Reconstruction of small-animal SPECT data acquired with a space-variable-focusing collimator

Objectives In small animal SPECT, the collimator is a key component strongly impacting the trade-off between spatial resolution and sensitivity. Parallel-hole, pinhole and multi-pinhole collimators have been largely investigated. Here, we simulated the first spatially-variable-focusing cone-beam collimator with focal lengths inside the field-of-view (FOV) and increasing from the center to the edge of the collimator and developed a reconstruction algorithm suited for this geometry.

Methods The simulated detector is composed of 131×131 CZT pixels, 0.75x0.75x5.0 mm3 in size. The detector energy resolution is 8% at 140.5 keV with intrinsic spatial resolution of 0.5 mm. The 20 mm thick collimator had 0.15 mm thick tungsten septa. SPECT acquisitions of Tc99m line sources and of a Derenzo phantom were simulated with no background activity. Scatter and attenuation were not corrected for. The system matrix entries were calculated by ray tracing without (noSA) and with modeling (SA) of the solid angle effect. We also modeled the point spread function (PSF) in the image-space (noSA-PSF and SA-PSF). Reconstruction was performed using OSEM with 20 subsets, setting the number of iterations to target the same noise level in all reconstructed images.

Results The detector equipped with the collimator had a sensitivity of 0.011 % with a FOV of ∼70 mm in diameter in the transaxial direction and ∼50 mm axially. Spatial resolutions from the line sources were 1.2, 0.7, 0.5 and 0.5 mm for noSA, SA, noSA-PSF and SA-PSF respectively. The 1.35 mm hot rods of the Derenzo phantom were clearly distinguished in all reconstructions, with rods to inter-rod activity ratios of 4.5, 28, 50 and 47 for noSA, SA, noSA-PSF and SA-PSF respectively.

Conclusions Data acquired using this original collimation can be reconstructed without artifact using an OSEM algorithm. The system matrix model had a significant impact on image quality, with a 58 % improvement in spatial resolution and a tenfold increase in rod-to-background activity ratios with SA-PSF compared to noSA.