List-mode image reconstruction for the high resolution ClearPET™ Neuro system

Conventionally, positron emission tomography data are sorted into sinograms prior to image reconstruction. This leads to a loss of spatial and temporal resolution due to histogramming of events. An alternative data representation is the so-called list-mode format which allows to store more attributes of each individual event, like involved detectors, energy, time, gantry state or even the depth of interaction. The format has the advantage, that it keeps the data in their highest possible temporal and spatial resolution. The additional number of attributes, if considered in the system model, can lead to a significant improvement in image quality. We develop a subsetized list-mode expectation-maximization image reconstruction with a ray-tracing technique to compute an enhanced system model. The algorithm approximates the three-dimensional coincidence response function for two arbitrarily oriented cubic detector elements, incorporated into crystal matrices. The geometry of the involved detector modules is evaluated at event time, tracing the ray paths of gamma quanta from one detector to the other through the field of view. The algorithm has been assessed using list-mode data for the ClearPET™ Neuro, and is compared to a standard ordered-subset expectation-maximization reconstruction method for projections.