全自動超音波不同掃瞄的腫瘤對位

Tumor Correspondence in Different Views of Automated Breast Ultrasound

周懿升(YI-SHENG JOU)

指導教授 ： 張瑞峰

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2014年)

https://doi.org/10.6342/NTU.2014.10653

摘要

傳統的乳房腫瘤檢測儀器是使用手持式超音波，這樣產生的影像品質相當仰賴操作者的操作。全乳房自動超音波則是一自動化掃瞄乳房區域的儀器，可較不依賴操作者的經驗及使得影像重製成為可能。為了確保掃描範圍涵蓋整個乳房，每個乳房約需要掃描三個方位的影像，而每次掃瞄後所產生的三維立體影像由數百張切片所組成，逐張檢查全部的掃瞄影像是相當費時的。
電腦輔助乳房腫瘤偵測的發展便是自動找出全自動乳房超音波中的可疑區域，讓醫生可以更有效率地找出腫瘤，而本研究是基於電腦輔助乳房腫瘤偵測的結果，從不同方位的掃描影像中，將有重覆掃瞄到的可疑區域做對應，進一步強調出可疑區域的位置可靠性及減少醫生檢查重複區域所花費的時間。
本研究利用了不同方位掃描影像中的可疑區塊間的特徵差異性，做為能否對應的條件，量化特徵包括形態特徵、灰階值、紋理特徵以及位置資訊。實驗中的區域可分為可合併的區域與不可合併的區域，由放射師所定義出的可合併區域有51對，其中25對可合併區塊是腫瘤區塊對，26對是非腫瘤區塊，每對是指某個可疑組織可在二個不同方位的掃描影像中對應到。由實驗結果得知，本研究在整體預測可合併區塊的對應率達到80.39%(41/51)時，錯誤率為5.97%(4/67)。在此結果下，對腫瘤與非腫瘤區塊進行分析，41組正確合併區塊中20組是腫瘤區塊21組是非腫瘤區塊，4組錯誤的皆為非腫瘤區塊。因此，腫瘤的可合併區塊的對應率為80.00%(20/25)錯誤率為0.00%(0/25);非腫瘤可合併區塊的對應率為80.76%(21/26)而錯誤率為9.52%(4/42)。總結，本論文所提出的基於使用量化特徵差異的腫瘤對應方法，可應用於對應電腦輔助乳房腫瘤偵測的結果。

關鍵字

乳癌 ； 電腦輔助乳房腫瘤偵測 ； 全乳房自動超音波 ； 腫瘤對應

並列摘要

Automated breast ultrasound (ABUS) system is developed to automatically scan the whole breast to reduce operator-dependent. Generally, three passes of different orientations are necessary to cover a breast in the scanning. A pass generates an ABUS image volume composed of more than 300 2-D slices. To reduce the review time, computer-aided detection (CADe) systems were proposed to automatically detect breast tumors in individual ABUS pass. This study further analyzed whether the detected regions in a pass are the same regions in other passes. The tumor correspondence algorithm used the criteria of clock, relative distance, and distance to nipple to remove low-likelihood mapping pairs. The discrimination of remaining mapping pairs was performed by quantitative morphology, intensity, texture and location features in a logistic regression model. As a result, the mapping rate could achieve 80.39% (41/51) with error rate of 5.97% (4/67). For tumor regions, the mapping rate was 80.00% (20/25) with the error rate of 0.00% (0/25). For non-tumor regions, the mapping rate was 80.76% (21/26) and the error rate was 9.52% (4/42). In conclusion, the performance of the proposed tumor correspondence algorithm would be helpful to detect the same regions in different passes that can reduce the reviewing time for radiologists.

並列關鍵字

Breast cancer ； computer-aided detection ； automated whole breast ultrasound ； tumor correspondence

參考文獻

[2] E. A. Sickles, R. Filly, and P. Callen, "Benign breast lesions: ultrasound detection and diagnosis," Radiology, vol. 151, pp. 467-470, May 1984.

[3] T. M. Kolb, J. Lichy, and J. H. Newhouse, "Comparison of the performance of screening mammography, physical examination, and breast us and evaluation of factors that influence them: an analysis of 27,825 patient evaluations1," Radiology, vol. 225, pp. 165-175, October 2002.

[4] K. M. Kelly, J. Dean, W. S. Comulada, and S.-J. Lee, "Breast cancer detection using automated whole breast ultrasound and mammography in radiographically dense breasts," European radiology, vol. 20, pp. 734-742, 2010.

[5] E. Wenkel, M. Heckmann, M. Heinrich, S. Schwab, M. Uder, R. Schulz-Wendtland, et al., "Automated Breast Ultrasound: Lesion Detection and BI-RAD Classification-a Pilot Study," Rofo, vol. 180, pp. 804-808, 2008.

[6] Y.-H. Chou, C.-M. Tiu, J. Chen, and R.-F. Chang, "Automated full-field breast ultrasonography: the past and the present," Journal of Medical Ultrasound, vol. 15, pp. 31-44, Aug 2007.