Dual Collimation Acquisition for High Resolution, Low Noise SPECT

We present Dual Collimation (DC) imaging for SPECT as a technique to improve effective sensitivity with little loss of resolution. For some imaging tasks, DC can overcome the inherent tradeoff between resolution and sensitivity of gamma camera collimators. Simulations were used to investigate the system performance without the expense of collimator fabrication. We have implemented DC for SPECT using simulations of SPECT acquisition using two collimated gamma cameras. Projection data are acquired using two collimators simultaneously: l)a high sensitivity (HiSen) collimator and 2)a high resolution (HiRes) collimator. Both projection data sets are used to reconstruct a single image using an Inverse Monte Carlo (IMOC) technique with Maximum Likelihood Expectation Maximization (MLEM). Accurate modeling of each collimator is included in the IMOC technique and the reconstructed image is constrained to be consistent with both projection sets. Monte Carlo modeling simulated projection acquisition from two phantoms: 1) a resolution-noise and 2) a cold defect contrast phantom. Projections were simulated for acquisitions of 200K counts with a dual gamma camera HiRes system, yielding 1.9M for for an equal time scan using DC (1.8 million counts in the HiSen; 100K in the HiRes). After 300 iterations (10 min), %rms noise decreased for DC (compared to HiRes with equal scan time) by a factor of 7 with an increase in reconstructed line-source width from 9 to 10 mm FWHM. For a 2 cm radius defect, contrast decreased 4% [from 0.90 to 0.86] while contrast-to-noise (CNR) increased 210%. For al cm defect, contrast decreased 40% while CNR increased 200%. Thus for a select class of imaging tasks, Dual Collimation for SPECT can provide significant improvements in signal-to-noise with only a modest sacrifice of system resolution and contrast.