B. M. W. Tsui, Z. Zhao, E. C. Frey, and G. T. Gullberg, "Comparison between ML-EM and WLS-CG algorithms for SPECT image reconstruction", IEEE Trans. Nucl. Sci., vol. 38, pp. 1766-1772, 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....are partially supported by NSF Grant DMS 9623077. 2 Partially supported by NSF Grant MIP 9702856. PET images presented in [1, 2] indicate that this method yields algorithms which converge faster and produce images with greater contrast and detail than both the ML EM and data based WLS methods [9]. However, not much is known about the mathematical properties (such as convergence) of these algorithms. The aim of this paper is to develop algorithms for minimizing the same WLS cost function as in [1, 2] which possess better mathematical properties. To do this, we adopt the same basic ....

....ij 0. In the ideal model for PET, the column sums of P are all 1, since p ij represents the conditional probability that an emission in the j th pixel is detected in the i th detector tube. Therefore we assume that P m i=1 p ij = 1; for all j = 1; 2; n. Traditional WLS methods (cf. [6, 9]) require the estimation of the error covariance matrix from the data, which introduces additional errors. We refer to this method as the data based WLS method. The idea in [1, 2] is to use the fact that the variance and mean of Poisson data are equal, thus replacing the data based variance with ....

B. M. W. Tsui, Z. Zhao, E. C. Frey, and G. T. Gullberg, "Comparison between ML-EM and WLS-CG algorithms for SPECT image reconstruction", IEEE Trans. Nucl. Sci., vol. 38, pp. 1766-1772, 1991.

....than filtered backprojection. In particular, the incorporation of the spatially varying or spatially invariant point response reduces noise at resolutions comparable to single pixel 5 ML EM (or filtered backprojection) and offers improved resolution with increased noise at further iterations [5, 44, 45, 46, 47, 48]. 2.7 Summary This background chapter described the nature of some of the most significant artifacts in SPECT imaging attenuation and point response. Building on this background, we will show that the location and type of attenuation artifact depends on the relative 5 Single pixel means ....

B. M. W. Tsui et al., "Comparison between ML-EM and WLS-CG algorithms for SPECT image reconstruction," IEEE Trans. Nucl. Sci., vol. 38, no. 6, pp. 1766--1772, 1991.

.... algorithm is proposed in [31] Increasing changes at each iteration by an overrelaxation parameter is proposed in [25] and described in more generality in [17] Minimizing a least squares function f(x) kn Gamma P T k 2 2 using conjugate gradient algorithm, introduced in [38] is used in [19, 37] and with a preconditioned conjugate gradient algorithm in [18] Parallel implementations have been reported for hypercube iPSC 2 and shared memory multiprocessor BBN Butterfly GP1000 [5] Alliant FX 8 shared memory multiprocessor [13] vector machine Cray 1 [17] special parallel architecture ....

Tsui, B., Zhao, X., Frey, E., and Gullberg, G. Comparison between ML-EM and WLS-CG algorithms for SPECT image reconstruction. IEEE Trans. Nucl. Sci., 38:1766--1772, 1991.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC