超音波鈣化點對比增強成像技術

Calcification Contrast Enhanced Ultrasound Imaging Technique

林玉茹(Yu-Ju Lin)

指導教授 ： 李夢麟

國立清華大學/電機資訊學院/電機工程學系/碩士(2013年)

摘要

本研究中提出三項影像處理技術，分別利用鈣化點於超音波訊號及影像上的特徵來增強鈣化點在超音波影像上的對比。首先從訊號上的觀點來看，基於鈣化點與正常組織於頻域上的差異特徵，分別在兩個中心頻率下進行造影，第一種成像技術為利用帶通濾波器之對比成像方法，對時域訊號處理並成像，以及第二種利用短時距傅立葉轉換之對比成像方法，對頻域訊號處理並成像。從影像上的觀點，計算斑點雜訊的變異數與平均值的比值，提出第三種依據斑點雜訊因子之成像方法，計算斑點雜訊的變異數與平均值的比值，來從影像中偵測鈣化點並加強對比。
在模擬中驗證此三種技術之可行性，以及比較不同參數之影響及成像結果。第一項方法中，帶通濾波器需探討最佳頻率組合及頻寬的影響，但因對比度不夠，幾乎所有情況皆難以判斷鈣化點的位置；第二項方法，也需探討最佳頻率組合及平移視窗大小之影響，鈣化點在其中三組頻率組合有較佳的對比，但並沒有大於原始影像，對比度仍待改善。第三項方法，需探討平移視窗大小之影響，得到的斑點雜訊因子之影像結果對比最好，能夠準確的偵測鈣化點且影響品質極佳，但有其他非鈣化點的亮點也會被增強，有誤判的疑慮。
因此三項成像技術都有要改善的空間，若能解決訊號比對上的偏差問題，有機會可以增強帶通濾波器成像方法的對比。短時距傅立葉轉換成像方法的對比也需加強，若能與斑點雜訊因子之方法結合，有機會在增強對比的同時，並驗證影像中出現的亮點，有哪些真實為鈣化點，而非其他強散射之組織，藉以增強判讀的準確性。

關鍵字

鈣化點 ； 乳癌 ； 超音波

並列摘要

Calcifications are important indicators of breast cancer. Since there is speckle noise in ultrasound images, it is difficult to detect calcifications from breast gland and fat. In this study, we propose three calcification contrast enhanced ultrasound imaging techniques to improve the contrast of calcifications to other tissues in ultrasound images. From signal view, we utilized the signal characteristics of calcifications in frequency domain, since there is more enormous variation of back scattering magnitude of calcifications than that of other tissues in breast in signal spectrum. We used band pass filter (BPF, method 1) and short-time Fouier transform (STFT, method 2) to enhance contrast of calcifications. From image view, we used speckle factor (SF) imaging to enhance the calcification contrast (method 3).
Simulations were performed to verify our idea. We used the Faran model to predict the complex amplitude of echoes from calcifications. We compare amplitudes of calcifications and other tissues between different center frequencies, which represents the contrast. Several frequency are chosen then to enhance the contrast of calcifications to other tissues. The selected frequency sets are employed to design band pass filter and apply short-time Fourier transform for contrast enhanced imaging. In addition, different bandwidth, sliding window size and the size of calcifications are simulated to optimize the contrast.
Furthermore, calcifications manifest as isolated strong scatterers in the scattering volume, so the variance become large when we calculate the variance of intensity in a small window. Speckle factor which is defined as the standard deviation divided by the mean of each window is used to form images where the contrast of calcifications is enhanced. The calcifications can be identified via the binary images obtained from the speckle factor images.
The best results of the three methods are obtained by speckle factor images. Since the binary images are derived from the speckle factor, the background is clear and the image contrast is the best and all the calcifications of different sizes could be detected precisely. There are more problems to solve in the other two methods. If we modify the signal mapping when using band pass filter and try to enhance the contrast in the method using short-time Fourier transform, the image quality and contrast of calcifications would be improved.

參考文獻

[1]行政院衛生署, 99 年死因統計分析, pp. 13–17, 2010

[9]蔡明芳, 鍾瑞容, 莊茂德, “乳房檢查報告的判讀─簡介BI-RADS,” 基層醫學. vol. 23, pp. 384–386, 2008

[3] M. P. Morgan, M. M. Cooke, G. M. McCarthy, “ Microcalcifications Associated with Breast Cancer : An Epiphenomenon or Biologically Significant Feature of Selected Tumors ?,” J. Mammary Gland. Biol. Neoplasia. vol. 10, pp. 181–187 , 2005

[5] E. A. Sickles, “Mammography features of ‘early’ breast cancer,” American Journal of Roentgenology, vol. 143, pp. 461-464, 1984

[6] A. Fandos-Morera, M. Prats-Esteve, J. Tura-Soteras, and A. Traveria-Cros, “Breast tumors: Composition of microcalcifications,” Radiology, vol. 169, pp. 325-327, 1988