臺灣博碩士論文加值系統

傳統乳房疾病的超音波影像診斷，不但仰賴醫師的專業經驗，更受到當時診斷環境、個人主觀意識等因素的影響。為了增加診斷的正確性及降低人為因素，我們致力於乳房腫瘤之電腦輔助診斷系統研究與開發，其主要目的是為了希望能夠早期發現乳癌病灶以及減少不必要組織切片檢查個數。雖然過去傳統的電腦輔助診斷系統已經達到了合理的成效，但是它們通常使用著不可靠的特徵。更確切的說，他們所建議的影像特徵非常敏感於系統參數的設定。此外，它們也並未善用在不確定病例中可提供決定性判斷的不顯著特徵。本研究之不同於過去電腦輔助診斷乳房腫瘤之超音波影像的研究，主要在於可以不必限定超音波的機型和設定，同時藉由仿照臨床醫師的診斷模式，我們也提出了二層分類結構的電腦輔助診斷系統。此系統，除了在臨床上可以提供醫師對乳房腫瘤之建議性的診斷，也可以做為將來開發教學系統以協助較無經驗的醫師學習分辨良惡性乳房腫瘤。為了克服傳統電腦輔助系統的問題，我們提出了一套有效的幾何性特徵和二層的新分類結構，並考慮那些不顯著特徵的診斷價值。在第一層部份是經由logstic discriminant function 選取顯著特徵，再由multi-layer feed-forward neural network進行分類。在第二層部份是以constrained maximization problem的方法，充份地利用那些其餘不顯著特徵。在檢驗幾何性特徵能力方面，我們以第一組160例病灶資料為訓練組，第二組111例可疑病灶進行測試，其表現能力accuracy、sensitivity、specificity、positive prediction vlaue 和negative prediction value 分別為94.6％、97.2％、90.0％、94.5 ％、和94.7 ％。整體而言，我們所提出二層分類結構電腦輔助診斷系統於160例可疑病灶，其表現能力accuracy、sensitivity、specificity、positive prediction vlaue 和negative prediction value 分別為93.8％、91.3 ％、95.6 ％、93.5 ％、和94.0 ％。藉由考慮不顯著特徵，我們將分類的accuracy、sensitivity和specificity各提高了3.2 ％、4.3 ％ , 和2.2 ％。最後，我們請兩位放射科醫師就同樣160例乳房超音波影像，進行專業診斷區別良惡性腫瘤。可以很清楚地看出本研究所提出的演算法，在ppv和npv方面都較優於兩位臨床醫師10％。即表示我們提出的電腦輔助診斷系統，將至少可以減少10％不必要組織切片檢查的個案數。

To early detect the malignant breast lesions and to reduce the number of unnecessary biopsies, computer-aided diagnosis (CAD) of sonographic breast lesions have been studied extensively. Although reasonable performances have been achieved, the conventional CAD algorithms generally suffer the unreliable features and the inability to utilize the insignificant features, which may be decisive in determining the equivocal cases for the significant features. To overcome these two problems, in this paper, we proposed a set of highly effective geometrical features and a new two-level classification scheme, which took into account the value of the insignificant features. The first-level classification was composed of a logistic discrimination function for selecting the significant features and a multi-layer feed-forward neural network for classification. The second-level classification has been formulated as a constrained maximization problem designed to fully utilize the insignificant features. The experimental results show that the best classification accuracy, sensitivity, specificity, positive prediction value and negative prediction value for the proposed two-level classification scheme with 160 lesions were as high as 93.8%, 91.3%, 95.6%, 93.5%, and 94.0%, respectively. By considering the insignificant features, we have succeeded in increasing the classification accuracy, the sensitivity and the specificity by 3.2%, 4.3%, and 2.2%, respectively.

誌謝……………………………………………………………………I
中文摘要………………………………………………………………II
英文摘要………………………………………………………………IV
目錄……………………………………………………………………V
圖表目錄………………………………………………………………VII
第一章緒論……………………………………………………………1
1.1 前言…………………………………………………………….1
1.2乳房之簡介……………………………………………………..2
1.2.1乳房的解剖及生理…………………………………………2
1.2.2乳房之常見腫瘤……………………………………………3
1.3 乳房超音波影像之電腦輔助診斷研究回顧………………….4
1.4 研究動機與目的……………………………………………….6
1.5 論文架構……………………………………………………….7
第二章材料與方法……………………………………………………8
2.1 特徵擷取……………………………………………………….9
2.1.1 幾何性特徵……………………………………………….9
2.1.2 區域性特徵……………………………………………….23
2.2 特徵選取與分類……………………………………………….29
2.2.1 第一層之特徵選取……………………………………….31
2.2.2 第一層之分類法………………………………………….33
2.2.3 第二層之特徵選取與分類法…………………………….36
2.2.4 特徵選取與分類之摘要………………………………….41
2.3 過去文獻之分析方法………………………………………….43
第三章實驗結果與能力分析…………………………………………46
3.1本研究提出的電腦輔助演算法的能力分析…………………..47
3.2能力比較………………………………………………………..54
第四章結論……………………………………………………………59
參考文獻………………………………………………………………...62

