针对漫射光学成像(DOT)选择不同的逆问题求解算法将直接影响成像质量的问题,对4种常用重建算法的重建效果进行了比较研究,采用双异质模型,利用Rytov方法对扩散近似方程进行近似,假设散射系数是已知常量,重点关注吸收系数图像的重建.通过使用Tikhonov正则化、截断奇异值分解(TSVD)、代数重建算法(ART)和同时迭代重建技术(SIRT)4种常用重建算法,对不同条件下的双异质图像的逆问题进行求解,研究不同算法重建后图像的空间分辨率、抗噪声能力和成像速度.研究结果表明:低噪声条件下,Tikhonov正则化算法重建图像的空间分辨率最好；当信噪比小于20 dB时,TSVD算法具有最佳的抗噪声性能；当源-探测器结构不变,成像的感兴趣区域的尺寸和体素划分固定时,Tikhonov正则化算法的速度最快.总体上讲,子空间技术优于代数重建技术,研究结果对不同成像条件下选择合适的重建算法具有一定的指导意义.%Different reconstruction algorithms in diffuse optical tomography may result in different image qualities due to the nature of inverse problem that is underestimated and ill-conditioned. A dual-heterogeneity model is presented and the Rytov solution to the diffusion approximation equation is employed, where the scattered field is assumed to be slowly variable in space. Furthermore, the scattering coefficient fis is regarded as a known spatially constant, and thus, the reconstruction of the absorption coefficient jia is solely focused on. Data simulated from the dual-heterogeneity model at various conditions are used to investigate spatial resolution, time consumption for imaging and noise sensitivity of four common reconstruction techniques respectively. The comparative research indicates that Tikhonov method is superior to the other three methods in the performance of spatial resolution when the noise gets relatively low. The truncated singular value decomposition (TSVD) is the most insensitive to noise as the signal to noise ratio is less than 20 dB. When the positions of sources and detectors, the region of interest, and the volume of each voxel are fixed,Tikhonov method takes the least time to image. Generally, the performance of the subspace techniques is better than the one of the algebraic techniques in the aspects mentioned above.
展开▼
