Separation of a Mixture of Simultaneous Dual-Tracer PET Signals: A Data-Driven Approach

In developing a signal separation algorithm, an obvious challenge relates to the ways in which the photon measurements from dual-tracer positron emission tomography (PET) are overlapped. Aside from this issue, the ill-conditioning and the noisy measures are still obstacles to achieve accurate and robust result. In this paper, we develop a novel approach from a data-driven perspective that separates dynamic dual-tracer PET signals into individual-tracer components explicitly with a simultaneous-injection single-scan. Our approach is divided into two phases: training and reconstruction. In the training phase, our proposed framework will learn the potential relationship of different frames for dynamic dual-tracer PET data. In the reconstruction phase, given the sinogram of a dual tracer, the PET activity map of an individual tracer can be obtained directly. Experiments on Monte Carlo simulation data sets are conducted as validation. The experimental results have demonstrated the superior performance of the proposed approach by comparing with the ground truth of the individual-tracer.