Montreal Heart Inst., Que.
IEEE Trans Med Imaging. 1989;8(3):245-50. doi: 10.1109/42.34713.
An approach to image analysis and processing, called holospectral imaging, is proposed for dealing with Compton scattering contamination in nuclear medicine imaging. The method requires that energy information be available for all detected photons. A set of frames (typically 16) representing the spatial distribution at different energies is then formed. The relationship between these energy frames is analyzed, and the original data is transformed into a series of eigenimages and eigenvalues. In this space it is possible to distinguish the specific contribution to the image of both primary and scattered photons and, in addition, noise. Under the hypothesis that the contribution of the primary photons dominates the image structure, a filtering process can be performed to reduce the scattered contamination. The proportion of scattered information removed by the filtering process is evaluated for all images and depends on the level of residual quantum noise, which is estimated from the size of the smaller eigenvalues. Results indicate a slight increase in the statistical noise but also an increase in contrast and greatly improved ability to quantitate the image.
提出了一种称为全谱成像的图像分析和处理方法，用于处理核医学成像中的康普顿散射污染。该方法要求所有检测到的光子都具有能量信息。然后形成一组（通常为 16 个）代表不同能量的空间分布的帧。分析这些能量帧之间的关系，并将原始数据转换为一系列本征图像和特征值。在这个空间中，可以区分原始和散射光子对图像的特定贡献，此外还有噪声。在假设原始光子的贡献主导图像结构的前提下，可以进行滤波处理以减少散射污染。通过滤波处理去除的散射信息的比例针对所有图像进行评估，并且取决于剩余量子噪声的水平，该水平是从小的特征值的大小估计得出的。结果表明，统计噪声略有增加，但对比度增加，并且大大提高了图像定量的能力。
