橫波乳房彈性攝影之腫瘤診斷

Tumor Diagnosis of Shear Wave Breast Elastography

林怡婷(Yi-Ting Lin)

指導教授 ： 張瑞峰

國立臺灣大學/電機資訊學院/資訊網路與多媒體研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.02056

摘要

乳癌一直是全球女性的十大死因之一，而腫瘤的硬度也已經被證實為分辨良性與惡性腫瘤的主要特徵。過去幾年來，醫師普遍使用乳房彈性超音波來評估病患的腫瘤彈性硬度。這項技術需人為壓迫乳房腫瘤來求得腫瘤彈性硬度。不同於傳統的乳房彈性超音波，本次實驗使用的橫波乳房彈性超音波只需利用聲波輻射便可取得腫瘤彈性硬度。在傳統的乳房彈性超音波上，腫瘤診斷是基於腫瘤內部的彈性資訊，而在橫波乳房彈性超音波上，重要的診斷資訊卻是來自於腫瘤外部而非內部的彈性硬度。此篇論文的目的為針對影像做自動切割輪廓並擷取出特徵來診斷腫瘤良惡性。首先，我們會藉由Level set切割方法自動地切割出腫瘤的輪廓，比起利用醫生的手動圈選腫瘤更能維持切割結果的一致性。接著，藉由腫瘤輪廓與影像資訊來擷取出B-mode與彈性特徵。最後，除了利用B-mode與彈性特徵分別來診斷腫瘤良惡性，也結合兩者來加以診斷腫瘤。本實驗中由112個經過病理驗證的病例進行測試，其中包含個58良性與54個惡性的病例。經由實驗結果，當使用B-mode特徵時，腫瘤分辨的準確度為84.82%；當使用彈性特徵時，腫瘤分辨的準確度為91.07%；當結合B-mode與彈性特徵時，腫瘤分辨的準確度為94.64%。根據實驗結果的統計分析，將B-mode與彈性特徵結合時，腫瘤分辨的準確度會有顯著的提升。

關鍵字

彈性 ； 橫波 ； 乳房 ； 腫瘤 ； 腫瘤切割

並列摘要

The breast cancer is always one of the ten leading death causes for women around the world. The strain of the tumor has been confirmed to be the main feature of distinguishing benign and malignant tumors. In the past years, the physician has used the sonoelastography with manual compression to obtain the tumor strain. Different from the conventional sonoelastography, this study adopts the new shear wave elastography which uses the acoustic radiation to generate the tumor strain. In the conventional sonoelastography, the tumor diagnosis is based on the elasticity information inside the tumor. However, in the new shear wave elastography, the important diagnostic information is outside the tumor rather than inside the tumor. The purposes of this paper are automatically segmenting the tumor contour for the image and extracting the features to diagnose benign and malignant tumors. First, we use the level set segmentation method to automatically cut out the tumor contour. Comparing with the manually circled tumor, our scheme can maintain the consistency of the segmentation results. Then, the tumor contour and image information are applied to extract the B-mode and elastographic features. Finally, in addition to use either B-mode or elastographic features to diagnose benign and malignant tumors, a combination of both feature set is also utilized for diagnosis. In this study, we use 112 biopsy-proved breast tumors composed of 58 benign and 54 malignant cases. The experimental results illustrate that the accuracy in distinguishing tumors using B-mode features is 84.82%, whereas 91.07% using elastography features, and 94.64% combining B-mode and elastographic features. Based on statistical analyses of experimental results, the accuracy of classifying tumors using the combined feature set is significantly improved.

並列關鍵字

elastography ； shear wave ； breast ； tumor ； tumor segmentation

參考文獻

[1] D. M. Regner, et al., "Breast lesions: Evaluation with US strain imaging - Clinical experience of multiple observers," Radiology, vol. 238, pp. 425-437, Feb 2006.

[3] C. C. Bhatia KS, Tong CS, Yuen EH, Ahuja AT., "Shear Wave Elasticity Imaging of Cervical Lymph Nodes," Ultrasound in Medicine & Biology, vol. 38, pp. 195-201, 2012.

[4] A. Itoh, et al., "Breast disease: Clinical application of US elastography for diagnosis," Radiology, vol. 239, pp. 341-350, May 2006.

[5] W. K. Moon, et al., "Analysis of elastographic and B-mode features at sonoelastography for breast tumor classification," Ultrasound in medicine & biology, vol. 35, pp. 1794-1802, 2009.

[6] P. W. Andrew Evans, Kim Thomson, Denis McLean, Katrin Brauer, Colin Purdie, Lee Jordan, Lee Baker, Alastair Thompson, "Quantitative shear wave ultrasound elastography: initial experience in solid breast masses," Breast Cancer Research, vol. R 104, 2010.