Bruyant P P, Sau J, Mallet J J
Nuclear Spectroscopy and Image Processing Research Group, Biophysics Laboratory, Claude Bernard University, Lyon, France.
J Nucl Med. 2000 Nov;41(11):1913-9.
Because of the limited number of projections, the mathematic reconstruction formula of the filtered backprojection (FBP) algorithm may create an artifact that streaks reconstructed images. This artifact can be imperfectly removed by replacing the ramp filter of the FBP with an ad hoc low-pass filter, the cost being the loss of contrast and definition. In this study, a solution was proposed to increase, by computational means, the number of projections to reduce the artifact at a lower cost. The cost was a postacquisition process, which was reasonably time consuming.
The process was called interpolation of projections by contouring (IPC). First, level lines were plotted on the sinogram to delimit isocount regions; then, the regions containing the interpolated points were found, and to each point was assigned the intensity of its isocount region. Using this process, the data could be resampled, allowing an increase in the number of projections or the number of pixels by projections. A phantom study of bone scintigraphy was performed to compare the slices obtained with and without the IPC process with the true image. A clinical case was also presented.
The phantom study showed that with the IPC process, the reconstructed slice was closer to the model, inside and outside the body, when the sinogram was resampled to multiply by 2 or 3 the number of projections, with the same number of pixels per projection. In the clinical study, the streak artifact was reduced, especially outside the body, although only a ramp filter was used.
The IPC process succeeded in reducing the streak artifact. This process did not require any modification in acquisition and was not operator dependent. The increase in the number of projections is likely a necessary but not a sufficient condition to reduce the streak artifact: if not corrected, the attenuation could be a limiting factor in the removal of this artifact when the number of projections increases.
由于投影数量有限,滤波反投影(FBP)算法的数学重建公式可能会产生一种伪影,使重建图像出现条纹。通过用特设的低通滤波器代替FBP的斜坡滤波器,可以不完全消除这种伪影,代价是对比度和清晰度的损失。在本研究中,提出了一种通过计算手段增加投影数量以降低成本来减少伪影的解决方案。成本是一个采集后处理过程,相当耗时。
该过程称为轮廓投影插值(IPC)。首先,在正弦图上绘制等高线以界定等计数区域;然后,找到包含插值点的区域,并为每个点分配其等计数区域的强度。使用此过程,可以对数据进行重采样,从而增加投影数量或每个投影的像素数量。进行了骨闪烁显像的体模研究,以比较有无IPC过程获得的切片与真实图像。还展示了一个临床病例。
体模研究表明,使用IPC过程,当对正弦图进行重采样以使投影数量乘以2或3,且每个投影的像素数量相同时,重建切片在体内外都更接近模型。在临床研究中,条纹伪影减少了,尤其是在体外,尽管只使用了斜坡滤波器。
IPC过程成功减少了条纹伪影。该过程不需要对采集进行任何修改,也不依赖操作员。投影数量的增加可能是减少条纹伪影的必要但非充分条件:如果不进行校正,当投影数量增加时,衰减可能是消除这种伪影的限制因素。