Improved quantification in 123I cardiac SPECT imaging with deconvolution of septal penetration

OBJECTIVES (123)I is becoming an important radionuclide for cardiac imaging. Multiple, low-abundance, high-energy photons associated with (123)I imaging can cause septal penetration in the collimators and degrade quantification of the (123)I cardiac uptake. This study presents a method for the deconvolution of septal penetration (DSP) for improving quantification in (123)I cardiac single photon emission computed tomography (SPECT). METHODS Distance-dependent point spread functions were measured for low-energy high-resolution collimators on a dual-head SPECT system. The measured point spread functions were used in two-dimensional (2-D) and three-dimensional (3-D) models of the collimator response, respectively. 2-D DSP and 3-D DSP were then developed and implemented using iterative reconstruction. A cardiac torso phantom with an internal calibration source was designed with various heart-to-calibration ratios (HCRs) simulating different levels of a patient's uptake. SPECT acquisitions of the phantom were performed using optimized acquisition and processing parameters for (123)I cardiac SPECT. HCRs were calculated using planar projection and tomographic reconstructions. The paired t-test and regression analysis were used to compare the HCRs given by different calculation methods. RESULTS SPECT produced more accurate HCRs than planar imaging. The slopes of the regression lines for SPECT using filtered back-projection were statistically significantly higher than those for planar imaging (0.2118 +/- 0.0297 vs. 0.0819 +/- 0.0070, P = 0.0001). 2-D DSP and 3-D DSP yielded similar HCRs that were close to the true HCR. The slopes of the regression lines for 2-D DSP and 3-D DSP were 0.9203 +/- 0.0523 and 0.9101 +/- 0.0304, respectively. The DSP HCRs were significantly more accurate than those calculated without DSP (P < 0.0001). CONCLUSION DSP significantly improves quantification in (123)I cardiac SPECT imaging. 2-D DSP with its less computational burden shows promise for implementation in clinical practice so as to allow the use of the widely available low-energy, high-resolution collimators for quantitative I cardiac SPECT imaging.