以模型為基礎之乳房X光攝影微鈣化群立體定位及重建

Model-Based Three-Dimensional Reconstruction of Clustered Microcalcifications from Two Mammographic Views

黃秉洋(Bing-Yang Huang)

指導教授 ： 郭文嘉

元智大學/資訊學院/資訊管理學系/碩士(2007年)

摘要

乳房X光攝影是截至目前為止被認為早期偵測乳癌方法中，考量明確度及敏感度與成本因素效益評估上最為可行的方法。近年來在外科醫學的研究結果中顯示，乳房攝影中微鈣化群聚的外型輪廓與空間分佈的情形是作為惡性腫瘤診斷的重要指標依據，然而由於攝影時投影方式及乳房受擠壓的影響，在所得的乳房攝影影像中微鈣化群聚在三維立體空間分佈的關係經常是混淆不明確的，因此近期較先進乳房攝影技術發展的趨勢已朝向結合多個相位的乳房攝影結果作為三維立體重建的參考基礎。
本論文主要提出一個以模型為基礎之乳房X光攝影微鈣化群立體定位及重建的方法。利用乳房攝影中兩個不同相位的影像(頭腳相及斜側相影像)完成定位及重建的工作，同時發展由二維影像重建三維乳房表示方法及以參數為基礎的受擠壓變形乳房模型。我們首先以等高集合法切割出乳房的輪廓，並以此輪廓透過所提乳房模型作三維立體重建，進而利用遺傳演算法求得頭腳像及斜側像重建比對時所需的最佳參數，完成三維立體模型重建及微鈣化群的立體位置定位。本論文除了可協助醫師重建微鈣化群在乳房中三維空間的資訊，提昇診斷的準確性外，同時重建出病人的乳房立體模型，作為乳房重建之臨床醫學上術前評估重建乳房之體積、形狀及位置等資訊，提供術前模擬實體評估之解決方案，有效減少醫療上資源的浪費與整體效率的提昇。

關鍵字

三維立體重建 ； 微鈣化群聚 ； 模型基礎 ； 乳房X光攝影 ； 等高演算法 ； 基因演算法

並列摘要

Mammography (breast X-ray) is considered currently by far the best trade-off between specificity/sensitivity and cost for the detection of breast cancer in its early stages. It has been revealed that the shape of cluster and the spatial distribution of individual microcalcifications within it are important indicators of its malignancy in recent clinical researches. However, the three-dimensional (3-D) distribution is confounded in mammographic images of clustered microcalcifications with image projection and breast compression. Therefore, recent advances in mammographic technologies have led to new approaches to generate and combine multiple views of the breast for 3-D disease reconstruction.
In this project, we proposed a model-based method for reconstructing 3-D microcalcification clusters from two mammographic views (Cranio-Caudal, CC view and Medio-Lateral Oblique, MLO view). We develop a 3-D breast representation and a parameterised breast compression model reconstructed from two-dimensional images. The level-set method is first used to segment and identify the contour of breast region. The contour is then used to reconstruct the 3-D breast representation through the proposed parameterized breast model. Furthermore, generic algorithm is used to estimate the most optimal parameters for matching the two breast representation volumes reconstructed from two different mammographic views respectively. Corresponding calcifications in the two views are then matched using both the geometric constraint and the matching criterion to build the 3-D reconstructed clusters. The purpose of this project is to make a full size solid breast model to assist reconstruction of 3-D breast information and microcalcification cluster localization. Doctors can also use this solid model for process of evaluation and planning before the surgeries to promote the efficiency of diagnosis, medical treatment qualities and reduce the waste of medical resource.

並列關鍵字

3-D reconstruction ； microcalcification clusters ； model-based ； X-ray mammography ； level-set method ； generic algorithm

參考文獻

[1] L. W. Bassett, “Mammographic analysis of calcifications,” Radio Clin. North Am., 1992, no. 30, pp. 93-105.

[2] M. Lanyi, Diagnosis and Differential Diagnosis of Breast Calcifications. Berlin, Germany: Springer-Verlag, 1986.

[3] J. N. Wolfe, Xeroradiography of the Breast. Springfield, IL: Thomas, 1982.

[4] H. P. Chan, B. Sahiner, K. L. Lam, N. Petrick, M. A. Helvic, M. M. Goodsitt, and D. D. Adler, “Computerized analysis of mammographic microcalcifications in morphological and texture feature spaces,” Med. Phys., vol. 25, pp. 2007–2019, 1998.

[5] Y. Jiang, R. M. Nishikawa, D. E.Wolverton, C. E. Metz, M. L. Giger, R. A. Schmidt, C. J. Vyborny, and K. Doi, “Malignant and benign clustered microcalcifications: automated feature analysis and classification,” Radiology, vol. 198, pp. 671–678, 1996.