1990; 3: 1345-1348. 11. Akiyama I, Saito T, Nakamura M. Taniguchi N, Itoh K. Tissue characterization by using fractal dimension of B-scan image. Ultrasonic Symposium Proceedings 1990; 3: 1353-1355. 12. Cohen FS, Georgiou G. Detecting and estimating structure regularity of soft tissue organs from ultrasound images. Proceedings of IEEE International Conference on Image Processing 1995; 2: 488-491. 13. Gauch JM. Image segmentation and analysis via multiscale gradient watershed hierarchies. IEEE Trans. Image Processing 1999; 8: 69-79. 14. Vincent L, Soille P. Watersheds in digital spaces: an efficient algorithm based on immersion simulation. IEEE Trans. PAMI 1991; 13: 583-597. 15. Karaman K, Kutay M, Bozdagi G. An adaptive speckle suppression filter for medical ultrasonic imaging. IEEE Trans. Med. Imaging 1995; 14: 283-292. 16. Yang PM, Huang GT, Lin JT, et al. Ultrasonography in the diagnosis of benign diffuse parenchymal liver diseases: A prospective study. J Formosan Med Assoc 1988; 87: 966-977. 17. Goyal AK, Pokharna DS, Sharma SK. Ultrasonic diagnosis of cirrhosis reference to quantitative measurements of hepatic dimensions. Gastrointest Radiol 1990; 15: 32-34. 18. Vilgrain V, Lebrec D, Menu Y, Scherrer A, Nahum H. Comparison between ultrasonographic signs and the degree of portal hypertension in patients with cirrhosis. Gastrointest Radiol 1990; 15: 218-222. 19. Joseph AEA, Saverymuttu SH, Al-Sam S, Cook MG, Maxwell JD. Comparison of liver histology with ultrasonography in assessing diffuse parenchymal liver disease. Clinical Radiology 1991; 43: 26-31. 20. Ferral H, Male R, Cardiel M, Munoz L, Ferrari FQ. Cirrhosis: Diagnosis by liver surface analysis with high-frequency ultrasound. Gastrointest Radiol 1992; 17: 74-78. 21. Aubé C, Oberti F, Korali N, et al. Ultrasonographic diagnosis of hepatic fibrosis or cirrhosis. Journal of Hepatology 1999; 30: 472-478. 22. Xu Y, Wang B, Cao H. An ultrasound scoring system for the diagnosis of liver fibrosis and cirrhosis. Chinese Medical Journal 1999; 112: 1125-1128. 23. Botros NM. A PC-Based tissue classification system using artificial neural networks. IEEE Trans. Instrumentation and Measurement 1992; 41: 633-638. 24. Keller JM, Chen S, Grownover RM. Texture description and segmentation through fractal geometry. Comput Vis Graph Image Process 1989; 45: 150-166. 25. Landini G, Rippin JW. How important is tumor shape? Quantification of the epithelial-connective tissue interface in oral lesions using local connected fractal dimension analysis. J Pathol 1996; 179: 210-217. 26. Daugman JG. Two-dimensional spectral analysis of cortical receptive field profiles. Vision Research 1980; 20:847-856. 27. Li KC. Sliced inverse regression for dimension reduction,(with discussions). J Amer Statist Assoc 1991; 86: 316-342. 28. Huang HC, Chen CM, Wang SD. Adaptive ultrasonic speckle reduction based on the slope facet model. Ultrasound in Medicine and Biology, revised, 2003. 