Improvement of image quality in multispectral PET by energy space smoothing and detector space normalization

By allowing independent data processing in each energy frame, multispectral PET has the potential to improve sensitivity and to support more accurate energy dependent scatter correction in high resolution PET. However, statistical fluctuations associated with the use of multiple energy windows and short acquisition times seriously undermine this potential. In this work, the authors show that this limitation can be overcome without resolution loss, a) by filtering data in the energy space to suppress statistical fluctuations and b) by multispectral normalization of detector efficiency in the spatial domain to eliminate systematic fluctuations. The effectiveness of these corrections was investigated by comparing images acquired in different energy frames with and without energy space filtering. The sharpness and contrast in different energy frames improved significantly. The standard deviation decreased in both the hot regions and background. The FWHM and FWTM evaluated from the images of a line source confirmed that smoothing in the energy space does not degrade image resolution. The work demonstrates that smoothing in the energy space in conjunction with multispectral normalization in the detector space provides adequate data for further processing such as energy-dependent scatter correction