Kubo N
Department of Radiological Technology, College of Medical Technology, Hokkaido University.
Kaku Igaku. 1995 Apr;32(4):395-403.
To improve the quality of single-photon emission computed tomographic (SPECT) images, a restoration filter has been developed. This filter was designed according to practical "least squares filter" theory. It is necessary to know the object power spectrum and the noise power spectrum. The power spectrum is estimated from the power spectrum of a projection, when the high-frequency power spectrum of a projection is adequately approximated as a polynomial exponential expression. A study of the restoration with the filter based on a projection power spectrum was conducted, and compared with that of the "Butterworth" filtering method (cut-off frequency of 0.15 cycles/pixel), and "Wiener" filtering (signal-to-noise power spectrum ratio was a constant). Normalized mean-squared errors (NMSE) of the phantom, two line sources located in a 99mTc filled cylinder, were used. NMSE of the "Butterworth" filter, "Wiener" filter, and filtering based on a power spectrum were 0.77, 0.83, and 0.76 respectively. Clinically, brain SPECT images utilizing this new restoration filter improved the contrast. Thus, this filter may be useful in diagnosis of SPECT images.
为提高单光子发射计算机断层扫描（SPECT）图像的质量，已开发出一种恢复滤波器。该滤波器是根据实用的“最小二乘滤波器”理论设计的。有必要了解目标功率谱和噪声功率谱。当投影的高频功率谱充分近似为多项式指数表达式时，功率谱可从投影的功率谱中估计得出。对基于投影功率谱的滤波器恢复进行了研究，并与“巴特沃斯”滤波方法（截止频率为0.15周期/像素）和“维纳”滤波（信噪功率谱比为常数）进行了比较。使用了位于充满99mTc的圆柱体中的两个线源模型的归一化均方误差（NMSE）。“巴特沃斯”滤波器、“维纳”滤波器以及基于功率谱的滤波的NMSE分别为0.77、0.83和0.76。在临床上，利用这种新的恢复滤波器的脑SPECT图像提高了对比度。因此，该滤波器可能对SPECT图像的诊断有用。