29. Chen CM, Lu HHS, Hsiao AT. A dual snake model of high penetrability for ultrasound image boundary extraction. Ultrasound in Medicine and Biology 2001; 27: 1651-1665. 30. Chen CM, Chou YH, Han KC, Hung GS. Computer-aided diagnosis of breast lesions on sonogram using nearly setting-independent features and artificial neural networks. Radiology, to appear, Feb. 2003. 31. Chenyang Xu and Jerry L.Prince ,”Gradient Vector Flow:A new Exerrnal Force for Snakes.” 32. Chenyang Xu,Student Member,IEEE,and Jerry L. Prince,Senior Member,IEEE “Snakes,Shapes, and Gradient Vector Flow” 33. M. Kass ,A. Witkin,and D.Terzopoulos. Snakes: Active contour models. Int. J. Computer Vision. 34. Chang CC and Lin CJ, LIBSVM: a library for support vector machines, 2004. Software available at http://www.csie.ntu.edu.tw/ ~cjlin/libsvm. 35. Hsu CW, Chang CC and Lin CJ, “A Practical Guide to Support Vector Classification,” 2004. Available at http://www.csie.ntu.edu.tw/ ~cjlin/libsvm. 36. 國立台灣大學醫學工程學研究所 許淑雅碩士論文 ﹕肝硬化超音波影像之電腦輔助診斷﹕幾近不受參數設定影響之影像特徵擷取 37. 國立台灣大學醫學工程學研究所 蔡漪琳碩士論文 ﹕以肝臟超音波影像紋理特徵區別肝硬化之研究 | |
dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/31523 | - |
dc.description.abstract | 肝臟疾病在台灣是相當普遍的疾病,每年都在國人十大死因的排行榜居高不下,而肝硬化以及肝癌引起的肝衰竭便是造成死亡的主要原因,但目前針對肝臟疾病的檢測方式以肝硬化來說,通常使用超音波影像當做第一線的檢查工具,然而此種方式的診斷能力取決於醫師的經驗,而象徵肝硬化前奏的肝纖維化檢測目前更是無法以肉眼作為診斷的依據,目前現行的肝纖維化病理檢查大多採用liver biopsy,但是liver biopsy採用的侵入式檢查造成病人心理上的壓力,且手術後需側躺六小時造成病人相當程度的不方便,因此在本研究中提出了一套電腦輔助分類系統以期望在第一線的超音波檢查能得到肝臟纖維化程度的初步資訊,以降低liver biopsy的必要性。
在本研究中,肝纖維化的分類主要是依據兩種臨床用之肝臟超音波特徵:粗糙回音紋理 (coarse echo-texture)以及結狀肝表面 (nodular liver surface)。在粗糙回音紋理部分首先會針對紋理上的顆粒感加以擷取其特徵,這一部分我們稱之為cell-based特徵,包含了以下三組特徵,代表顆粒大小的cell size特徵、代表明顯顆粒數量的cell number特徵、代表顆粒對比強度的cell contrast特徵,除此之外,還有將肝臟質地的區域同質性量化得到的region size特徵。而在增加了分類的數量後為了繼續維持夠高的準確度又開發了兩個新的特徵值包含wavelet decomposition所抽出的D2影像之variance,以及針對肝硬化引起肝表面產生凹凸起伏的現象加以擷取此特徵,經由surface extraction、loess smooth以及residual的計算最後會得到代表肝表面起伏程度的surface lump number特徵。 在不同的機器參數設定之下,影像的特性將會產生不小的差異。然而在相關的電腦輔助診斷研究當中,大多沒有針對機器參數所造成的影響加以考量。本研究為了解決此一問題,提出了參考脾臟紋理的方法,由於機器參數的資訊會同時出現在肝臟以及脾臟的紋理上面,因此上述由紋理所抽取出的特徵值便可以藉由脾臟所計算出來的特徵值將機器以及個人之間潛在的差別消除,和以往只採用肝臟影像的方法比較起來也較具可信度,而與需要在所有影像上固定機器參數的方法比較起來則是方便許多。 本論文的結果分為兩個部份,一部分的實驗資料將其為分兩類( cirrhotic and non-cirrhotic),此一部分的實驗資料包含三種機器分別為Toshiba SSA-380A cirrhotic case 90個 non-cirrhotic case 97個;Toshiba SSA-370A cirrhotic case 34個 non-cirrhotic case 40個;Aloka SSD-4000 cirrhotic case 31個 non-cirrhotic case 30個,此部份採用了logistic regression function當做分類器,準確率在各種不同的評估研究中約為91%至97%,這些評估研究包含同張影像所抓的4個不同ROI對結果所產生的差異、三台不同機器之間所造成的差異、將三種機器影像混合產生機器以及ROI所造成的差異以及同一個病人取得的四張影像個別抓取ROI所產生的影像間差異,將會有許多的探討。另一部分的實驗資料為具有biopsy tissue proof的資料共106個case,在不同的fibrosis stage case 數目上分別有A 12個、B 7個、C 46個、D 20個、E 5個、F 15個和G 1個。本研究將此部份之實驗資料分為三類採用了support vector machine當做分類器並且延用了第一部分所找出的cell contrast、region size特徵值並且加入新的D2 variance以及結狀肝表面所擷取出的surface lump number特徵加入判別。其中嘗試了許多不同的fibrosis stage組合在組合為class1:A,B class2:C,D;class3:E,F,G得到約九成的準確率為最佳組合。 | zh_TW |
