實現於繪圖處理器與場效可編輯邏輯匝陣列之敏捷超音波開發平台

醫用超音波影像系統在臨床醫學上是十分普遍使用的一種診斷工具，在一般超音波系統中，一般是以硬體方式進行超音波成像系統的演算法建置，硬體架構下雖然可以達到較快的成像速度與較少的資源使用，但其硬體描述語言較艱深難以學習，以及所需建置時間較長，偵錯不易，故發展起來較受限制。近年來，由於繪圖處理器平行處理的盛行，部分超音波影像系統也開始採用此架構，與硬體系統相反，GPU雖然在偵錯與發展上較為容易，然而其資源使用以及運算速度上則比硬體系統遜色。本研究主旨有二，一是利用實現兩大平行程式語言：CUDA及OpenCL之超音波影像系統。二是以OpenCL為基準，建立超音波敏捷開發平台。本研究利用ALTERA所提供之OpenCL SDK將程式轉為FPGA的硬體描述語言，建立一個跨平台，能在GPU與FPGA上快速開發的超音波影像系統，除符合程式通用性原則外，還可擷取軟體開發較以且硬體運算速度較快兩者優點。基於上述三種超音波程式設計方法，本研究分別以一維陣列系統及高頻超音波系統為目標系統，成功建構出B-mode、C-mode、PW-mode以及M-mode四種模式，並比較在不同系統架構或程式語言下，記憶體搬移、運算速度等效能評比。結果顯示，在OpenCL與CUDA中，在四種模式都可以達到即時成像標準(30幀/秒)，而在 FPGA 部分,本研究成功以OpenCL 為媒介建構出基於FPGA 之超音波成像系統，但仍有最佳化之工作需要完成以充分發揮本平臺之優勢。

關鍵字

繪圖處理器；平行程式設計； FPGA ；超音波成像系統

並列摘要

Ultrasound imaging is widely used as a diagnostic tool in clinical medicine. In a conventional ultrasound system, imaging algorithms are implemented in hardware. Although the hardware-based system can generally achieve higher processing speed with less hardware resource usage, system development is relatively a long process. Recently, with the advancement of graphics processing units (GPU’s) and parallel programming, software-based ultrasound imaging has been developed. Compared with the hardware-based system, GPU’s have more flexibility in algorithm designs but the GPU’s resource usage and computation speed may not be as good as those of a hardware-based system. To this end, this research has two purposes. First, ultrasound imaging algorithms are implemented in two common GPU programming languages, CUDA and OpenCL. Second, based on OpenCL, an agile ultrasound development platform that exploits the advantages of both the hardware based system and the software-based system is constructed. This research uses OpenCL SDK provided by ALTERA to convert OpenCL code to FPGA based language. In this research, we successfully implement two systems, including a 1D array system and a single element high frequency system, with four major imaging modes, including B-mode, C-mode, PW-mode and M-mode. Performance of the hardware based system was compared against that of the software based system, including memory copy and computation times. Our results show that both CUDA and OpenCL can reach real-time imaging frame rate (i.e., > 30 frame/s). For the FPGA-based system, code conversion from OpenCL is successful and several optimization tasks are necessary to fully exploit the advantages of the agile platform.

並列關鍵字

GPU ； Parallel Programming ； FPGA ； Ultrasound Imaging System

參考文獻

[1] 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.

[4] National Instrument, FPGA Fundamentals, 2012.

[14] J. Bercoff, M. Tanter and M. Fink, "Supersonic shear imaging: a new technique for soft tissue elasticity mapping," IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, vol 51, no. 4, pp. 396-409, 2004.

[15] M. Daga, A. M. Aji and W. Feng, "On the efficacy of a fused CPU+ GPU processor (or APU) for parallel computing," Symposium on Application Accelerators in High-Performance Computing, pp. 141-149, 2011.

[2] T. Y. Lee, “GPU-based frequency domain beamforming for real-time ultrasound 3D imaging,” National Taiwan University, Master Thesis, 2014.

Google Scholar

國際替代計量

實現於繪圖處理器與場效可編輯邏輯匝陣列之敏捷超音波開發平台

全文下載

主題瀏覽