Evaluation of scattering models in brain SPECT imaging

Quantification accuracy in SPECT imaging is impaired by a number of factors including noise, attenuation, collimator/detector response and scatter. Recent advances in SPECT technology have considerably improved the possibilities of quantification in nuclear neuroimaging. On the one hand, fan-beam collimators on multihead systems offer a good trade-off between resolution and noise. On the other hand, transmission imaging systems which enable us to obtain attenuation maps are now available. The scattering and the spatially variant fan-beam collimator response are relevant degrading effects in neuro-transmission SPECT studies. Monte Carlo simulation (MC) is the most general method for detailed modelling of scatter although it implies long computation times. The scattering in the collimator was neglected for low energy photons as those of ^99mTc agents. Thus, the scattering models in ^99mTc brain neurotransmission studies were focused on scatter interactions inside the object. The SimSET Monte Carlo code was used and the collimator module included a new probability density function per unit of solid angle for fan-beam collimators. This modelling of the collimator/detector blurring effect accelerated the code in a 2-fold factor. The assessment of the Monte Carlo-based scatter estimate is based on comparison with sinograms which in turn are based on simulated projection data of numerical phantoms. The numerical phantom was implemented by using experimental data obtained from a CT image of an anthropomorphic striatal phantom. The fully 3D reconstruction which included attenuation and PSF corrections was based on composite images of the 8 slices showing the basal ganglia activity, 128×128 pixels each. Good agreement was found between the original and the estimated scatter distributions. Our findings indicate that the proposed methodology is expected to be useful for obtaining accurate scattering distributions in SPECT projections.