Quantitative cardiac SPECT imaging - A Monte Carlo method for estimating system matrices

Objectives Statistical errors in the elements of the system matrix calculated from a Monte Carlo method cause large speckled noise in human cardiac SPECT images. The challenge is to estimate the system matrix with sufficient precision while excluding excessive computation time.

Methods System matrices modeling scatter and attenuation were estimated for a cardiac SPECT/CT human study applying the Monte Carlo software package SimSET to attenuation maps obtained from the x-ray CT. The system matrices were designed to reconstruct 24 transaxial slices from parallel projections while accounting for scatter and attenuation within as well as scatter spillover from outside the slices. To reduce computation time in estimating the system matrices a multiresolution approach was used which includes coarser resolution in the background outside the heart. The analytical calculation of the collimator response was convolved with the estimates of the system matrix. Smoothing the system matrix was also used to reduce the noise in reconstructed images. Transaxial slices (128x128) through the heart were reconstructed from 60 projections over 180° using the estimated system matrices and 50 iterations of the ML-EM algorithm.

Results A profile through the heart shows better resolution recovery for attenuation and scatter correction than for attenuation correction alone. The resolution is improved even more when scatter spillover is corrected. While better resolution recovery is achieved the contrast also improves. When scatter correction is added to attenuation correction the image intensity decreases which is expected as attenuation correction alone overestimates the image intensity.

Conclusions Monte Carlo estimation of system matrices provides quantitative myocardial perfusion imaging and is possible with advancements in computing hardware and SPECT/CT systems that offer high resolution attenuation distributions.

Research Support NIH R01-EB00121 and R01-EB007219, and U.S. Department of Energy Contract DE-AC02-05CH11231