利用監督式區域平均法提升超音波B-Mode影像品質之研究

The Study of Enhanced Ultrasound B-Mode Images Quality by Applying Supervised Local Means Method

莊慧怡(Hui-Yi Chuang)

指導教授 ： 黃詠暉

共同指導教授 ： 陳泰賓(Tai-Been Chen)

義守大學/電機資訊學院/資訊工程學系/碩士(2015年)

https://doi.org/10.6343/ISU.2015.00131

摘要

超音波掃描儀具有即時性、非侵入性、方便性以及無輻射性等優點；目前仍是臨床第一線經常使用之造影儀器之一。然而影像品質常取決探頭老化以及雜訊干擾程度。因此讓醫學影像品質更好，提供優良的診斷價值是重要課題之一。
本研究採用監督式區域平均法進行影像品質提升。利用腹部造影假體進行取像，由GE E9型音波造影儀之取像為監督者；而GE L9及Sonosite 180+取像為測試組。造影實驗因子包括2種深度(12和20公分)以及10個取像位置。開發5種線性與非線性區域平均演算法提升測試組影像品質。評估演算法處理前後影像對比度與訊雜比，透過相關性及迴歸分析，判定影像品質提升率。
分析結果顯示，測試組影像品質提升以線性模型為佳，其中中間提升值演算法對測試組影像之對比度提升率達14.53%；而線性轉換與中間值提升轉換之加權平均演算法對對測試組影像之訊雜比的提升率達1.19%；反之非線性模型的演算法對影像品質提升皆無明顯改善。從本實驗得知，線性模型對提升影像品質的對比度及訊雜比有相當大的幫助。

關鍵字

影像品質 ； 監督式區域平均法 ； 訊雜比 ； 對比度

並列摘要

The merits of ultrasound tomography are real-time, none invasive, convenience, and none-radiation. Therefore, the ultrasound tomography is the one of used to image in clinical. However, the quality of images is often dominated by exhaustion of probe and scattering noise. Hence, it is an important issue that improves quality of images and provides accurate diagnostic judgment.
In this study, the quality of images was improved by supervised local-means algorithms. The supervised images were obtained by scanner of GE E9 on abdominal phantom. The testing images were provided by scanners of GE L9 and Sonosite 180+. The experimental factors included two imaging depths and ten imaging locations. Five linear and none-linear local-mean methods were designed to enhance quality of testing images. The improvement rates of image quality were involved contrast (CR) and signal-to-noise (SNR) ratios estimated by correlation coefficient and regression analysis for presented methods.
The analytical results were shown that linear local-mean method could generate better quality of testing images than other methods. Especially, the CR of testing images by linear local-mean method of Median Correction was prompted to 14.53%. The SNR of testing images by the weighted between Median Correction and linear methods was prompted to 1.19%. However, the rates of promotion were not significantly by none-linear local-mean method for testing images. The linear local-mean approach could be used to enhance quality of testing images via evaluated CR and SNR in this study.

並列關鍵字

Quality of Image ； Supervised Local-Means Method ； SNR ； CR

參考文獻

1. Eunseop Yeom et.al. Improvement of ultrasound speckle image velocimetry using image enhancement techniques. Ultrasonics 54 205-216, 2014 .

2. Hojong Choi , K. Kirk Shung. Novel power MOSFET-based expander for high frequency ultrasound systems. Ultrasonics 54 121-130, 2014 .

3. Hongliang Ren et.al. Tubular Structure Enhancement for Surgical Instrument Detection in 3D Ultrasound. 33rd Annual International Conference of the IEEE EMBS Boston, Massachusetts USA, August 30 - September 3, 2011.

5. G. Vegas-S´anchez-Ferrero, Realistic Log-compressed Law for ultrasound image recovery. 2011 18th IEEE International Conference on Image Processing. Annual Conference , 2029-2032, 2011.

6. J. Seabra and J. Sanches, On estimating de-speckled and speckle components from b-mode ultrasound images, Rotterdam (NED), ISBI’10 , 284-287, 2011.