透過您的圖書館登入
IP:3.14.80.45
  • 學位論文

實現於繪圖處理器上之超音波即時三維影像頻域成像方法

GPU-Based Frequency Domain Beamforming for Real-Time Ultrasound 3D Imaging

指導教授 : 李百祺
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


醫用超音波成像系統在臨床醫學上是被十分普遍使用的一種診斷工具,在一般的二維影像成像用途中,此系統會先以一維陣列探頭收取射頻資料,接著再利用延遲相加波束形成產生影像資料,最後,在加上適當的影像處理技術後,輸出影像品質清晰,且幀率可達即時成像標準以上的結果。 然而,在這樣的成像系統基礎下,若我們希望更進一步地將成像系統中所使用的一維陣列探頭替換為二維陣列探頭,以此來獲取目標空間中的三維影像時,首先,射頻資料量便會因為二維陣列探頭的使用而呈倍數成長,增加了系統在處理資料時的負擔,再者,更大的問題則是,延遲相加波束形成會因為射頻資料量的增加,以及三維影像所需形成之波束數量增加兩因素,更大幅度地增加其所需要的計算時間,導致整個成像系統無法達到即時成像的目標,有鑒於此,本研究選擇以加速波束形成的方式作為切入點,期望能提供三維超音波即時成像系統一可能的解決方案。 為了能加速波束形成,本研究採用了平面波頻域波束形成與繪圖處理器上之平行程式設計兩項技術,一來利用平面波頻域波束形成最大的特點,只需要一組射頻資料即可產生一組影像資料,而非僅一條波束,來根本上大幅地降低波束形成所需的計算時間,二來,再利用平面波頻域波束形成非常適合以平行方式處理的特色,搭配上繪圖處理器強大的平行運算能力,更進一步地加速波束形成以實現即時成像的目標。 最後,本研究以 Field II 產生實驗所需之模擬射頻資料,並在 Matlab 中進行平面波頻域波束形成與合成影像技術之模擬,透過分析所得影像之 PSF、Side Lobe 與 CNR 來驗證所得影像之正確性與品質,接著,本研究再於搭載了繪圖處理器的個人電腦上,以 CUDA 語言實現平面波頻域波束形成之平行程式設計,並以 Nsight 分析其運算時間,根據實驗結果,在 64 x 64 通道 x 4096 取樣點之二維陣列探頭模擬環境下,等效幀率能達到每秒 30 幀之即時成像標準,此即證明了本研究結果確實可作為三維超音波即時成像系統之可能解決方案。

並列摘要


Medical ultrasound imaging system is generally used as a diagnosis tool in clinical medicine. For general 2D imaging purpose, the system uses 1D array transducer to transmit/receive RF data, do the delay and sum beamforming (DAS), add some adequate imaging processing techniques, and finally, output the good quality image result in real-time. However, if 1D array transducer is replaced by 2D array for 3D imaging purpose, large RF data size will make the DAS too slow to keep the system real-time. Therefore, this research combines two techniques to speed up beamforming, that is, plane wave frequency domain beamforming (PWFDBF) and parallel programming on GPU. First, PWFDBF can use just one set of RF data to beamform one frame image. This feature will reduce the complexity and processing time of beamforming funamentally and enormously. Second, the powerful parallel processing ability of GPU will make PWFDBF ever faster. Field II is used in this research for creating simulated RF data. Then, Matlab is used for simulating PWFDBF and compounding imaging technique, verifying the correctness of the image result and doing the analytis of image quality. Finally, parallel programming PWFDBF is implemented on the PC with GPU by CUDA programming language. And Nsight is used for speed analysis. According to the experiment result, equivalent frame rate of PWFDBF can reach to the real-time standard(30 frames/s) under the simulated environment which is a 2D array transducer with 64 x 64 channels x 4096 samples. The result proves that the research is indeed a possible solution for 3D ultrasound imaging system according to the final processing time analysis result.

參考文獻


[4] Y. F. Li and P. C. Li, “Software beamforming: Comparison between a phased array and synthetic transmit aperture,” Ultrasonic imaging, vol. 33, pp. 109–118, 2011.
[5] M. O’Donnell, “Efficient parallel receive beam forming for phased array imaging using phase rotation,” in Proc. IEEE Ultrason. Symp., vol. 3, 1990, pp. 1495–1498.
[6] M. L. Li and P. C. Li, “Improved fourier-transform-based parallel receive beam for- mation,” Ultrasonic imaging, vol. 25, pp. 73–84, 2003.
[7] J. Lu, “2d and 3d high frame rate imaging with limited diffraction beams,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 44, no. 4, pp. 839–856, 1997.
[9] P.Kruizinga,F.Mastik,N.deJong,A.vanderSteen,andG.vanSoest,“Plane-wave ultrasound beamforming using a nonuniform fast fourier transform,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 59, no. 12, 2012.

延伸閱讀