dc.description.abstract | Prevailing in Taiwan, liver diseases have been one of the top-ten causes of death for years. In particular, liver cirrhosis and liver cancers have been the major causes of death for the liver diseases. With the advance of the ultrasound imaging technology, sonography has been the first-line check-up tool for diagnosis of liver cirrhosis nowadays. Nevertheless, the diagnosis accuracy of liver cirrhosis is highly dependent on the experiences of medical doctors. It is almost impossible to differentiate the stages of liver fibrosis using sonograms based on visual perception. Currently, liver biopsy is the common approach for the pathological analysis of liver fibrosis. However, liver biopsy is an invasive check-up, which not only makes a patient under a great pressure, but also requires a patient to keep lying down for six hours. To reduce the needs for liver biopsy, this study aims to develop a computer-aided diagnosis (CAD) system to assist stage differentiation of liver fibrosis based on liver sonography.
Classification of liver fibrosis, in this study, is based on two classes of sonographic features: coarseness of echo-texture and nodularity of liver surface. Three categories of mathematical features are extracted to characterize the coarseness of echo-texture. The first category is cell-based features, which attempt to describe the grain-like nature of a liver sonogram. It consists of three pairs of features, namely, cell-size, cell-number and cell-contrast, which quantify the grain size, grain number and grain contrast, respectively. The second category is region-based features, which measure the regional homogeneity. These two categories of features are the basis for differential diagnosis of cirrhotic and non-cirrhotic livers. The third category is multiresolutional texture energy emphasizing the energy of the second level details, called D2 variance. For nodularity of liver surface, a novel feature called surface lump number is extracted through three computational steps: surface extraction, loess smooth and residual estimation. It is well known that the image features would vary with system setting. However, most previous CAD algorithms did not take the effect of system setting into account. To solve this problem, a spleen-reference approach is proposed in this study. Because the effect of system setting may appear in both liver and spleen sonograms, it is anticipated that this effect may be alleviated by simultaneously presenting the sonographic features of liver and spleen sonograms. Therefore, the proposed CAD algorithm is expected to be more robust than the conventional approaches and more convenient than those approaches requiring fixing the system setting. The results of this study are composed of two parts. The first part is to classify the images into two classes, namely, cirrhotic and non-cirrhotic. Three ultrasound imaging systems are used, including Toshiba SSA-380A, Toshiba SSA-370A and Aloka SSD-4000, for which the number of cirrhotic and non-cirrhotic cases are (90, 97), (34, 40), and (31, 30), respectively. The