Accelerated PET water activation acquisition with signal separation methodology

Positron emission tomography (PET) water activation studies can be considerably accelerated when the time between the successive water bolus injections is reduced. However, when considering shorter interinjection times, as short as 5 min, there will be a contaminating residual PET signal in brain tissues attributed to the previous injections. In this paper a signal separation methodology is proposed to allow shorter (<10 min) interinjection times.

The contaminating signal in the frames of interest is estimated by extrapolating the decaying tail of the previous signal. The method requires the dynamic PET recording of the decaying signals between two injections in order to extrapolate the tail, which can be done post-reconstruction using a weighted least squares method. Several extrapolation functions are investigated, including a quadratic function, a decaying exponential, and a biologically inspired function. The biologically inspired function is based on the one-tissue compartmental model for [(15)O]H2O and makes use of a generating function estimated from the total trues coincidence rate. To evaluate the proposed method and extrapolation functions, one- and two-dimensional simulation studies were performed, using interinjection times as low as 5 min. The resulting corrected images are compared to the conventionally obtained images with an interinjection time of 10 min, which allows the previous signal to almost completely vanish.

Among all considered extrapolation functions the biologically inspired function was found to give the best results. The bias introduced when considering shorter interinjection times (<10 min) could be almost completely removed by subtracting the extrapolated remaining activity from the total measured signal of interest. Compared to the standard method of using longer interinjection times the resulting images have a slightly increased variance. Nonetheless the observed increases are small compared to the total variance and the resulting activation maps were visually very similar.

The simulation results show that the proposed method can significantly reduce the total scan time (e.g., when considering 12 bolus injections, the total scan time can be reduced from 2 to 1 h). Extensive and very realistic simulations were used in this work, paving the way for future in vivo validation of the method.