近年來由於磁振造影的技術進步，磁振影像亦可用來量測肺部微灌流。肺部的區域血流供應與氣體供應是否良好匹配決定了氧氣的交換率，因此對比劑微灌流影像是一種有效檢查肺部疾病的方法，病患注射對比劑後，經過MRI等醫療設備觀察對比劑流動的狀況，可以讓醫師知道肺部的血流狀況，是否發生血塊堵塞(肺栓塞)的現象。為了分析肺部組織血流的情況，需要複雜的曲線擬合參數計算，一般都需花上幾十分鐘甚至幾個小時，醫師就不能立即診斷病人的病況。近年來，通用圖形平行運算(GPGPU)逐漸能加速科學運算的技術，並用在演算法能夠被平行處理，能有效加快需要大量平行計算的資料。我們的研究中，提出以通用圖形平行運算單元運用在曲線擬合的運算上，減少整個運算過程的時間。當應用在Levenberg-Marquardt 演算法中，平行運算的計算時間改善到3秒的計算時間。總結，我們提出的通用圖形平行運算有希望去降低曲線擬合的計算時間。

Due to the technical development of the medical image in recent years, MRI is utilized to evaluate pulmonary perfusion. After injection of contrast agent, the washing-in and washing-out of contract agent in tissues is quantified through a dynamic scan. Then, the blood flow analysis of the patient can be determined and provided for the follow-up diagnosis. The quantification analysis of lung tissues is to obtain perfusion parameters by using gamma curve fitting. Pixel-by-pixel curve fitting of perfusion generally takes minutes or hours by MATLAB system. Recently, the parallel computing using general-purpose computation on graphics processing units (GPGPU) shows able to accelerate the scientific computing if the algorithm can be parallelized. In this study, GPGPU parallel computation is proposed to reduce the whole calculation time of gamma-curve fitting by Levenberg-Marquardt algorithm. Applying GPU program on the 7-slice perfusion data set, the parallel algorithm reduced the computation time to ~3 seconds. We conclude that the GPU computing is a promising method to accelerate curve fitting.

Axel, L. (1980). "Cerebral blood flow determination by rapid-sequence computed tomography: theoretical analysis." Radiology 137(3): 679-686.
Flynn, M. (1972). "some computer organizations and their effectiveness." IEEE Trans Comput C-21: 948.
Hansen, M. S., D. Atkinson, et al. (2008). "Cartesian SENSE and k-t SENSE reconstruction using commodity graphics hardware." Magn Reson Med 59(3): 463-468.
Hatabu, H., E. Tadamura, et al. (1999). "Quantitative assessment of pulmonary perfusion with dynamic contrast-enhanced MRI." Magn Reson Med 42(6): 1033-1038.
Knoll, F., M. Unger, et al. (2010). "Fast reduction of undersampling artifacts in radial MR angiography with 3D total variation on graphics hardware." MAGMA 23(2): 103-114.
Levenberg, K. (1944). "A Method for the Solution of Certain Non-Linear Problems in Least Squares." The Quarterly of Applied Mathematics 2: 164–168.
Levin, D. L., Q. Chen, et al. (2001). "Evaluation of regional pulmonary perfusion using ultrafast magnetic resonance imaging." Magn Reson Med 46(1): 166-171.
Marquardt, D. (1963). "An Algorithm for Least-Squares Estimation of Nonlinear Parameters." SIAM Journal on Applied Mathematics 11(2): 431–441.
McGraw, T. and M. Nadar (2007). "Stochastic DT-MRI connectivity mapping on the GPU." IEEE Trans Vis Comput Graph 13(6): 1504-1511.
Meier, P. and K. L. Zierler (1954). "On the theory of the indicator-dilution method for measurement of blood flow and volume." J Appl Physiol 6(12): 731-744.
Ohno, Y., H. Hatabu, et al. (2004). "Quantitative assessment of regional pulmonary perfusion in the entire lung using three-dimensional ultrafast dynamic contrast-enhanced magnetic resonance imaging: Preliminary experience in 40 subjects." J Magn Reson Imaging 20(3): 353-365.
Roujol, S., B. D. de Senneville, et al. (2009). "Online real-time reconstruction of adaptive TSENSE with commodity CPU/GPU hardware." Magn Reson Med 62(6): 1658-1664.
Sorensen, T. S., T. Schaeffter, et al. (2008). "Accelerating the nonequispaced fast Fourier transform on commodity graphics hardware." IEEE Trans Med Imaging 27(4): 538-547.
Joachim Wuttke: lmfit - a C/C++ routine for Levenberg-Marquardt minimization with wrapper for least-squares curve fitting, based on work by B. S. Garbow, K. E. Hillstrom, J. J. Mor&eacute;, and S. Moshier. Version <3.0>, from http://www.messen-und-deuten.de/lmfit/.