classifier used for this part is logistic regression function. The prediction accuracies range from 91% to 97% for various assessments, including the effects due to four different ROIs, three different ultrasound imaging systems, heterogeneous image properties, and four different scanning views of the same subject. The second part is to classify the images into three classes. The number of cases used in this part is 106, all of which have tissue proof. The numbers of cases for the fibrosis stages A-G are 12, 7, 46, 20, 5, 15, and 1, respectively. The classifier is support vector machine. The features include cell contrast, region size, D2 variance, and surface lump number. While different combinations of the 7 stages of liver fibrosis have been evaluated, we are able to attain the prediction accuracy of approximately 90% for the combination of: class 1-A, B, class 2-C, D, and class 3-E, F, G. | en |
dc.description.provenance | Made available in DSpace on 2021-06-13T03:14:13Z (GMT). No. of bitstreams: 1 ntu-95-R93548053-1.pdf: 1423397 bytes, checksum: ee38bbf10f056ae7d63837900c1126fb (MD5) Previous issue date: 2006 | en |
dc.description.tableofcontents | 第一章 研究背景與動機 1
1.1 研究背景 1 1.2 動機 2 第二章 文獻探討 5 第三章 特徵擷取 8 3.1 選取ROI影像 9 3.2 分兩類所使用之特徵 11 3.2.1 Multi-Scale Gaussian Filter 12 3.2.2 Laplacian Filter 14 3.2.3 Gray Scale Dilation 14 3.2.4 Watershed Transform 15 3.2.5 Cell Contrast 特徵 18 3.2.6 Cell Erosion 19 3.2.6.1 Merge Cells 20 3.2.6.2 Erosion 21 3.2.7 Region Size 數學影像特徵 24 3.3 分三類所使用之特徵 27 3.3.1 D2 variance 27 3.3.2 Surface lump number 29 3.3.2.1 Surface extraction 29 3.3.2.1.1 Preprocess 30 3.3.2.1.2 Dijkstra Algorithm 31 3.3.2.1.3 Snake deformation 33 3.3.2.2 Curve smooth 37 3.3.2.3 Surface lump number 39 第四章 方法的比較 41 4.1 Ogawa的方法 41 4.2 Yeh et. al的方法 43 第五章 特徵值的分類 48 5.1 分類方法 48 5.1.1 Logistic Regression Function 48 5.1.2 Support vector machine 52 5.2 Cross-validation 54 第六章 實驗結果與討論 56 6.1 資料來源 56 6.2 特徵值 57 6.3 分兩類之結果以及比較 58 6.3.1 本研究之結果 59 6.3.2 Ogawa的結果 60 6.3.3 Yeh的結果 62 6.3.4 從Az值統計出的CV值比較各種參數對不同系統的變異程度 64 6.3.5 T-test之效能比較分析 65 6.4 分三類之結果 66 第七章 結論 70 | |
dc.language.iso | zh-TW | |
dc.title | 肝臟超音波影像肝纖維化程度之電腦輔助分類 | zh_TW |
dc.title | Computer-aided Classification of Liver Fibrosis in Ultrasound Images | en |
dc.type | Thesis | |
dc.date.schoolyear | 94-2 | |
dc.description.degree | 碩士 | |
dc.contributor.coadvisor | 楊培銘(Pei-Ming Yang) | |
dc.contributor.oralexamcommittee | 鄭國順(Kuo-Shung Cheng) | |
dc.subject.keyword | 肝纖維化,肝硬化,邏輯回歸函數,支持向量機,結狀肝表面,粗糙回音紋理, | zh_TW |
dc.subject.keyword | Liver Fibrosis,Liver Cirrhosis,Logitic Regression function,Support Vector Machine,Nodular Liver Surface,Coarse Echo Texture, | en |
dc.relation.page | 75 | |
dc.rights.note | 有償授權 | |
dc.date.accepted | 2006-08-04 | |
dc.contributor.author-college | 工學院 | zh_TW |
dc.contributor.author-dept | 醫學工程學研究所 | zh_TW |
顯示於系所單位: | 醫學工程學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-95-1.pdf 目前未授權公開取用 | 1.39 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。