Normalization of multispectral data in positron emission tomography

Abstract

Data acquisition using more than one energy window has been proposed principally for scatter correction in positron emission tomography (PET). However, prior to scatter correction, such data must be corrected for various distortions including detector efficiency and spectral nonuniformities. The aim of this work is to describe and validate a normalization technique that can minimize these undesirable effects. As with conventional methods using broad windows, the technique reduces the non-uniformity of data acquired in several narrow windows, while performing an additional task unique to multispectral acquisition, i.e. to restore symmetry in data acquired by mirror window pairs of individual detectors. Non-uniformity and asymmetry reductions have been achieved by averaging each pair of detectors over all detectors and averaging each mirror window pair, respectively. The method has been validated using data with different degrees of spatial and spectral distortion, and has been found to be capable of restoring the greatest spectral asymmetric acquisition possible with our animal PET system. Unlike the broad-window imaging, a match of spectral composition of photon fluence between the calibration and measurements is not a requirement. In conclusion, multispectral data acquisition in PET is possible with fairly non-uniform detector response. The success of the technique, however, depends on detectors with very stable characteristics.