Resolution recovery reconstruction in Compton camera consisting of DSSD and 4 CZTs using shift-variant point spread function

Objectives A portable Compton camera based on electronic collimation has an advantage over SPECT and PET for simultaneous multi-tracer imaging in molecular, radiation therapeutic and nuclear medical studies. However, the spatial resolution of the volumetric reconstruction suffers from the measurement uncertainties and is variant over field-of-view (FOV) due to the detection model based on the conical surface integration. In this study, we propose and evaluate the strategy on incorporation of the shift-variant point spread function (SV-PSF) into the system model of resolution recovery reconstruction in Compton camera.

Methods The variant resolutions over the 3D FOV were measured with 35 point sources of 140 keV. The point sources were located 1 cm apart at 5 radial distance from center of yz-plane and 7 axial distance from the Compton camera. The measured resolutions of 35 point sources were fitted into the exponential function of radial (r) and axial (d) distances, f(r,d)=A*exp(Br+Cd). The fitting surface function was used to incorporate Gaussian SV-PSF into the listmode OSEM (LMOSEM) algorithm. The performance of LMOSEM with SV-PSF was compared with shift-invariant PSF (SIV-PSF) using 9 point sources and IEC-like phantom.

Results The coefficients (A, B, C) for fitting surface of SV-PSF were not identical between x-axis (5.8, 0.0032, 0.019) and yz-palne (6.1, 0.0022, 0.013). LMOSEMs with SV-PSF of 9 point sources and IEC-like phantom provided better qualitative and quantitative image quality (x 1.5) than LMOSEMs with SIV-PSF.

Conclusions The LMOSEM with SV-PSF yielded better resolution and contrast recovery over the FOV than without PSF and SIV-PSF. The volumetric and multi-tracer imaging of the Compton camera for molecular and nuclear medical applications may be possible by LMOSEM with SV-PSF with good spatial resolution