〔1〕Jackson VP, “The role of US in breast imaging,” Radiology, Vol. 177, 1990, pp. 305-311.
〔2〕F. Winsberg et al., “Detection of radiographic abnormalities in Mammograms by means of optical scanning and computer analysis,” Radiology, Vol. 89, 1967, pp. 211-215.
〔3〕J. Roehrig et al., “Clinical results with R2 imagechecker system,” Digital Mammography, N.Karseemeijer et al., eds., Kluwer Acdemic, Dordrecht, The Netherlands, 1998, pp.395-400.
〔4〕Stavros AT et al., “Solid breast nodules : use of sonography to distinguish between benign and malignant lesions.” Radiology, Vol. 196, 1995, pp. 123-134.
〔5〕V. Goldberg et al., “Texture analysis of ultrasound B-scans to aid diagnosis of cancerous lesions in the breast.” 1988 Ultrasonics symposium, 1988, pp. 839-842.
〔6〕V. Goldberg et al., “Improving the distinction between benign and malignant breast lesions: the value of sonographic texture analysis.” Ultrasonic Imaging, vol.19, 1992, 1475-1481.
〔7〕BS Garra et al., “Improving the distinction between benign and malignant breast lesion: the value of sonographic texture analysis.” Ultrasonic Imaging, vol. 15, 1993, pp. 267-285.
〔8〕Y Zheng et al., “Reduction of breast biopsies with a modified self-organizing map.” IEEE Trans Neural Networks, vol. 8, 1997, pp. 1386-1396.
〔9〕C Ruggiero et al., “Automatic recognition of malignant lesions in ultrasound images by artificial neural networks.” Proceedings of the 20th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, vol. 20, 1998, pp. 872-875.
〔10〕ML Giger et al., “Computerized analysis of lesion in US images of the breast.” Acad Radiol, vol. 6, 1999, pp. 665-674.
〔11〕Y Joseph et al., “Application of artificial neural networks for diagnosis of breast cancer.” IEEE, 1999, pp. 1755-1759.
〔12〕RD Chen et al., “Computer-aided diagnosis applied to US of solid breast nodules by using neural networks.” Radiology, vol 213, 1999, pp. 407-412.
〔13〕RD Chen et al., “Breast cancer diagnosis using self-organizing map for sonography.” Ultrasound in Med. & Bio., vol. 206, 2000, pp. 405-411.
〔14〕RF Chang et al., “computer-aided diagnosis for surgical office-based breast ultrasound.” ARCH SURG, vol. 135, 2000, pp. 696-699.
〔15〕DR Chen et al., “Texture analysis of breast tumors on sonograms.” Seminars in Ultrasound, CT, and MRI, vol. 21, 2000, pp. 308-316.
〔16〕洪國賢，“腫瘤形狀特徵多變數分析技術應用於超音波乳癌診斷研究”，陽明大學，2000。
〔17〕P Skaane et al., “Analysis of sonographic features in the differentiation of fibroadenoma and invasive ductal carcinoma.” AJR, vol. 170, 1998, pp. 109-114.
〔18〕DH Foley, “Considerations of sample and feature size.” IEEE Trans. Inform. Theory, vol. 18, 1972, pp. 618-626.
〔19〕K Fukunaga, “Effect of sample size in classifier design.” IEEE Trans. PAMI, vol. 11, 1989, pp. 873-885.
〔20〕GJ Mclachlan, Discrminant analysis and statistical pattern recognition, John Wiley & Sons, Inc., 1992, New York.
〔21〕SJ Raudys et al., “Small sample size effects in statistical pattern recognition.” IEEE Trans. PAMI, vol. 13, 1991, pp. 252-264.
〔22〕LM F, “Analysis of the dimensionality of neural networks for pattern recognition.” Patt. Recog., vol. 23, 1990, pp. 1131-1140.
〔23〕V Dutt et al., “Adaptive speckle reduction filter for log-compressed B-scan images.” IEEE Trans. Med. Imaging, vol. 15,1996, pp. 802-813.
〔24〕V. Cherkassky et al., Learning from data: concepts, theory, and methods. John Wiley & Sons, inc., 1998, New York, U.S.A.
〔25〕WR Dillon et al., Multivariate analysis : method and applications. John Wiley & Sons, inc., 1984, New York, U.S.A.
〔26〕DC He et al., “Texture discrimination based on an optimal utilization of texture features.” Pattern Recognition, vol. 21, 1989, pp. 141-146.
〔27〕JM Zurada et al., Introduction to artificial neural systems, PWS Publishing Company, 1992, Boston, U.S.A.
〔28〕RJ Prokop, The technique of standard moments for global feature object representation, Master Thesis, Cornell University, 1990, Ithaca, New York, U.S.A.
〔29〕RM Haralick et al., Computer and robot vision, Addison-Wesley Publishing Company, Inc., 1993, New York, U.S.A.
〔30〕W Leucht, Teching atlas of breast ultrasound, Georg Thieme Verlag, 1992, New York, U.S.A.
〔31〕張金堅，“乳房醫學”，民73，台北健康世界雜誌社。
〔32〕陳啟明等人，“乳房疾病Breast diseases”，民80，台北九州。