Zixiong Cao

Simulation of a pinhole-collimator insert for small animal PET using GATE

In previous work we demonstrated the feasibility to conduct SPECT imaging using a multi-pinhole collimator insert in a small animal PET scanner. A simplified analytical simulation method was used, modeling mainly geometric attributes, to optimize the multi-pinhole configuration to achieve the best trade-off between resolution and sensitivity for specific applications. To more accurately optimize the pinhole configuration more physical attributes need to be modeled, including attenuation and scatter in both the object and collimator. In this work we used GATE, a Monte Carlo simulation platform based on the well-validated Geant4 libraries, to better model our system. New simulations have been performed using the geometry of the MOSAIC (Philips) system and our simple prototype cylindrical insert (lead, 12 pinholes, pinhole diameter 1.0 mm, insert diameter 60 mm). Measurements of sensitivity (0.22%), resolution (FWHM=1.6 mm), and scatter fraction (28%) using point and capillary sources from GATE-simulated data are in good agreement with measured experimental data. Having validated this simulation environment with this baseline test, we are iteratively testing various objects, pinhole configurations, and collimator materials to design our next generation of optimal pinhole configurations for specific imaging tasks.

Determining the lesion detectability index along the axial and radial direction for multi-pinhole SPECT

The improved resolution capabilities of pinhole SPECT makes it an ideal imaging modality to study various human disease models in small animals. However, the sensitivity profile of the sampled image space is non-uniform and depends on the relative location of the pinhole and the imaging geometry used. Hence, it will be desirable to place the target volume in the same plane as the pinhole for better count statistics. In this work we study the lesion detectability index of single pinhole (SP) and multi-pinhole (MP) for lesions placed at various locations in the field-of-view (FOV). A uniform cylinder filled with activity equivalent to those observed in a mouse scan was simulated (14 MBq, 10 minutes study). Lesions (1.4 mm) were placed at predetermined locations and multiple noise realizations were performed. The reconstructed images were evaluated using non-prewhitening (NPWF) analysis to accurately study the lesion detectability in the axial and radial directions. The overall lesion detectability of MP was found to be uniform and about 1.6 times better than that of the SP case. Our studies shows that the improved sensitivity (84 cps/MBq for SP and 612 cps/MBq for MP) and uniform sampling of the image space by the MP results in improved lesion detectability of a larger volume for the same acquisition time.